Executive Summary

BLUF: Italy’s data-centre pipeline has expanded far faster than its credible electricity-demand outlook, grid-development cycle, permitting capacity and public disclosure framework.

By January 2026, Terna recorded 449 data-centre connection applications representing 78.79 GW, up from approximately 30 GW in December 2024; the requested capacity is therefore a project queue, not a realistic consumption forecast. Da rifiuto a risorsa: la nuova sfida della transizione ecologica e digitale – Terna – February 2026.

Even a small realization rate would create material consequences: 10% of the queue equals 7.9 GW, a continuous load comparable to an entire national industrial subsector and requiring approximately 69 TWh annually at full utilization.

Italy is not yet building an equivalent volume of firm, geographically matched, low-carbon electricity, storage, substations and transmission capacity.

The central risk is not a nationwide shortage in annual energy alone, but local scarcity of connection capacity, dispatchable power, transformers, cooling water, land and network resilience, especially around Lombardy and Milan.

France possesses a structural advantage through abundant nuclear generation and surplus low-carbon electricity; Germany is coupling data-centre expansion with a national capacity strategy; the United Kingdom is openly preparing preferential connection and pricing mechanisms for designated AI zones.

Italy’s present architecture remains predominantly reactive: applications arrive first, while network reinforcement, generation, water accounting, ownership transparency and public-interest conditionality follow later.

The most probable 2031 outcome is neither completion of the entire queue nor collapse of the sector, but a selective build-out in which foreign hyperscalers and infrastructure funds secure the most valuable grid nodes while part of the reinforcement cost is socialized.

The strategic question is therefore not whether Italy should host data centres, but which computing workloads justify consuming scarce Italian electricity, water, land and network capacity—and who captures the resulting economic and informational value.

Italy’s AI Power Gap: Data Centres Without the Energy

Italy is racing to host the infrastructure of artificial intelligence before deciding how much electricity, water and industrial capacity it is prepared to allocate to it. The country’s data-centre boom is therefore no longer merely a technology story. It is becoming a test of energy security, industrial policy and economic sovereignty. By January 2026, requests to connect data centres to Italy’s national transmission system had reached a scale that bears little relation to the generating capacity actually under construction. The danger is not that every announced project will materialise. It is that even a fraction will concentrate in already congested areas, forcing steelmakers, railways, hospitals, ports and manufacturers to compete with computing platforms for the same scarce megawatts.

The 78.79-Gigawatt Queue

At the end of 2024, data-cententre developers had submitted requests to Terna for approximately 30 gigawatts of connection capacity. By the end of 2025, the figure had risen to 69.05 GW, spread across 411 applications. At the end of January 2026, Terna counted 449 applications for 78.79 GW. The requested volume had therefore increased by more than 160% in thirteen months. (Terna Blog Energia)

This is not a forecast of actual consumption. Connection queues include alternative sites, phased projects, speculative reservations and developments that may never obtain finance, customers or planning approval. Yet the queue matters because it represents claims on future substations, high-voltage lines, land and network investment.

The arithmetic exposes the strategic scale. If only 5% of the requested capacity became operational and ran continuously, it would correspond to almost 4 GW of load and approximately 34.5 terawatt-hours a year. At 10%, consumption could approach 69 TWh—more than one fifth of Italy’s present annual electricity demand. Even after allowing for lower utilisation, the issue ceases to be marginal.

The International Energy Agency estimates that global data-centre electricity consumption reached approximately 415 TWh in 2024 and could more than double to around 945 TWh by 2030. AI-accelerated servers are expected to account for almost half of the increase. The agency also warns that the local impact is far greater than the global percentage suggests because large centres are geographically concentrated and require unusually reliable power. (IEA)

Milan’s Invisible Congestion

Italy’s difficulty is spatial before it is national. Data centres are not distributed according to where unused electricity is abundant. They cluster around fibre routes, customers, cloud ecosystems, land availability and existing substations. In Italy, that means above all Lombardy and the Milan region, already the centre of the country’s financial, industrial and digital economy.

A new hyperscale campus does not compete with an abstract national electricity surplus. It competes at a specific grid node with urban networks, electrified railways, hospitals, metro systems, manufacturing plants, logistics platforms and new industrial connections. It also demands redundancy: operators commonly require multiple electricity feeds, backup generation and contractual reliability that ordinary industrial consumers may not obtain.

This distinction is crucial. Italy could produce enough additional renewable electricity annually and still face severe bottlenecks during evening peaks, low-wind periods or local network constraints. Solar generation available in southern Italy at midday does not automatically power a server campus near Milan at night. Transmission, storage and dispatchable capacity determine whether contractual “green power” corresponds to electricity physically available when the machines are running.

The IEA explicitly distinguishes between renewable energy purchased contractually and the generation mix physically supplying a data centre through the local grid. It projects that renewables will meet almost half of the worldwide increase in data-centre demand to 2030, but natural gas and coal together could still supply more than 40% of the additional electricity. (IEA)

Italy’s Gas Penalty

The expansion is occurring while Italy remains structurally exposed to expensive electricity. The European Commission’s 2026 Country Report, published on 3 June 2026, found that Italian electricity prices in the first half of 2025 were the fourth highest in the European Union for households and the sixth highest for large companies. Gas set the wholesale electricity price during 61.4% of hours in 2024, while fossil fuels still represented 51% of Italy’s electricity mix in 2025, compared with 29% across the EU. (Economy and Finance)

The industrial consequences are already measurable. Energy-intensive industries account for approximately 3.7% of Italian gross value added, but their production has declined by 12% since 2021. Introducing large, inflexible digital loads into this market without additional generation and network investment risks amplifying the cost disadvantage faced by chemicals, metallurgy, paper, glass, automotive suppliers and advanced manufacturing.

The political question is therefore who finances the infrastructure. If transmission reinforcements, substations and reserve capacity are funded through general network tariffs, part of the cost of hosting global cloud operators may be transferred to Italian households and companies. A data centre that purchases renewable certificates can still increase the system’s requirement for gas-fired generation, balancing services and network expansion.

Italy should consequently require every major project to disclose not only its nominal connection request, but its expected hourly load, commissioning timetable, cooling technology, backup generation, storage capability and contribution to the reinforcement of the grid node it intends to occupy.

Water Behind the Cloud

Electricity is only the most visible resource. Large data centres also consume land and, depending on their cooling architecture, substantial quantities of water. The true impact varies according to climate, server density, cooling technology and the origin of the water used. A facility cooled primarily by air cannot be evaluated in the same way as one using evaporative systems, just as recycled industrial water cannot be treated like drinking water drawn from a stressed municipal supply.

The European Union has begun to close this information gap. The recast Energy Efficiency Directive introduced mandatory monitoring of the energy performance and environmental footprint of significant data centres. Delegated Regulation EU 2024/1364, adopted on 14 March 2024, established common indicators for electricity, renewable energy, water consumption and waste-heat reuse. The Commission required operators to begin reporting to a European database in 2024 and annually thereafter. (Energy)

These rules are a starting point, not an allocation policy. Municipal authorities still need project-level evidence before granting planning permission: the source of cooling water, drought restrictions, discharge conditions, emergency operating procedures and the possibility of feeding recovered heat into nearby residential or industrial networks.

Public incentives should be conditional on measurable efficiency rather than corporate declarations. Power-usage effectiveness, water-usage effectiveness and heat recovery must become contractual obligations, verified after commissioning. A project that fails to meet them should lose preferential treatment.

France’s Nuclear Advantage

The comparison with France illustrates why identical data-centre ambitions produce different risks. In its 2025–2035 Generation Adequacy Outlook, French transmission operator RTE identified approximately 4.3 GW of data-centre projects and estimated that their consumption could rise from around 5 TWh in 2025 to approximately 15 TWh by 2030, assuming 60% of the identified pipeline materialises. (RTE)

France can approach this demand from a position Italy does not possess: abundant low-carbon generation, a large nuclear fleet and an electricity surplus that RTE considers favourable to electrification and reindustrialisation. The French challenge is connecting and locating projects efficiently; the Italian challenge is also producing sufficient firm electricity at competitive prices.

This does not make France immune to congestion or water constraints. It does mean that each additional French data centre is less likely to increase dependence on gas-fired generation. Italy, by contrast, risks importing servers while importing or burning more gas to supply them.

The competitive consequence may become decisive by 2030. France can offer low-carbon electricity as an industrial advantage. Italy may offer strategic geography, fibre connectivity and proximity to Mediterranean cable routes, but unless it expands generation and storage, those advantages will be discounted by higher and more volatile power costs.

Germany Chooses Capacity

Germany has responded by turning data-centre infrastructure into an explicit national strategy. On 18 March 2026, the federal cabinet approved the country’s first national data-centre strategy, targeting at least a doubling of installed capacity to more than 6 GW by 2030 and a fourfold increase in computing power dedicated to artificial intelligence. The programme contains 28 measures covering energy, efficiency, waste heat, connection procedures, land and municipal taxation. (Bundesregierung)

Berlin has also acknowledged the speculative character of connection queues. Applications without credible construction plans are to be deprioritised so that mature projects can obtain grid access more rapidly.

The strategic difference is not that Germany has solved the energy problem. Its electricity system also faces high costs and network congestion. The difference is that the government has defined a desired capacity, a timetable and conditions for integration.

Italy still lacks an equivalent public doctrine. The number of connection applications has become more visible than the number of megawatts the country actually wishes to host. A queue is being mistaken for a strategy.

Britain Allocates Power Politically

The United Kingdom has moved even further towards deliberate allocation. The government opened applications for AI Growth Zones on 30 April 2025, requiring candidate locations to demonstrate access to at least 500 MW of power by 2030, together with suitable land, water and digital connectivity. (GOV.UK)

On 13 November 2025, the Department for Science, Innovation and Technology published a package intended to shorten connection times by as much as five years, mobilise up to £100 billion of additional investment and reserve or reallocate network capacity for projects considered strategically important. It also proposed regional electricity support from April 2027 of up to £24 per MWh in Scotland, £16 in Cumbria and £14 in north-east England, where additional demand could reduce grid-constraint costs. (GOV.UK)

The British approach is controversial because it openly ranks electricity users. But it at least makes the choice visible. Projects are rewarded when they locate where power is abundant or where their demand improves system economics.

Italy currently risks doing the opposite: approving developments where investors prefer to build, then asking the network and industrial system to accommodate them.

The Sovereignty Test

The decisive variable is not the nationality of the building but control of the computing value chain. A data centre may stand on Italian soil while its ultimate owner, cloud platform, servers, software, encryption systems and economic rents remain foreign.

A serious national register should therefore identify beneficial ownership, principal tenants, installed computing capacity, public subsidies, electricity contracts, water sources and the categories of public-sector data hosted. It should distinguish hyperscale cloud facilities from colocation property, sovereign cloud infrastructure and high-performance computing centres serving research or industry.

Physical location alone does not guarantee sovereignty. Control also depends on encryption keys, remote administration, software dependencies, access rights, governing law and the ability of public authorities to ensure continuity during emergencies.

This leads to the most sensitive unanswered question: what happens during rationing? Hospitals, railways, ports, water systems, telecommunications, defence facilities and strategic factories require explicit priority rules. Private data centres must not obtain de facto immunity from curtailment merely because they negotiated superior connection contracts.

The Price of Computation

Italy should not slow digital investment. It should price it correctly. Projects bringing proprietary research, high-value employment, sovereign public computing or industrial AI deserve a different assessment from speculative colocation developments whose principal asset is a reserved grid connection.

By 2031, the most probable outcome is that only a limited share of today’s 78.79 GW queue will be built. Yet that limited share could still transform electricity planning in northern Italy. The risk is therefore not an impossible national blackout caused by every announced campus. It is the gradual transfer of scarce network capacity towards operators whose economic contribution has never been measured against the industrial activity they may displace.

The policy principle should be simple: no major data centre without credible additional power, transparent ownership, verified water sustainability, flexible-load capability and a quantified contribution to the Italian economy. Artificial intelligence may be intangible, but its infrastructure is not. It occupies land, consumes water, loads transmission lines and competes for capital.

Italy must decide whether it is building a national computing economy—or merely leasing its electricity system to one.

U.S.–Saudi Nuclear Deal: Proliferation Risk to 2031

Italy’s AI Electricity Bubble: Who Gets the Next Megawatt?

Italy is preparing to host the factories of artificial intelligence without yet deciding which part of its existing economy must make room for them. The issue is no longer whether data centres create investment, but whether the electricity system can accommodate their expansion while manufacturers electrify production, railways and ports decarbonize, hospitals reinforce resilience and households replace gas with electricity. By January 2026, data-centre developers had submitted 449 connection applications to Terna for 78.79 GW—a volume equivalent to more than half Italy’s entire generating capacity. Most projects will never be built. Yet even an unsuccessful application can reserve engineering attention, influence network planning and delay another investor. Between now and 2031, control over substations, transmission corridors and firm electricity will become a decisive instrument of Italian industrial policy. (Terna Blog Energia)

The Impossible Queue

The Italian data-cententre pipeline has ceased to be a credible forecast and become a market for future electrical optionality. Requested capacity rose from approximately 30 GW at the end of 2024 to 69.05 GW through 411 applications at the end of 2025, before reaching 78.79 GW one month later. The average application was therefore close to 175 MW, the load of a major industrial complex rather than a conventional computing facility. (Terna Blog Energia)

If all 78.79 GW operated continuously, annual consumption would approach 690 TWh. Italy’s total electricity requirement in 2024 was 311.9 TWh, supported by 263.2 TWh of net domestic production, 55.9 TWh of imports and 137.6 GW of gross efficient generating capacity. The full queue is physically impossible. But this is the wrong reassurance. If only 5% of requested power were commissioned and operated at an average 60% load, it would consume roughly 20.7 TWh annually. At 10%, consumption would exceed 41 TWh—enough to become one of the largest new categories of Italian electricity demand. (Terna)

France offers a warning against reading applications literally. RTE reports that data centres connected during the previous two or three years initially consumed, on average, only 20% of the capacity requested, because server deployment is gradual and operators reserve large safety margins. The implication is double-edged: actual demand grows more slowly, but valuable capacity can remain contractually occupied and physically underused for years. (rte-france.com)

The Northern Bottleneck

Italy’s conflict will not occur uniformly across the country. It will concentrate around Milan and other northern nodes where fibre networks, corporate customers, skilled labour and high-voltage infrastructure already converge. At the same time, much of Italy’s future solar and wind generation is planned in the South and on the islands. A power-purchase agreement with a Sicilian or Apulian renewable project does not transport electricity to Lombardy. Transmission lines, substations, storage and balancing resources must still bridge the geographical distance.

This mismatch transforms the data-centre boom into a network question. Italy can possess sufficient annual renewable production and still lack dependable power at a particular substation during an evening peak, a low-wind winter period or a summer heatwave. Its exposure is amplified by limited interconnection. European Commission data place Italy’s electricity interconnection level at 4.68% in 2025, far below the EU’s 15% target. The Commission’s 2026 Country Report also states that Italian electricity prices remained among the Union’s highest: in the first half of 2025 they were the fourth highest for households and the sixth highest for large businesses. Gas set the Italian electricity price in 61.4% of hours during 2024, while fossil fuels still represented 51% of the electricity mix in 2025, compared with 29% across the EU. (Economy and Finance)

A data centre can purchase renewable certificates; a steel mill cannot operate on certificates. Both ultimately depend on the same physical system.

The Industrial Contest

The next megawatt can power an AI-training cluster, an electric-arc furnace, a pharmaceutical process, a railway substation, shore power in a port, a hospital extension or thousands of vehicle chargers. These uses are not economically equivalent, even when they offer the same connection payment.

Hyperscalers possess a structural advantage. They can sign long-duration contracts, finance dedicated substations and provide stronger credit guarantees than many industrial companies. Electricity suppliers therefore have a rational commercial incentive to favor them. Yet the private value of a contract is not the same as its national value. A foreign-owned colocation facility may produce construction expenditure and property revenue but limited permanent employment, while high-margin cloud revenues and intellectual property remain abroad. A steel, chemical or pharmaceutical plant may retain more employment, exports and supply-chain value in Italy but face greater cyclical risk and weaker financing conditions.

The allocation criterion should therefore be value per constrained megawatt, not application chronology. Government should measure domestic value added, employment, avoided imports, fiscal retention, strategic autonomy, demand flexibility and the cost of grid reinforcement. A data centre hosting sovereign public computing and Italian industrial AI should rank differently from one serving foreign entertainment or non-urgent model training. A manufacturer investing in efficient electrification should rank differently from an obsolete plant seeking subsidized power without a credible transition plan.

France’s Surplus

France is better positioned because its data-centre strategy rests on a large low-carbon electricity base dominated by nuclear generation. RTE projects data-centre consumption of 15–20 TWh in 2030 and 23–28 TWh in 2035, when the sector could represent approximately 4% of French demand. RTE argues that this remains manageable relative to a national production surplus of roughly 20%. (rte-france.com)

Paris can therefore pursue digital expansion and reindustrialization without immediately forcing a zero-sum choice at national level. It is identifying locations near the high-voltage network and testing whether data centres can become flexible system resources rather than inflexible consumers. The strategic advantage is not simply cheaper electricity. It is the ability to offer investors a credible physical path from announced capacity to energization.

Italy lacks that margin. It cannot imitate France by treating every major request as an opportunity to absorb surplus generation. It must first determine whether the power exists at the requested location and hour.

Germany’s Filter

Germany has chosen a more explicit industrial doctrine. On 18 March 2026, the federal cabinet approved the country’s first national data-centre strategy. Digital Minister Karsten Wildberger told the Bundestag on 16 April that Germany intends to at least double data-centre capacity to more than 6 GW by 2030 and quadruple AI computing power. The programme contains 28 measures covering energy, sustainability, sites and digital sovereignty. (Bundesregierung)

Its most important provision may be the least spectacular: speculative grid applications without genuine construction intent are to be deferred. Germany has recognized that an undisciplined queue is itself an economic cost. A connection request must correspond to land, financing, planning progress and a credible construction programme.

Berlin’s difficulty is price. Its data centres compete with automotive production, chemicals, steel and the electrification of industrial heat. Germany is therefore attempting to expand computing without allowing globally financed technology companies to outbid its productive base. Italy has not yet published an equivalent national ranking doctrine.

Britain’s Priority

The United Kingdom has moved further by openly prioritizing artificial-intelligence infrastructure. Its April 2026 Compute Roadmap estimates that the country will need at least 6 GW of AI-capable data-centre capacity by 2030, three times the capacity currently available. The government wants several nationally significant sites of at least 500 MW, including one AI Growth Zone exceeding 1 GW by 2030. (GOV.UK)

Applicants must demonstrate access to 500 MW by 2030, sufficient water and discharge capacity, and at least 100 acres of developable land by 2028. The British model does not wait for projects to cluster randomly around London. It designates zones and combines power, planning, land and investment policy. (GOV.UK)

The November 2025 delivery plan goes further: speculative demand will be removed from queues, capacity may be reserved or reallocated to strategic projects, and developers may construct their own high-voltage infrastructure. A 500 MW data centre located where additional demand reduces network constraints could receive electricity-cost support of up to £24/MWh in Scotland, £16/MWh in Cumbria or £14/MWh in north-east England. The government estimates that the package could shorten access to power by five years, save a qualifying 500 MW facility as much as £80 million annually and unlock up to £100 billion of investment. (GOV.UK)

London has made its choice visible. Italy is making one implicitly through the queue.

Europe’s Double Mandate

Brussels is encouraging digital capacity and industrial electrification at the same time. The European Commission aims to triple EU data-centre capacity by 2035. Since 2024, centres with power demand above 500 kW have entered a common reporting framework covering energy, water and sustainability performance. On 27 March 2026, the Commission opened consultation on a Union-wide rating scheme, while minimum performance standards are under preparation. (Energy)

But the same electricity is also expected to decarbonize factories, buildings and transport. On 17 July 2026, the Commission presented an Electrification Action Plan intended to raise electricity’s share of final European energy use from approximately 23% to 46% by 2040 and potentially reduce fossil-fuel imports by €260 billion annually. Electricity, however, still often costs around three times as much as gas, while grid connections can require years. (European Commission)

The European strategy therefore contains a structural tension: Europe wants to become both an AI continent and an “electro-continent,” but has not established a common hierarchy for scarce connections. National governments will decide which ambition receives priority.

The Italian Rule

Italy should create a Strategic Connection Test for every large load above a defined threshold. Approval should depend on five conditions: verified project maturity; additional generation or storage; compatibility with the relevant grid node; measurable demand flexibility; and demonstrable national value.

Capacity reservations should expire unless developers meet staged milestones for land acquisition, permits, financing, equipment orders and actual utilization. Projects creating dedicated network costs should finance them. Data centres should separate essential cloud services from deferrable AI training so that non-critical computing can be reduced during system stress. Manufacturers that invest in energy efficiency should retain part of the released capacity for future electrification instead of seeing it immediately absorbed by another customer. Rail, ports, hospitals and water systems should receive forward network reservations based on approved public investment plans.

The decisive formula is political rather than technological: no project should receive scarce electricity merely because it arrived first or can pay more. It should receive it when the value retained in Italy exceeds the opportunity cost imposed on the rest of the economy.

The Cost of Delay

By 2031, Italy will not have built 78.79 GW of data centres. It may, however, have allocated the most valuable nodes of its future electricity system before deciding what they are worth. That is the real bubble.

France is using surplus low-carbon power. Germany is filtering speculative applications. Britain is creating politically designated AI zones. The European Union is building transparency and efficiency rules. Italy still risks allowing the connection queue to become its industrial strategy.

The country does not have to choose between artificial intelligence and manufacturing. It must choose between unmanaged demand and governed development. Data centres can strengthen Italian productivity, public administration and technological sovereignty—but only when they bring generation, storage, flexibility, fiscal value and strategic computing with them. Otherwise, Italy will not be importing artificial intelligence. It will be exporting electricity, land and industrial optionality to foreign balance sheets.

Iranian Nuclear Trajectory: Power Projection, Proliferation Risks and Allied Enablers (2026-2031 Outlook)

---

Navigational Index

I. The Physical Balance Sheet

Electricity demand, connection queues, transmission bottlenecks, generation adequacy, cooling water, land use and competing industrial loads.

II. The Sovereignty Balance Sheet

Ownership, cloud concentration, public-administration data, contractual control, emergency curtailment, cyber resilience and the geographic location of strategic computing.

III. Italy’s Energy-Operator Balance Sheet: Who Can Power the Data-Centre Expansion?

IV. The 2026–2031 Allocation Conflict

Competition between data centres, manufacturing, transport electrification and public services; comparative strategies in France, Germany, the United Kingdom and the wider European Union.

---

Master Abstract

The Queue Is Not the Grid

Italy is experiencing an extraordinary divergence between the amount of data-centre capacity requested from the electricity system and the amount that can plausibly be constructed, powered and economically utilized by 2031. Terna reported approximately 30 GW of data-centre connection requests at the end of 2024; this increased to more than 50 GW across over 300 initiatives by 30 June 2025, reached 69.05 GW through 411 applications by the end of 2025, and climbed to 78.79 GW through 449 applications by the end of January 2026. Piano di sviluppo della rete – Terna – 2025; Data center e futuro: la sfida energetica che accende l’Italia – Terna – 2025; Da rifiuto a risorsa: la nuova sfida della transizione ecologica e digitale – Terna – February 2026. The increase cannot be interpreted as committed physical demand because connection queues contain overlapping locations, alternative project configurations, speculative reservations, phased developments and projects whose financing, permits, customers or equipment supply remain uncertain. France provides a useful empirical correction: RTE observes that recently connected data centres consume, on average, only around 20% of the power originally requested, partly because operators reserve contingency margins and partly because computing occupancy rises gradually. Les data centers en chiffres clés – RTE – 2026. Even after applying severe attrition, however, the Italian queue remains systemically important. If only 5% became fully utilized, the resulting 3.94 GW continuous load would consume approximately 34.5 TWh per year; a 10% realization would imply 7.88 GW and about 69 TWh; a 15% realization would exceed 103 TWh. These are analytical conversions based on continuous operation and should not be treated as forecasts, yet they reveal the scale mismatch: Italy cannot assess the queue as a normal property-development cycle because a data-centre campus is simultaneously a power-sector investment, a transmission reservation, a water and land claim, a telecommunications node and a long-duration allocation of industrial capacity. Terna itself states that new infrastructure or upgrades will be necessary to connect the requested loads. The first intelligence conclusion is therefore decisive: Italy is not facing 78.79 GW of certain demand, but it is facing 78.79 GW of claims on future optionality, and those claims can influence grid planning, land values, substation priorities and the bargaining power of other industrial consumers before most servers are installed.

The Conflict Is Spatial, Not Merely National

The emerging conflict cannot be understood by comparing annual Italian generation with annual data-centre consumption; the binding constraints are geographic, temporal and contractual. Data centres demand high-quality electricity continuously, require redundant supply routes, tolerate voltage disturbances poorly and often cluster where fibre connectivity, customers, land, cooling infrastructure and existing high-voltage substations already coincide. Terna identifies Lombardy and northern Italy, particularly the Milan area, as the principal concentration of new requests. Data center e futuro: la sfida energetica che accende l’Italia – Terna – 2025. This means that the relevant competition is not an abstract contest between artificial intelligence and national industry, but a local allocation struggle among data centres, steelmaking, chemicals, electrified railways, ports, hospitals, manufacturing clusters, urban distribution systems, battery factories, vehicle charging and heat electrification. The European Commission recognizes that data centres impose substantial energy and climate impacts and can require large quantities of cooling water; it has therefore introduced mandatory energy-performance reporting and a common Union sustainability-rating scheme under the recast Energy Efficiency Directive and Delegated Regulation EU 2024/1364. Energy performance of data centres – European Commission – 2026; Commission adopts EU-wide scheme for rating sustainability of data centres – European Commission – March 2024. Yet reporting does not itself solve allocation. A facility can contract renewable guarantees of origin while remaining physically dependent during constrained hours on the local system’s actual generation, storage, interconnection and dispatchable reserve. The International Energy Agency explicitly distinguishes contractual procurement from the physical electricity mix supplying the facility and projects that data-centre electricity consumption will rise from approximately 485 TWh in 2025 to 950 TWh in 2030, while AI-focused consumption triples. Executive Summary, Key Questions on Energy and AI – International Energy Agency – April 2026; Energy supply for AI – International Energy Agency – April 2025. The relevant Italian disclosure standard must therefore move beyond annual renewable claims and require hourly or sub-hourly physical matching, water-source identification, drought-mode cooling plans, backup-generation inventories, grid-node disclosure, heat-reuse feasibility, curtailment capability and the attribution of reinforcement costs. Without these controls, the country may grant high-reliability electrical capacity to globally mobile computing projects while asking less mobile Italian manufacturers to absorb higher network charges, longer connection delays or reduced investment certainty.

NATO Nuclear Posture Evolution: Expansion Risks and AI Integration 2026-2031 — Strategic Deterrence Outlook

Europe Is Choosing Different Allocation Regimes

The European comparison indicates that Italy’s vulnerability derives not simply from having less electricity than its competitors, but from lacking a visible doctrine governing how strategic computing should enter the energy system. France begins from a materially stronger supply position: RTE describes abundant low-carbon generation, a large electricity surplus and approximately 4.3 GW of data-centre projects expected to materialize within its rapid-decarbonization pathway. Its planning range places data-centre electricity use at 15–20 TWh in 2030 and 23–28 TWh in 2035, with the latter equal to roughly 4% of national consumption. Les bilans prévisionnels – RTE – 2025; Les data centers en chiffres clés – RTE – 2026. Germany, already Europe’s largest data-centre market, adopted a national strategy in 2026 targeting at least a doubling of capacity to more than 6 GW by 2030 and a fourfold increase in AI computing power. Its approach combines grid-connection queue reform, competitive electricity prices, renewable procurement, waste-heat treatment, water-saving cooling and measures against speculative connection applications. Bundesdigitalminister zur Rechenzentrumsstrategie – Bundesregierung – April 2026; Mehr Rechenpower für Deutschland – Bundesregierung – March 2026. The United Kingdom has gone further toward explicit prioritization: its AI Growth Zones framework identifies delayed power access as the largest barrier, proposes removing speculative demand from connection queues, reserving or reallocating capacity to strategic projects, and granting location-dependent electricity-cost support where data centres reduce network constraints. A 500 MW facility could receive support of up to £24/MWh in Scotland, £16/MWh in Cumbria, or £14/MWh in north-east England, subject to implementation from April 2027. Delivering AI Growth Zones – UK Department for Science, Innovation and Technology – November 2025. These models expose Italy’s unresolved choice. It can treat applications on a largely first-come, technically assessed basis; it can prioritize sovereign and economically additive workloads; it can create energy-rich computing zones; or it can require developers to bring generation, storage and grid reinforcement with them. Each regime distributes costs and strategic control differently. The United Kingdom is already preparing to rank projects; Germany is linking expansion to a national capacity objective; France can exploit a favorable generation balance. Italy, by contrast, risks allowing the connection queue itself to become policy.

Ownership Determines Whether the Load Is Strategic

Electricity consumption becomes economically defensible only when the value retained by the host state exceeds the system resources displaced, subsidized or reserved. This requires separating four business models that are frequently combined in public announcements: hyperscale proprietary cloud, colocation real estate, sovereign or regulated cloud, and AI high-performance computing. A foreign-controlled hyperscale campus may bring construction expenditure, property taxes, limited operational employment and improved latency, but it may also repatriate high-margin cloud revenues, import servers and software, lock customers into proprietary platforms and control the contractual architecture through which Italian public and private data are processed. A colocation operator may own only the powered building while tenants control the hardware, data and computing economics. A publicly supported sovereign facility can preserve jurisdictional control but still depend on non-European chips, accelerators, virtualization software and cloud-management layers. The physical location of data in Italy is therefore not equivalent to Italian control over encryption keys, administrative access, orchestration software, remote maintenance, model weights, metadata or legal exposure. The principal analytical question must be reformulated from “How many data centres are being built?” to “Which layers of the computing stack are owned, governed, taxed and interruptible in Italy?” A mandatory national register should identify ultimate beneficial ownership, financing counterparties, tenants, cloud platforms, installed IT capacity, committed electrical capacity, actual peak and average load, public incentives, water source, cooling method, heat-recovery potential, contractual renewable supply, backup fuel, network-reinforcement obligations and the categories of public-sector data hosted. It should also disclose whether a project has obtained priority connection rights, capacity reservations, tax relief, public land, guarantees, subsidized power contracts or resilience classifications. This is not an argument for publishing sensitive security diagrams. It is an argument for giving government a consolidated balance sheet of what it is exchanging. Without that balance sheet, Italy may believe it is importing artificial-intelligence capability when it is principally exporting grid access, land, water, construction rights and legally secure operating space to external balance sheets.

Five Competing Hypotheses for 2031

A structured Analysis of Competing Hypotheses produces five principal explanations for the Italian pipeline. H₁ — Productive Digital Reindustrialization: the queue represents a genuine structural shift in which Italy attracts cloud, AI inference, sovereign workloads and associated digital industries, producing enough productivity, tax revenue and strategic capacity to justify infrastructure expenditure. H₂ — Grid-Option Accumulation: developers are reserving scarce connection capacity and land before obtaining final tenants or financing; much of the queue will never operate, but it can still distort planning and exclude competing investments. H₃ — Foreign Infrastructure Arbitrage: international operators exploit Italy’s location, fibre routes, political incentives and improving renewable pipeline while retaining most intellectual property and platform rents outside the country. H₄ — Energy-System Modernization Catalyst: data-centre demand accelerates transmission, storage, renewable generation, demand flexibility and advanced cooling infrastructure that later benefits the wider economy. H₅ — Industrial Crowding-Out: insufficiently conditioned data-centre expansion consumes high-quality grid capacity and transfers network costs to households and industry, weakening electrification and manufacturing competitiveness. Current evidence favors a mixed posterior rather than one dominant hypothesis. Using an explicitly judgmental Bayesian baseline—not an official forecast—the initial 2026 probabilities are assessed at P(H₁)=0.22, P(H₂)=0.30, P(H₃)=0.24, P(H₄)=0.14 and P(H₅)=0.10. The extraordinarily rapid rise from 30 GW to 78.79 GW, combined with foreign-capital intensity and the absence of equivalent commissioned generation, raises H₂ and H₃. Terna’s acknowledgment that significant infrastructure reinforcement will be required raises H₅, but the high attrition normally observed in connection queues limits its immediate probability. Conversely, European reporting rules, renewable expansion and possible co-location with storage support H₄, although only where reinforcement costs and flexibility duties are contractually internalized. The five hypotheses are not mutually exclusive in physical reality; they function as competing dominant explanations. The most probable composite outcome is that speculative and duplicate applications are progressively filtered, a smaller set of well-financed projects secures premium locations, northern-grid congestion becomes politically salient, and the state introduces project-ranking criteria after—not before—the first allocation conflicts become visible.

Five-Year Outlook: The Bottleneck Moves from Land to Power

The 2026–2031 trajectory can be divided into three phases. During 2026–2027, Italy’s dominant issue will be queue validation: determining which applications possess land control, planning status, financing, anchor tenants, equipment procurement and credible energization schedules. Grid scarcity will increasingly be expressed through connection delays, conditional offers, phased capacity, requests for private substations and pressure for transmission upgrades. During 2028–2029, the focus will shift from announcements to system integration. The projects that survive will compete for transformers, high-voltage equipment, skilled engineering, water authorizations and firm power contracts. Public debate will intensify when annual renewable procurement is shown to differ from hourly physical supply and when municipal authorities compare limited permanent employment with water, land and grid impacts. During 2030–2031, data centres will become part of the security-of-supply architecture. Operators may be required to provide dispatchable demand reduction, batteries, on-site generation, thermal storage or interruptible-load services. The decisive regulatory issue will concern curtailment hierarchy: whether computing loads are interrupted before hospitals, railways, urban networks and strategic manufacturing during system emergencies, or whether private high-reliability contracts effectively shield some facilities. The European Commission expects the Union to triple data-centre capacity by 2035, while the International Energy Agency warns that approximately 20% of planned global projects could face delay unless grid risks are addressed. Energy performance of data centres – European Commission – 2026; Executive Summary, Energy and AI – International Energy Agency – April 2025. A provisional Monte Carlo architecture for Italy therefore requires at least six stochastic variables: project-queue realization, average utilization, power-usage effectiveness, renewable commissioning, transmission-delay duration and industrial-demand recovery. Across conservative parameter ranges, the most decision-relevant risk is not that all 78.79 GW materialize; that outcome is physically implausible by 2031. The material tail risk is that only 5–10% materializes in concentrated northern nodes without commensurate generation, storage and flexible-load obligations. A nationally modest load can become locally destabilizing when it is spatially concentrated and contractually inflexible.

I. The Physical Balance Sheet: Italy’s Data-Centre Power Constraint

The 78.79 GW Queue Is a Claim on the Future, Not a Forecast

The starting point for any rigorous physical assessment is the distinction between requested connection capacity, contracted capacity, energized capacity, average electrical load and productive computing output. At the end of December 2024, requests by data-centre developers for connection to Italy’s national transmission grid amounted to approximately 30 GW. By 30 June 2025, Terna counted more than 300 initiatives exceeding 50 GW; at the end of 2025, the queue had reached 411 applications and 69.05 GW; by the end of January 2026, it had expanded to 449 applications and 78.79 GW. Piano di sviluppo della rete – Terna – 2025; Data center e futuro: la sfida energetica che accende l’Italia – Terna – 2025; Da rifiuto a risorsa: la nuova sfida della transizione ecologica e digitale – Terna – February 2026. The queue therefore increased by approximately 163% in thirteen months, a pace that cannot plausibly represent an equivalent acceleration in secured customers, imported accelerator chips, construction readiness, financing, planning approval and electricity-generation availability. It instead represents a mixture of credible campuses, alternative sites submitted by the same economic interests, phased expansions, projects without final tenants, speculative reservations of substation capacity and applications designed to preserve strategic optionality while developers compare European jurisdictions. The queue should consequently be treated as an electricity-option book: each application seeks a right, or at least a priority position, over a scarce future network node. Even projects that are never completed can affect the system because Terna must study their impact, reserve engineering resources, assess reinforcements and incorporate credible portions of demand into long-term planning. The relevant policy failure would be to confuse queue size with investment realized while simultaneously allowing the queue to influence grid development as though all entries possessed equal maturity.

Italian data-centre connection indicator	December 2024	June 2025	December 2025	January 2026	Intelligence significance
Requested connection power	~30 GW	>50 GW	69.05 GW	78.79 GW	Measures claims on network access, not actual operating demand
Number of initiatives or applications	Not stated in cited summary	>300	411	449	Rapid application growth indicates duplication and optionality risk
Implied average request at January 2026	—	—	—	~175 MW	Equivalent to a very large industrial load for each average application
Increase from December 2024	Baseline	>67%	~130%	~163%	Far faster than grid, generation and permitting construction cycles
Geographic concentration	Northern Italy	Northern Italy	Northern Italy	Lombardy–Milan dominant	Creates local congestion even when national annual energy appears adequate

The average January 2026 request, obtained by dividing 78.79 GW by 449 applications, is approximately 175 MW. That is not the profile of a conventional enterprise server room; it is the scale of a hyperscale campus, a major industrial complex or a cluster of multiple halls developed in phases. The physical significance becomes clearer when converted into annual energy. One gigawatt operating continuously for one year requires approximately 8.76 TWh. If the whole queue operated at full load, it would require roughly 690 TWh annually, more than twice Italy’s current national electricity requirement and therefore physically impossible within the 2031 horizon. Italy’s electricity requirement in 2025 was approximately 311.3 TWh, while the industrial-consumption index ended the year slightly lower overall, despite recovery in steel, engineering, cement and food production during the second half. I come IMCEI – Terna – February 2026. The impossibility of the full queue does not make the issue harmless. A realization rate of only 5% would produce nearly 3.94 GW of connected capacity. At 55% average utilization, this would consume approximately 19 TWh per year; a 10% realization at 70% utilization would consume approximately 48 TWh. The crucial analytical problem is therefore not whether 78.79 GW will materialize—it will not—but whether a smaller, geographically concentrated fraction will emerge faster than local transmission reinforcement, dispatchable generation, storage and industrial electrification. A national system can appear adequate in aggregate while individual 380 kV nodes, transformer banks and metropolitan corridors become effectively saturated.

Analytical realization case	Share of January 2026 queue built	Connected facility power	Assumed average utilization	Indicative annual demand	Interpretation
Screening case	3%	2.36 GW	40%	8.3 TWh	Material but absorbable with geographically disciplined siting
Central case	5%	3.94 GW	55%	19.0 TWh	Comparable to a major new national demand category
Accelerated case	7%	5.52 GW	65%	31.4 TWh	Requires major generation, storage and transmission coordination
Stress case	10%	7.88 GW	70%	48.3 TWh	Creates direct allocation conflict with industrial electrification
Upper-tail case	15%	11.82 GW	75%	77.7 TWh	Incompatible with business-as-usual grid development

Values are analytical conversions, not official Terna forecasts. Annual demand equals connected power multiplied by average utilization and 8,760 hours.

Connection Capacity and Electricity Consumption Are Different Assets

International operational evidence provides a critical correction to the Italian queue. RTE, the French transmission-system operator, reported that France had approximately 18 GW of signed data-centre connection requests across around 80 projects by May 2026, up from approximately 5 GW at the end of 2024. Yet data centres connected to the French transmission grid during the preceding two or three years were using, on average, only around 20–22% of the power they had requested. Operators indicated that reaching approximately 80% consumption could take ten to fifteen years because expensive processors and server capacity are installed gradually as customers and workloads emerge. Les data centers en chiffres clés – RTE – 2026; Bilan électrique 2025: Consommation – RTE – 2026. This distinction has two opposing consequences for Italy. First, it reduces the probability that the entire requested capacity becomes simultaneous peak demand. Second, it increases the probability that scarce network capacity will remain underutilized for long periods after being contractually reserved. A campus that obtains a 200 MW connection but initially consumes only 40 MW may prevent another industrial customer from accessing capacity that is technically reserved but commercially dormant. The network therefore faces a capacity-hoarding problem, not merely an energy-consumption problem. Conventional first-come procedures can reward the earliest application rather than the project delivering the greatest national value per megawatt. France has responded by identifying fast-track sites close to the 400 kV network and by preparing mechanisms to prioritize mature projects. Germany’s 2026 national data-centre strategy explicitly calls for speculative applications without genuine construction intent to be set aside. Bundesdigitalminister zur Rechenzentrumsstrategie – Bundesregierung – April 2026. The United Kingdom has gone further, proposing reservation and reallocation mechanisms through which freed capacity can be reassigned to projects classified as strategically important, including AI growth zones, industrial electrification and transport infrastructure. Accelerating electricity network connections for strategic demand – UK Government – March 2026. Italy’s absence of a similarly visible maturity-ranking doctrine creates a strategic asymmetry: private developers can optimize applications across multiple sites, while the state lacks an equivalent mechanism to optimize scarce capacity across sectors.

The Bottleneck Will Emerge at Specific Nodes, Not Across Italy Uniformly

Italy’s data-centre problem is geographically concentrated because digital infrastructure does not select locations on the basis of national generation averages. It selects sites according to the simultaneous availability of high-capacity electricity, redundant fibre, proximity to corporate clients, international cable routes, secure land, permitting certainty, water or alternative cooling resources and access to skilled maintenance personnel. Terna identifies Lombardy, particularly the wider Milan area, as the principal concentration of new requests. Data center e futuro: la sfida energetica che accende l’Italia – Terna – 2025. The northern system already carries the densest concentration of manufacturing, rail traffic, logistics, hospitals, electric mobility, commercial property and cross-border power flows. A new hyperscale cluster therefore does not compete against an abstract national consumer; it competes against other projects requiring access to the same high-voltage substations, transmission corridors and distribution interfaces. The physical constraint may appear in at least five forms: insufficient transformation capacity from 380 kV to 220 or 132 kV; limited thermal capacity on existing lines; short-circuit and voltage-stability limits; insufficient redundancy under an N−1 contingency; or delays in authorizing new overhead lines and substations. Terna’s 2025 Development Plan provides for more than €23 billion of network investment during 2025–2034, an increase in internal transport capability from approximately 16 GW to 39 GW, around 7–8 GW of additional exchange capacity in major corridors and a roughly 40% expansion of cross-border capacity. The plan also assumes integration of more than 65 GW of additional renewable capacity by 2030 and a requirement for approximately 71.5 GWh of new storage, excluding existing pumped hydro. Piano di sviluppo della rete – Terna – 2025. These investments are substantial, but they were not designed exclusively for data centres. They must also integrate geographically remote solar and wind generation, reinforce the islands, support industrial decarbonization, electrify ports, increase resilience to climate hazards and replace ageing equipment. Every megawatt allocated to digital demand must therefore be assessed against a portfolio of pre-existing national obligations.

Physical bottleneck	Observable indicator	Data-centre effect	Competing public or industrial requirement	Priority intelligence collection
380/220/132 kV transformation	Available MVA at relevant substations	Delayed or phased energization	Steel, rail, automotive and urban growth	Substation-level connection queue and reinforcement date
Transmission thermal limits	N−1 transfer capability	Connection conditional on new lines	North–south power transfer and renewable integration	Contingency studies and zonal congestion data
Reactive power and voltage control	Voltage-stability margin	Need for dedicated compensation equipment	Metropolitan quality of supply	Technical connection conditions
Short-circuit capacity	Fault-level margin	Additional switchgear and protection requirements	Existing industrial users	Node-specific fault studies
Distribution interface	Available high-voltage feeder capacity	Limits smaller urban facilities	Hospitals, transit and commercial electrification	DSO queue and transformer utilization
Equipment supply chain	Transformer and cable delivery time	Multi-year construction slippage	Entire energy transition	Procurement contracts and manufacturer lead times
Permitting	Authorization duration	Energization delayed despite completed building	Grid reinforcement nationally	Procedure status, appeals and land acquisition

Generation Adequacy Is a Question of Hourly Firmness, Not Annual Renewable Certificates

Italy can add large quantities of solar and wind generation while still failing to provide the hourly firmness required by high-availability computing loads. A hyperscale data centre can purchase annual guarantees of origin or sign a power-purchase agreement for electricity generated in another region, but the physical facility continues to draw from the local interconnected system every second. During evening peaks, prolonged low-wind conditions, summer heat events or network contingencies, its actual supply comes from the generation, storage, imports and reserve capacity available to the Italian system at that time. Terna’s development scenario anticipates more than 65 GW of additional renewables by 2030, alongside 71.5 GWh of storage excluding existing pumped hydro. Piano di sviluppo della rete – Terna – 2025. Those numbers are essential but not interchangeable: gigawatts measure instantaneous power, while gigawatt-hours measure stored energy duration. For illustration, 71.5 GWh could theoretically support a 7.9 GW stress-case data-centre load for roughly nine hours before accounting for losses, reserve requirements and competing consumers. It cannot cover a multi-day low-renewable period on its own. Italy also remains structurally dependent on electricity imports and gas-fired generation. Terna’s 2024 statistical summary reported electricity demand of approximately 311.9 TWh, net generation of 263.2 TWh, imports of 55.9 TWh and exports of 4.9 TWh. Pubblicazioni statistiche – Terna – 2025. The exact yearly balance will change, but the structural conclusion does not: digital-load expansion increases exposure not only to domestic generation delays but also to French nuclear availability, Alpine hydrology, Swiss and Austrian network conditions, Balkan interconnection performance, international gas prices and geopolitical disruption of fuel supply. A contract labelled “100% renewable” does not by itself remove these physical dependencies. Italy therefore needs a 24/7 energy-matching test, not merely annual certificate accounting, for very large campuses seeking privileged grid access or public incentives.

The generation challenge becomes more severe when the quality of load is considered. A traditional industrial plant may shut furnaces for maintenance, reschedule some production or respond to market prices, although many processes also require continuity. Cloud and AI operators generally design around stringent uptime commitments and may regard demand interruption as commercially unacceptable. Their facilities therefore include uninterruptible power supplies, batteries, diesel generators and multiple electrical feeds. Yet private backup does not automatically make them system-flexible. Batteries may be reserved for internal ride-through rather than offered to the grid, and backup generators may face fuel-duration, air-quality or environmental constraints. European regulation has begun to recognize this distinction. Commission Delegated Regulation EU 2024/1364 requires qualifying operators to report total energy consumption, installed IT power, backup-generator energy, renewable-energy use, battery capacity offered to the grid, grid-support functions, total water input, potable-water use and heat reuse. It also establishes PUE, WUE, ERF and REF as common sustainability indicators. Commission Delegated Regulation EU 2024/1364 – European Commission – March 2024. Italy should extend this reporting architecture into connection regulation by distinguishing between a rigid 200 MW request and a 200 MW campus capable of reducing 40 MW within seconds, operating batteries in balancing markets, shifting non-urgent training workloads and accepting controlled interruption under predefined emergency conditions. Two facilities with equal peak power can impose radically different system costs. Connection contracts that ignore this difference effectively reward inflexibility.

Cooling Water Is a Local Constraint Hidden by National Averages

Water risk must be analyzed at the catchment and municipal-network level rather than through Italy’s aggregate renewable-water resource. ISPRA estimated that Italy’s renewable water availability in 2025 was approximately 128 billion cubic metres, more than 7% below the long-term average, around 4% below the 1991–2020 climate average and approximately 19% lower than in 2024. ISPRA also confirmed a negative national trend in renewable-water availability extending from 1951 to 2025, while the Centre and South continued to experience moderate to high water severity, with Sicily remaining under high severity throughout 2025. Nel 2025 risorse idriche in calo rispetto al 2024 – ISPRA – March 2026; Mari sempre più caldi: nel 2025 secondo valore più alto mai registrato – ISPRA – July 2026. Data-centre water use varies enormously by cooling architecture, climate, operating temperature, workload density and whether the measurement includes only direct on-site consumption or also the water consumed in electricity generation. Evaporative systems can reduce electricity used for mechanical cooling but increase direct water withdrawal; dry cooling reduces on-site water demand but can consume more electricity and lose efficiency during extreme heat. Liquid cooling is increasingly necessary for high-density AI racks, but it does not eliminate heat rejection: the heat must ultimately be transferred to air, water, a district-heating system or another sink. National averages therefore conceal the real conflict. A campus located in water-abundant northern territory may still compete with municipal supply, agriculture, industry and ecological flows during a drought; a coastal facility may use seawater but confront corrosion, discharge-temperature and marine-permitting constraints; a facility relying on drinking water imposes a different public cost from one using reclaimed wastewater. The EU database publishes water data only in aggregated form at Member-State and Union level, partly to protect commercial confidentiality. That architecture is valuable for benchmarking but insufficient for local permitting because the environmental impact occurs at the individual basin and utility level.

Cooling configuration	Electricity implication	Direct water implication	Principal physical vulnerability	Italian permitting test
Air-cooled mechanical chillers	Higher electricity in hot weather	Low direct process water	Summer peak demand and heat derating	Peak-temperature efficiency and emergency capacity
Evaporative cooling	Lower cooling electricity under suitable conditions	Potentially high water consumption	Drought restrictions and potable-water competition	Source, seasonal limit and drought curtailment plan
Hybrid cooling	Variable electricity–water trade-off	Moderate and controllable	Operational optimization failure	Mandatory seasonal operating envelope
Direct-to-chip liquid cooling	Reduces internal fan and cooling losses	Depends on external heat rejection	High-density heat concentration	Full system boundary, not rack-level claims
Immersion cooling	Potentially low PUE	Low direct water within IT loop	Fluid management and external heat rejection	Lifecycle and waste-fluid controls
Reclaimed-water cooling	Moderate electricity	Avoids potable-water use	Availability and treatment reliability	Binding supply agreement and quality standard
Seawater or river cooling	Potentially efficient	Very high withdrawal but limited consumption	Thermal discharge and ecosystem impact	Intake, discharge and climate-adaptation assessment
District-heat integration	May improve total energy utilization	System-dependent	Lack of nearby heat demand and seasonal mismatch	Verified heat off-taker before credit is granted

China’s regulatory architecture illustrates a more interventionist physical-planning doctrine. The National Development and Reform Commission, together with other agencies, adopted a dedicated green and low-carbon data-centre action plan in July 2024. National policy links computing location to power-system geography, promotes direct green-power supply and source-grid-load-storage coordination, and requires new large and ultra-large facilities to reach PUE of 1.25 or lower, with national hub projects not exceeding 1.2. China reported average PUE of approximately 1.3 across its eight national computing hubs, with leading facilities reaching considerably lower values. Data Center Green and Low-Carbon Development Special Action Plan – National Development and Reform Commission – July 2024; Response to National People’s Congress Recommendation No. 4677 – National Development and Reform Commission – November 2025. Beijing has gone further by applying progressively stricter PUE ceilings to larger projects and by encouraging annual renewable-energy shares to rise toward 100% for newer facilities by 2030. Provisions on Strengthening Energy-Saving Reviews of Data-Centre Projects – Beijing Municipal Development and Reform Commission – July 2023. The Chinese model cannot be transplanted mechanically into Italy, but it demonstrates the strategic alternative: computing demand can be directed toward energy-rich regions instead of allowing fibre proximity and real-estate economics alone to determine location. Italy’s equivalent would require linking major campuses to renewable zones, storage, grid reinforcement and reclaimed-water infrastructure before connection priority is granted.

Land Use Extends Far Beyond the Server Building

A data-centre proposal is frequently presented to municipalities as a building-development question, yet the actual territorial footprint includes the campus, security perimeter, internal roads, substations, transformers, backup generation, cooling equipment, fuel storage, water infrastructure, fibre routes and potentially new transmission lines. The most visible warehouse-like structure may therefore represent only part of the land consumed or sterilized. Power density varies considerably, but very large campuses seeking hundreds of megawatts can require tens of hectares before external infrastructure is counted. Their land impact is magnified when the site occupies areas adjacent to existing 220 or 380 kV substations, because those parcels possess scarcity value for industrial plants, storage projects, renewable interconnections and future grid expansion. A poorly designed planning system can permit the private capture of this locational scarcity without measuring its opportunity cost. The United Kingdom’s AI Growth Zone criteria provide a useful benchmark because applicants must demonstrate access to at least 500 MW by 2030, identify water availability and discharge arrangements, provide land-area information, address planning and connectivity, and show either secured network capacity or a credible behind-the-meter solution. AI Growth Zones: open for applications – UK Government – June 2025. The UK framework also favors deindustrialized areas, locations near low-carbon generation and storage, and sites where computing investment supports a wider innovation ecosystem rather than operating as an isolated power-consuming property asset. AI Growth Zones: expression of interest – UK Government – June 2025. Italy should apply an equivalent land-to-compute additionality test: how much permanent employment, domestic procurement, taxable value, sovereign computing, research capacity and industrial productivity is obtained per hectare and per megawatt? A project offering only colocation space and imported hardware should not receive the same planning priority as one integrating Italian research institutions, domestic cloud services, advanced manufacturing applications and recoverable heat.

Territorial asset consumed	Typical hidden extension beyond campus	Alternative claimant	Required disclosure
Industrial land	Security buffers, roads, cooling yards	Manufacturing, logistics, laboratories	Total sealed and reserved surface
Grid-adjacent land	Dedicated substation and line corridors	Storage, renewable connection, factories	Full electrical footprint
Water infrastructure	Intake, treatment and discharge works	Municipal supply and agriculture	Maximum daily and drought-period withdrawal
Fibre corridors	Redundant routes and landing connectivity	Telecom and public networks	Resilience architecture without sensitive detail
Air-quality envelope	Backup generators and testing	Existing industrial permits	Annual operating and emergency fuel profile
Heat sink	Cooling towers or dry coolers	Urban development and ecosystems	Maximum thermal rejection
Municipal infrastructure	Roads, fire services and security	Other commercial developments	Public-service cost allocation

Competing Loads Convert the Digital Expansion into an Industrial-Policy Choice

The physical conflict becomes explicit when data-centre demand is compared with other electricity uses that Italy is simultaneously attempting to expand. Steel producers require power for electric-arc furnaces and potentially hydrogen-based processes; chemical producers require electricity for heat, electrolysis and feedstock substitution; ports require cold ironing for ships at berth; railway systems require traction power and resilient substations; automotive electrification requires charging networks and battery-related manufacturing; hospitals require high-reliability supply; households and commercial buildings are moving toward heat pumps; and renewable-energy integration requires storage, converter stations and reinforced interregional transfer. Terna’s data show that the national requirement in 2025 was 311.3 TWh and that steel, engineering, cement and food were among the industrial segments expanding late in the year. I come IMCEI – Terna – February 2026. Data centres differ from these loads because their direct permanent employment per megawatt is often limited after construction, while their economic value is concentrated in software, cloud services, data exploitation and intellectual property that may be booked outside Italy. This does not mean that manufacturing should automatically outrank computing. High-performance computing, sovereign cloud, defence analytics, pharmaceutical research and industrial AI can generate major national value. It means that megawatts must be allocated according to workload additionality, not according to the generic label “digital infrastructure.” France’s approach is revealing: RTE expects data-centre consumption to rise toward 15–20 TWh in 2030 and 23–28 TWh in 2035, but France begins from a position of abundant low-carbon generation and a substantial electricity surplus. It has identified fast-track sites close to the 400 kV network and estimates that approximately 4.3 GW of data-centre projects should materialize in a rapid-decarbonization pathway, alongside industrial and hydrogen projects. Bilan prévisionnel 2025–2035 – RTE – 2025; Les data centers en chiffres clés – RTE – 2026. Italy cannot assume the same margin. Its ranking mechanism must therefore compare the economic and strategic return of each incremental megawatt under scarcity.

Competing load	Reliability requirement	Flexibility potential	Domestic value retention	Consequence of delayed connection
Hospital and health infrastructure	Extremely high	Very low	High public value	Direct safety and service risk
Rail traction and metropolitan transit	High	Limited	High national productivity	Transport bottlenecks and emissions
Steel and advanced metallurgy	High, process-specific	Medium with planning	High if domestic production retained	Deindustrialization and import dependence
Chemicals and pharmaceuticals	High	Process-dependent	High strategic value	Supply-chain and health-security exposure
Ports and cold ironing	Scheduled but high	Medium	Logistics and environmental value	Continued local pollution and fuel use
Automotive charging and batteries	Distributed	Potentially high	Depends on domestic industry	Slower transport electrification
Data-centre colocation	High	Technically medium, commercially uncertain	Variable and often externally captured	Lower digital capacity
Sovereign HPC and public cloud	High	Workload-dependent	Potentially very high	Strategic and administrative dependence
Residential heat electrification	Weather-sensitive	Medium with thermal storage	Broad social value	Continued gas dependence

Bayesian Assessment and Five-Year Physical Outlook

The evidence supports five competing physical hypotheses. H₁ — Queue Dissipation assumes most applications are speculative or duplicated and fewer than 3% of requested megawatts become operational by 2031. H₂ — Managed Integration assumes 3–6% is built through geographically diversified projects accompanied by grid reinforcement, storage, efficient cooling and flexible-load obligations. H₃ — Northern Congestion assumes 5–8% materializes primarily around Lombardy and other established hubs, creating local connection scarcity despite manageable national annual consumption. H₄ — Accelerated Hyperscale Build-Out assumes 8–12% becomes operational as foreign capital and AI demand advance faster than expected, requiring emergency network prioritization and increased gas or import dependence. H₅ — Infrastructure Delay assumes credible projects exist but transformers, substations, permits, generation and water authorizations delay energization beyond 2031. A judgmental Bayesian update based on the surge from 30 GW to 78.79 GW, French evidence of 20–22% initial utilization, Terna’s explicit warning that new infrastructure will be required, and European moves to purge speculative queues produces the following posterior assessment: H₁ 17%, H₂ 28%, H₃ 31%, H₄ 9%, H₅ 15%. These values are analytical judgments, not official probabilities. The most likely outcome is therefore neither harmless queue collapse nor full-scale electricity crisis, but concentrated northern congestion combined with selective project realization. During 2026, the key indicators will be application attrition, signed connection solutions and the appearance of maturity requirements. During 2027, procurement lead times for high-voltage equipment and the distribution of projects outside Lombardy will become decisive. During 2028, the first conflict between data-centre energization and industrial-electrification schedules is likely to become visible at specific nodes. During 2029, water permits, summer cooling demand and contractual flexibility will move from environmental side issues to operating constraints. By 2030–2031, Italy will either have created a transparent system for ranking loads and allocating reinforcement costs or will be forced to manage scarcity through ad hoc political intervention.

Hypothesis	Description	Posterior probability	Evidence that would increase probability	Evidence that would reduce probability
H₁	Most applications disappear	17%	High withdrawal rate; few land or finance milestones	Rapid signing of binding connection agreements
H₂	Controlled, diversified integration	28%	New location rules; storage and flexibility obligations	Continued Lombardy concentration without reforms
H₃	Localized northern congestion	31%	Substation delays; industrial projects displaced	Fast geographic diversification and network delivery
H₄	Accelerated hyperscale surge	9%	Large anchor tenants; AI hardware deliveries; self-built grid	Capital constraints or chip-supply slowdown
H₅	Projects delayed by infrastructure	15%	Transformer shortages; permitting appeals; water limits	Fast-track authorizations and standardized designs

A Monte Carlo framework should model at least eight uncertain variables: queue realization rate, annual commissioning profile, average utilization, facility PUE, hourly load flexibility, renewable and storage delivery, transmission-delay duration and industrial-demand recovery. A robust simulation would not use a single national annual-energy ceiling; it would divide Italy into electrical zones and major substation clusters, then evaluate hourly adequacy under normal and contingency conditions. The output should include probability distributions for unserved connection demand, congestion cost, reinforcement expenditure, gas-fired generation, imports, water withdrawal and industrial displacement. A central distribution using realization between 3% and 7%, utilization between 40% and 65%, and PUE between 1.15 and 1.35 would place 2031 data-centre electricity consumption broadly between 10 and 35 TWh, with a wider stress tail extending above 45 TWh. The principal systemic hazard is not the median energy value; it is correlation. High summer temperatures simultaneously increase cooling demand, reduce the efficiency of thermal equipment, stress water availability, elevate household air-conditioning loads and potentially constrain hydroelectric output. Low-wind winter periods can similarly coincide with heating electrification and reduced renewable generation. Data-centre campuses designed as constant loads intensify both conditions unless they can shift computing, reduce non-critical workloads, discharge batteries or activate controlled backup. The physical-balance model must therefore price coincident peak risk, not average annual consumption.

The Required Italian Physical-Control Architecture

Italy requires a national authorization and connection architecture that treats every large data centre as an integrated energy, water and territorial project. The first control should be a maturity gate: no application above a defined threshold should retain scarce network priority without verified land rights, planning progress, financial capacity, anchor demand and a schedule for IT deployment. The second should be a capacity-use obligation: projects that fail to reach staged utilization milestones should release unused megawatts or pay escalating reservation charges. The third should be locational differentiation: connection cost and timing should reflect whether a project aggravates or relieves congestion, with incentives for sites near surplus generation, storage and underused grid assets. The fourth should be physical energy matching: annual renewable certificates should be supplemented by disclosure of hourly supply, residual grid mix, storage and firm backup. The fifth should be water conditionality: potable water should be a last-resort cooling source, and each project should disclose maximum withdrawal, consumption, discharge, drought operation and reclaimed-water alternatives. The sixth should be industrial-impact assessment: every campus above 100 MW should be evaluated against competing connection requests in manufacturing, transport, hospitals, ports and urban infrastructure. The seventh should be emergency hierarchy: contracts must specify whether and how the campus can be curtailed during adequacy events, rather than allowing private service-level agreements to determine de facto national priority. The eighth should be cost causation: dedicated substations, network reinforcements and resilience upgrades attributable to the project should not be automatically socialized across all users. The EU reporting system already provides the measurement vocabulary—installed IT power, total energy, PUE, water, renewable factor, heat reuse, backup generation and grid-support services. Commission Delegated Regulation EU 2024/1364 – European Commission – March 2024. Italy’s missing step is to convert those indicators from retrospective sustainability reporting into prospective conditions for access to scarce physical infrastructure.

The strategic conclusion of the physical balance sheet is that Italy is not merely adding a new category of electricity consumer. It is assigning long-duration rights over the most valuable nodes of its future electrified economy. The 78.79 GW queue represents a warning signal because its scale exceeds any credible short-term development pathway, but the greater danger lies in the smaller subset that will actually be built. A 20–30 TWh data-centre sector could be integrated nationally, yet still inflict disproportionate local costs if concentrated around Milan, granted rigid supply rights, cooled with scarce potable water and energized before corresponding network or generation investments are commissioned. Conversely, the same load could strengthen the system if located near renewable surpluses, equipped with storage, capable of shifting workloads, connected through developer-funded infrastructure and tied to sovereign or industrially productive computing. The physical outcome will not be determined by PUE alone. A highly efficient facility can remain strategically harmful if it occupies the wrong grid node, uses the wrong water source, retains unused capacity or displaces higher-value electrification. By 2031, the decisive metric should therefore be national value delivered per constrained megawatt, adjusted for location, hourly firmness, water, land, flexibility and ownership. Without that metric, Italy will build data-centre shells while importing the chips, software, capital and platform rents, effectively selling scarce electrical capacity and territorial optionality. With it, computing infrastructure can become a controlled instrument of reindustrialization rather than an electricity bubble whose network costs are distributed across households and legacy industry.

II. The Sovereignty Balance Sheet: Who Controls Italy’s Strategic Computing?

Sovereignty Is a Stack of Enforceable Controls, Not the Address of a Server

The geographic presence of a data centre on Italian territory does not, by itself, establish Italian digital sovereignty. Sovereignty exists only when the state, the public administration or the regulated Italian customer retains effective authority across the entire operational stack: beneficial ownership of the infrastructure; jurisdiction over the provider and its subcontractors; control of encryption keys and privileged identities; knowledge of where primary data, replicas, backups, logs and metadata are processed; contractual rights to inspect and audit; technical capacity to move workloads; access to source code, documentation and skilled personnel; continuity rights during insolvency or geopolitical disruption; and the legal authority to determine whether a strategic workload must remain available, be isolated or be curtailed. A server physically installed in Milan can remain dependent on a foreign control plane, proprietary identity service, remote support team, non-European software licence, overseas security-operation centre and corporate parent subject to another jurisdiction. Conversely, a workload operated by a foreign technology vendor can provide a higher degree of Italian control if the architecture places cryptographic keys, operational personnel, administrative authority, audit rights and emergency command inside a legally insulated Italian structure. The Cloud Italia Strategy explicitly associates public-sector cloud adoption with autonomy, interoperability, reversibility, protection of data and reduction of vendor lock-in; it also acknowledges that the objective is not merely modernization but the creation of a national operating system for public services. The strategy classified public data and services as strategic, critical or ordinary, according to the damage their compromise could cause to national security, essential services, health, safety or socioeconomic welfare. It further set the target that, by 2026, 75% of digital public services should operate on secure cloud infrastructures and all strategic public-sector data and services should be hosted on infrastructures capable of ensuring strategic and decision-making autonomy. La Strategia Cloud Italia – Dipartimento per la trasformazione digitale and Agenzia per la cybersicurezza nazionale – September 2021; Classificazione di dati e servizi – Cloud Italia – January 2022; Le misure del Piano Nazionale di Ripresa e Resilienza – Cloud Italia – 2026.

Sovereignty layer	Superficial indicator	Effective control test	Principal Italian vulnerability
Physical location	Servers located in Italy	Are every production copy, backup, log and recovery image located in declared jurisdictions?	Replication or support processes may cross borders
Corporate ownership	Italian-incorporated operator	Who is the ultimate beneficial owner and who appoints management?	Italian legal entity may remain foreign-controlled
Legal jurisdiction	Contract governed by Italian law	Can a foreign parent or authority compel disclosure, suspension or technical action?	Extraterritorial legal exposure
Encryption	Data encrypted at rest	Who generates, stores, rotates and can override the keys?	Provider-controlled key management
Administrative authority	Italian customer owns account	Who controls root identities, hypervisor access and emergency credentials?	Privileged access retained by provider
Software sovereignty	Workload hosted domestically	Can the workload operate without proprietary foreign control planes or licences?	Licence termination and platform dependency
Personnel	Local operations team	Are critical engineering, incident-response and recovery skills available in Italy?	Remote dependency on external experts
Contractual reversibility	Exit clause exists	Has migration been technically tested within a defined time and data volume?	Paper exit rights without executable migration
Emergency command	National infrastructure status	Who decides service priority, isolation, shutdown or restoration?	Conflict between private SLA and public necessity
Economic sovereignty	Italian data-centre investment	Where are cloud margins, IP income and tax value booked?	Physical costs retained domestically, digital rents externalized

The Polo Strategico Nazionale Solves Part of the Problem but Does Not Eliminate Layered Dependency

The Polo Strategico Nazionale, or PSN, is Italy’s principal institutional response to the vulnerability created by fragmented public-administration computing. The government describes it as a high-reliability infrastructure designed to host strategic and critical public data and services, consolidate central-administration data centres and provide resilience, scalability, interoperability and security. Its four declared data-centre locations are Acilia and Pomezia in Lazio and Rozzano and Santo Stefano Ticino in Lombardy, a geographic pairing intended to provide operational continuity and fault tolerance across distinct sites. The ownership structure, however, illustrates why physical location and sovereign control must be examined separately. PSN’s capital is divided among TIM at 45%, Leonardo at 25%, CDP Equity at 20% and Sogei at 10%. PSN’s own transparency disclosure states that TIM is privately held through regulated capital markets; Leonardo is 30% owned by the Ministry of Economy and Finance but declares that it is not controlled by the ministry under the relevant Italian corporate-law definition; CDP Equity is wholly owned by CDP, itself more than 80% owned by the ministry; and Sogei is wholly owned by the ministry and subject to analogous public control. The same disclosure expressly notes that Sogei’s 10% holding is insufficient by itself to guarantee state control or dominant influence over PSN. This does not mean that PSN lacks public-interest safeguards: governance rights, the concession architecture, security requirements, contractual oversight and public-shareholder influence may collectively provide substantial control. It means that sovereignty cannot be inferred from a logo, government sponsorship or the presence of public shareholders. The decisive intelligence questions concern reserved matters, voting thresholds, board appointment rights, vetoes over subcontractors, change-of-control provisions, treatment of intellectual property, access to operational telemetry and the state’s rights during insolvency, conflict, sanctions or a severe cyber incident. Polo Strategico Nazionale – Cloud Italia – 2026; Società Trasparente – Polo Strategico Nazionale – 2026.

PSN ownership position	Equity share	Public-control characteristics	Sovereignty implication requiring verification
TIM	45%	Listed company held by private investors	Largest shareholder; evaluate governance influence, technology supply and change-of-control protections
Leonardo	25%	30% Ministry of Economy and Finance ownership; no declared statutory control	Defence and security capabilities are relevant, but state influence is not equivalent to sole control
CDP Equity	20%	Wholly owned by CDP; CDP more than 80% ministry-owned	Strong public financial influence, subject to corporate governance arrangements
Sogei	10%	Wholly ministry-owned in-house company	Direct public-administration competence, but share alone does not provide dominant influence
Combined public-influence perimeter	55% through Leonardo, CDP Equity and Sogei	Different legal forms and control intensities	Numerical aggregation does not automatically equal unified voting control
Data-centre geography	Four sites in Lazio and Lombardy	All declared on Italian territory	Concentration in two regions leaves correlated grid, fibre and regional-disaster questions
Statutory audit	EY appointed for PSN accounts	Audited reporting provides financial visibility	Financial audit does not substitute for operational-sovereignty audit

The four-site PSN architecture should be assessed against correlated-failure rather than simple site-count metrics. Two sites in Lazio and two in Lombardy can protect against the loss of one building, a local fire or a component failure, but the arrangement does not automatically protect against a regional grid event, telecom backbone disruption, common software defect, compromised identity system, supply-chain implant or simultaneous attack on replicated management infrastructure. Resilience depends on whether the sites use genuinely independent power routes, distinct substations, diversified fibre carriers, separately administered security domains, isolated backup systems and recovery mechanisms that cannot be corrupted through the same control plane. A mirrored copy is not a sovereign backup when malware, a malicious administrator or a defective automation process can modify both copies simultaneously. Similarly, a disaster-recovery site is not operationally independent when its authentication, orchestration, monitoring or patch-management systems depend on the same cloud region or software supplier as the primary site. Italy therefore requires a classified dependency map for each strategic public service, linking the visible PSN facility to the less visible layers below it: processor architecture, firmware, hypervisor, container platform, identity provider, database engine, encryption system, network-management platform, cyber-monitoring tools, remote-maintenance channels and specialist personnel. The public version of this map need not reveal exploitable details, but competent authorities must know which supposedly redundant services share the same hidden single point of failure. The NIS2 Directive recognizes data-centre and cloud providers as essential or important entities within the European cybersecurity regime and explicitly defines a data-centre service as the combined accommodation, interconnection and operation of IT equipment together with power-distribution and environmental-control infrastructure. It also places cloud and data-centre providers under jurisdictional and registration rules intended to prevent cross-border service structures from escaping regulatory visibility. Directive EU 2022/2555 on measures for a high common level of cybersecurity across the Union – European Parliament and Council – December 2022.

Cloud Concentration Converts Commercial Convenience into Systemic State Exposure

The cloud market creates sovereignty risk through concentration at three separate levels: concentration of expenditure among a small number of hyperscalers; concentration of workloads on proprietary technical services; and concentration of multiple public or regulated institutions on common underlying providers, regions and identity systems. The most authoritative current public investigation is the United Kingdom’s Competition and Markets Authority inquiry into cloud infrastructure services. Although its jurisdiction was British, its findings are directly relevant to Italy because the principal providers, licensing structures, committed-spend agreements, egress mechanisms and technical barriers operate across Europe. The CMA closed its investigation in July 2025 with a finding of an adverse effect on competition and recommended that the authority consider strategic-market-status investigations into Microsoft and Amazon Web Services, identified as the two largest providers. The official case record examined market concentration, switching, multi-cloud adoption, licensing, committed-spend agreements, technical barriers, egress charges, public-sector procurement and provider profitability. In March 2026, the CMA stated that Microsoft and Amazon had taken material steps concerning interoperability and cloud egress fees, while the authority launched a separate strategic-market-status investigation into Microsoft’s broader business-software ecosystem in May 2026. These actions establish a crucial sovereignty principle: cloud dependency does not originate only from where data are stored; it arises from the economic and technical mechanisms that make leaving a provider progressively more difficult. A public administration may technically own its data while remaining dependent on proprietary database formats, identity systems, serverless functions, AI APIs, security tools, developer pipelines and office-software licensing. A nominal multi-cloud architecture may still be concentrated when one provider controls identity, another hosts production workloads, and a third supplies the software licences required to operate both. Cloud Services Market Investigation – Competition and Markets Authority – July 2025; CMA Announces Package of Actions on Business Software and Cloud Services – Competition and Markets Authority – March 2026; Microsoft’s Business Software Ecosystem – Competition and Markets Authority – May 2026.

Concentration vector	Observable commercial practice	Hidden sovereignty effect	Required Italian countermeasure
Infrastructure concentration	Large share of workloads on one hyperscaler	Simultaneous exposure to outage, sanctions, policy change or account suspension	Sector-wide provider-concentration register
Software licensing	Preferential economics inside vendor’s own cloud	Migration becomes uneconomic even where technically possible	Procurement evaluation of licence portability
Committed-spend agreements	Customer prepays or commits large future expenditure	Creates financial penalty for diversification	Limit duration and require exit valuation
Data-egress architecture	Charges or operational complexity when moving data	Customer retains legal ownership but loses practical mobility	Tested extraction and transfer benchmarks
Proprietary PaaS	Databases, analytics and serverless tools tied to platform	Applications must be redesigned before migration	Open interfaces and architecture-neutral procurement
Identity concentration	Single identity provider authenticates multiple clouds	Failure compromises all nominally diversified environments	Independent emergency identity domain
Security-tool concentration	Same vendor protects endpoints, email, cloud and logs	One vulnerability creates correlated visibility loss	Diverse control layers and offline logging
Subcontractor concentration	Multiple prime contractors depend on same hyperscaler	Apparent supplier diversity masks common infrastructure	Full subcontracting-chain disclosure
Geographic region concentration	Many agencies select one cloud region	Regional outage becomes national service event	Region and failure-domain diversification
Personnel concentration	Migration knowledge held by contractor	Formal exit right cannot be executed internally	Public-sector skills and escrowed documentation

The European Data Act reduces several contractual barriers but does not itself guarantee sovereign portability. Regulation EU 2023/2854 requires cloud and other data-processing providers to facilitate switching, cooperate in good faith, make relevant interfaces and information available, and preserve security during migration. It provides for the gradual withdrawal of switching charges and states that, from 12 January 2027, providers may no longer impose switching charges for the switching process. It also requires information concerning procedures, machine-readable export formats, technical restrictions and the expected time needed to switch. These provisions significantly improve customers’ formal bargaining position, but the regulation’s concept of functional equivalence is strongest for basic infrastructure services and becomes more difficult at higher platform and software layers. Moving a virtual machine can be relatively straightforward; reproducing a complex public-service architecture built around proprietary identity, database, analytics, AI and event-management services may require extensive redesign. Italy must therefore distinguish data portability from service portability, application portability, security portability and operational portability. A ministry that can extract files but cannot reconstitute workflows, permissions, audit trails and interfaces within an acceptable recovery window is not sovereign in practical terms. Every strategic public-cloud contract should contain a quantified exit plan stating the total data volume, extraction rate, destination architecture, maximum tolerable downtime, treatment of metadata and logs, licence transfer conditions, required staff, expected cost and the sequence for validating service equivalence. That plan should be tested through partial migrations rather than accepted as documentary compliance. Regulation EU 2023/2854 on harmonised rules on fair access to and use of data – European Parliament and Council – December 2023; Rules on Fair Access to and Use of Data – EUR-Lex – 2024.

Public-Administration Data Require Workload-Level Mapping, Not Institution-Level Labels

Classifying an entire ministry, municipality or health authority as strategic, critical or ordinary is too coarse for an operational sovereignty model. A single institution can operate workloads with radically different consequences: a public website may be ordinary; payroll and tax systems may be critical; identity registries, emergency command systems, classified government communications or national-health datasets may be strategic. The same dataset can also change classification when combined with other information. Individually non-sensitive mobility, health, energy or procurement records can become strategically revealing when aggregated through AI analytics. Italy’s public-cloud governance should therefore establish a workload-level register linking each application to its data classification, service criticality, recovery objective, provider, region, subcontractors, encryption-key owner, administrative domain, software dependencies and exit pathway. The existing Cloud Italia framework provides the conceptual foundation by classifying information according to the damage caused by compromise of confidentiality, integrity or availability. The qualification regime also requires IaaS, PaaS and SaaS providers serving the public administration to obtain approval from the Agenzia per la Cybersicurezza Nazionale. The unresolved risk lies in continuous assurance: a service can be qualified at procurement and later change through new subcontractors, software updates, ownership transactions, control-plane relocation, AI integration or altered support arrangements. Qualification must therefore become dynamic rather than static. Every material change to data location, subcontracting, privileged support, encryption, incident response or corporate control should trigger reassessment. The register should also distinguish six geographic concepts often collapsed into the phrase “data are in Italy”: location of the active dataset; location of replicas and backups; location of metadata and telemetry; location of administrators; location of cryptographic-key control; and location of the legal entity able to compel or prevent provider action. The Digital Operational Resilience Act, although addressed to financial entities, offers a superior contractual model for strategic public workloads because it requires contracts to identify the countries or regions where services are provided and data processed, establish access and audit rights, define incident assistance, provide recovery and return of data, examine subcontracting and concentration risks, and maintain tested exit strategies for critical services. Il percorso di qualificazione dei servizi cloud della PA – Cloud Italia and Agenzia per la cybersicurezza nazionale – 2026; Regulation EU 2022/2554 on Digital Operational Resilience for the Financial Sector – European Parliament and Council – December 2022.

Data-location question	Minimum contractual answer	Sovereignty failure revealed
Where is active production data stored?	Named country, region and availability zone	Provider gives only generic “EU region”
Where are replicas and backups stored?	Complete replication and disaster-recovery geography	Backup crosses jurisdiction without customer visibility
Where are logs and metadata processed?	Security, billing, telemetry and diagnostic locations	Strategic activity exposed through metadata
Where are encryption keys controlled?	Key owner, hardware-security module location and override procedure	Provider or foreign parent can decrypt data
Where are administrators located?	Countries, entities and access conditions	Remote privileged access outside declared perimeter
Where is support delivered?	Named support entities and escalation locations	Emergency dependency on external jurisdiction
Where are subcontractors established?	Full chain for critical functions	Prime provider masks hyperscaler dependency
Where is the control plane hosted?	Region and resilience architecture	Italian data plane managed through foreign systems
Where are AI models and prompts processed?	Training, inference, logging and retention locations	Sensitive prompts leave sovereign environment
Where can service be restored?	Tested alternative region or provider	“Backup” cannot operate independently

Contractual Control Determines Whether Italy Can Act During a Crisis

The strongest sovereignty test is not normal operation but the moment when the interests of the Italian state diverge from those of the provider. Such divergence could arise during a cyberattack, sanctions dispute, diplomatic conflict, corporate insolvency, acquisition, export-control restriction, severe energy shortage, legal order from another jurisdiction or simultaneous outage affecting multiple European customers. A strategically adequate contract must establish in advance who can suspend remote administration, isolate compromised workloads, invoke emergency access, retain data, prioritize restoration, compel cooperation, obtain forensic images, activate backup licences and transfer operations to another provider. Standard commercial service-level agreements usually concentrate on uptime percentages and financial credits; they rarely provide the command authority needed for national continuity. Financial compensation after an outage does not restore public safety, tax collection, health records or emergency communications. Italy should therefore establish mandatory sovereign-control clauses for strategic and critical workloads. These clauses should include an Italian-law emergency operating licence for essential software; source-code or executable escrow where replacement is otherwise impossible; customer-controlled cryptographic keys; privileged-access recording; pre-authorized access for national incident-response authorities; prohibition on unilateral deletion or suspension during a dispute; minimum notice for subcontractor or ownership changes; inventory and priority rights over essential spare parts; and an obligation to maintain an operationally independent recovery environment. DORA again supplies a useful official model: it requires clear allocation of rights and obligations, complete service descriptions, notification of processing locations, incident assistance, audit and inspection rights, recovery and return of data during insolvency or discontinuation, termination rights, tested contingency arrangements and comprehensive exit strategies for services supporting critical functions. The regulation also treats provider concentration as a potential systemic threat to the European financial sector and requires customers to assess whether a provider is readily substitutable and whether multiple critical arrangements depend on the same provider or connected providers. Regulation EU 2022/2554 on Digital Operational Resilience for the Financial Sector – European Parliament and Council – December 2022.

Required sovereign clause	Commercial contract norm	Strategic contract requirement
Service suspension	Provider may suspend for breach or security reasons	No unilateral suspension of designated essential services without state-controlled escalation
Data deletion	Deletion after termination or non-payment	Mandatory preservation during disputes, incidents and public emergencies
Administrative access	Provider retains root-level access	Customer-controlled approval, recording and emergency revocation
Encryption keys	Provider-managed key service	Italian-controlled keys with independent hardware security
Incident support	Best-effort or paid premium support	Immediate cooperation, forensic preservation and authority access
Software continuity	Licence terminates with contract	Emergency continuation licence and escrow for essential workloads
Change of ownership	Limited notification	Government review and termination or step-in rights
Subcontracting	Provider may change subcontractors	Prior notice, risk review and prohibition for sensitive functions
Audit	Third-party certification accepted	Direct or pooled inspection rights for competent authorities
Exit	Data export tools supplied	Tested workload migration with time, cost and personnel commitments
Insolvency	General creditor and contract law applies	Data return, operational continuity and step-in mechanisms
Geopolitical disruption	Force-majeure clause	Defined continuity obligations, domestic staff and isolated operating mode

Emergency Curtailment Creates a Collision Between Energy Security and Digital Sovereignty

The previous physical-balance analysis established that very large data centres will become significant electricity consumers. The sovereignty question is what happens when available electricity is insufficient to satisfy every contracted load. A public cloud hosting health, taxation, identity or security systems may qualify as essential digital infrastructure, but the same building may also host advertising, entertainment, cryptocurrency, foreign enterprise workloads or non-urgent AI training. Curtailment cannot rationally be determined at building level alone. Italy needs the ability to distinguish essential computing from deferrable computing within the same campus and to enforce workload-aware demand reduction. A provider claiming universal mission-critical status could otherwise convert private service-level agreements into de facto priority over steel plants, rail systems, ports or urban networks. Conversely, indiscriminate disconnection of data centres could disable the digital systems coordinating those same services. The appropriate architecture is a hierarchy based on service function, not corporate identity. Tier S₁ should include national-security, emergency, identity and essential health workloads that must remain continuously available. Tier S₂ should include critical financial, tax, social-security and transport-control functions capable of limited degradation. Tier S₃ should include ordinary public and commercial workloads that can shift between sites or reduce performance. Tier S₄ should cover batch processing, AI training, non-urgent analytics and other deferrable loads. Connection agreements should require technical separation, real-time metering and automatic load shedding across these tiers. Backup generation and batteries should be integrated into the emergency plan, but operators should not receive priority merely because they possess private generators; fuel availability, emissions limits and the duration of a crisis determine whether such systems are credible. NIS2’s treatment of cloud and data-centre operators as regulated entities provides the cybersecurity perimeter, but Italy needs an additional energy-continuity doctrine specifying which workloads receive power, which can migrate, which can be degraded and which must stop. Directive EU 2022/2555 on measures for a high common level of cybersecurity across the Union – European Parliament and Council – December 2022.

Curtailment variable	Intelligence question	Required capability
Workload classification	Which applications are essential, critical, ordinary or deferrable?	Government-approved workload registry
Electrical separation	Can non-essential racks be disconnected without disabling essential systems?	Segmented power distribution and metering
Geographic mobility	Can workloads move to another Italian or European site?	Preconfigured secondary environment
Battery use	Are batteries reserved only for uptime or available to the grid?	Contracted emergency-dispatch rules
Backup fuel	How many hours of verified fuel are available?	Audited inventory and resupply plan
Network dependency	Will telecom failure make powered servers inaccessible?	Diversified carriers and emergency communication
Operator authority	Who issues and authenticates curtailment commands?	Joint energy-cyber command protocol
Private SLA conflict	Do customer contracts prohibit interruption?	Sovereign override clause
Foreign workload priority	Could external customers consume protected Italian capacity?	National emergency allocation rules
Recovery sequencing	Which services return first after interruption?	Tested national restoration plan

Cyber Resilience Must Address Common-Mode Failure and the Cloud Supply Chain

Cloud platforms can improve security by concentrating advanced defensive capabilities, but the same concentration can amplify a single defect, compromised credential, malicious software update or control-plane failure across thousands of customers. Italy’s sovereign architecture must therefore avoid the false binary in which local legacy systems are considered unsafe and hyperscale cloud automatically safe. The correct comparison concerns the probability and impact of failure across distinct architectures. Fragmented public data centres often lack patching, monitoring, redundancy and specialist personnel; Cloud Italia’s strategy noted that the overwhelming majority of public-sector data centres assessed in the earlier national census failed minimum requirements for security, reliability, processing capacity or efficiency. Consolidation is therefore necessary. Yet consolidation increases the potential blast radius. A compromised identity provider, orchestration layer or managed security platform can affect multiple ministries simultaneously. The security objective must be controlled concentration: enough consolidation to achieve professional defence and resilience, but enough architectural diversity and isolation to prevent one provider, software stack or credential system from becoming a national single point of failure. NIS2 requires cloud and data-centre providers to implement risk-management measures, address incident handling, business continuity, supply-chain security, vulnerability management, cryptography and access control. DORA adds detailed mapping of ICT assets, interdependencies, providers and critical functions and requires financial institutions to identify concentration and subcontracting risks. Italy should apply these concepts beyond finance to the whole strategic public-cloud perimeter. Each essential service should maintain a dependency graph and a failure-domain identifier showing whether nominally separate systems share the same hyperscaler region, content-delivery network, identity service, domain-name provider, security agent, software repository, certificate authority or remote-support organization. Red-team exercises should test not only ransomware and data theft but also loss of provider access, malicious support activity, corrupted backups, sanctions-based licence interruption, simultaneous cloud-region failure and manipulation of AI services used in administrative decisions. La Strategia Cloud Italia – Dipartimento per la trasformazione digitale and Agenzia per la cybersicurezza nazionale – September 2021; Directive EU 2022/2555 – European Parliament and Council – December 2022; Regulation EU 2022/2554 – European Parliament and Council – December 2022.

Common-mode cyber failure	Why conventional redundancy fails	Sovereign mitigation
Shared identity provider	Both primary and backup require same authentication service	Offline emergency identities and separate privileged domain
Shared cloud control plane	Multiple regions managed by same provider system	Independent recovery environment
Shared software update	Defective or malicious update reaches all sites	Staged deployment and isolated validation zone
Shared security agent	Security product failure disables or destabilizes endpoints	Diverse controls and safe-removal procedures
Shared certificate authority	Authentication and encryption fail simultaneously	Independent trust anchors and emergency certificates
Shared DNS or CDN	Services become unreachable despite healthy servers	Authoritative DNS diversity and direct-access capability
Shared backup credentials	Attackers encrypt or delete primary and backup data	Immutable offline copies with separate administration
Shared subcontractor	Multiple vendors rely on same hidden provider	Full chain-of-dependency registry
Shared AI service	Decision-support systems fail or produce manipulated output	Local fallback models and human override
Shared hardware firmware	Vulnerability affects all identical platforms	Hardware diversity for highest-criticality functions

France, Germany, China and the United Kingdom Reveal Four Different Sovereignty Models

European and global comparisons show that data sovereignty is moving beyond simple localization. France has developed a security-certification approach under which cloud services intended for the most sensitive uses must satisfy demanding technical and legal-protection conditions; the French concept emphasizes resistance to non-European extraterritorial access and operational control within an approved trust perimeter. Germany has favored detailed assurance frameworks and auditable cloud-security criteria, while maintaining a strong emphasis on public-sector procurement, federal control and industrial participation. The United Kingdom has adopted a more market-oriented model but has subjected cloud concentration, public procurement, licensing and switching barriers to direct competition scrutiny, acknowledging that dependence on two dominant providers can reduce customer choice and economic resilience. China uses a state-centric model that integrates localization, national-security review, cross-border-data controls and domestic cloud capacity into a broader concept of cyber sovereignty. Italy occupies an intermediate position: it has created a public cloud strategy, a national strategic infrastructure, a classification system and ACN qualification, but it continues to rely on global technologies and proprietary ecosystems. The relevant lesson is not to replicate any single foreign model. A rigid national-only model could sacrifice innovation, scalability and access to advanced AI tools; an unrestricted hyperscaler model could externalize legal control, software rents and strategic dependencies. Italy should pursue federated sovereignty: Italian physical and legal control for the highest-criticality workloads; European providers and interoperable platforms for sensitive services; qualified global clouds for ordinary workloads and carefully isolated advanced services; and mandatory portability between layers. The objective is not technological autarky, which is unattainable given foreign processor, software and networking dependencies, but the preservation of national decision authority. The state must always retain a technically credible alternative to accepting a provider’s decision, failure or legal constraint. The EU Data Act, NIS2 and DORA collectively move Europe toward this model by strengthening switching, risk management, location transparency, auditability and concentration oversight. Regulation EU 2023/2854 – European Parliament and Council – December 2023; Directive EU 2022/2555 – European Parliament and Council – December 2022; Cloud Services Market Investigation – Competition and Markets Authority – July 2025.

Sovereignty model	Primary governing instrument	Principal strength	Principal weakness	Italian lesson
French trusted-cloud model	High-assurance security and legal-protection certification	Strong protection for highly sensitive workloads	Smaller eligible supplier universe and potential higher cost	Reserve strongest legal insulation for strategic services
German assurance model	Detailed auditable security criteria and federal procurement control	Strong process discipline and industrial integration	Complex compliance and fragmented federal implementation	Standardize evidence and recurring technical audits
UK competition model	Market investigation, switching and strategic-market powers	Directly addresses concentration and licensing barriers	Does not automatically create national technological control	Treat competition as a resilience instrument
Chinese cyber-sovereignty model	Localization, security review and cross-border controls	Strong state command and domestic ecosystem protection	Reduced openness and strong state access	Location must be tied to control, but not at cost of democratic safeguards
Italian PSN and qualification model	Data classification, ACN qualification and national strategic infrastructure	Clear institutional architecture for PA migration	Layered foreign technology and governance dependencies remain	Build federated sovereignty and tested exits
EU regulatory model	Data Act, NIS2, DORA and common cloud rules	Interoperability, security, concentration oversight and customer rights	Implementation remains distributed across states and sectors	Use EU rules as minimum floor, not maximum ambition

Bayesian Assessment: Five Competing Sovereignty Outcomes to 2031

Five competing hypotheses define Italy’s 2026–2031 sovereignty trajectory. H₁ — Sovereign Consolidation assumes PSN, ACN qualification and contractual reform create genuine Italian command over strategic workloads while global technology remains contained behind enforceable controls. H₂ — Localized Dependency assumes data and servers move to Italy, but software, identity, cloud control planes and operational knowledge remain concentrated in foreign ecosystems; physical localization improves resilience without creating strategic autonomy. H₃ — Hyperscaler Capture assumes public and private cloud demand consolidates around a small number of global providers whose technical and commercial ecosystems become difficult to replace, leaving Italy with nominal regulatory authority but weak practical exit capacity. H₄ — Federated European Autonomy assumes Italy combines PSN, European providers, global clouds and interoperable architectures, allocating workloads by criticality and maintaining tested substitution paths. H₅ — Fragmented Reversal assumes migration problems, cost overruns, cyber incidents or institutional resistance lead agencies to retain fragmented legacy environments, producing neither cloud efficiency nor sovereignty. A judgmental Bayesian update assigns P(H₁)=0.20, P(H₂)=0.30, P(H₃)=0.22, P(H₄)=0.21 and P(H₅)=0.07. The highest probability currently belongs to localized dependency because Italy has made significant progress on physical hosting, classification and public-cloud governance while remaining deeply dependent on external platform, software and hardware ecosystems. H₃ remains material because international evidence demonstrates powerful concentration, licensing and switching effects. H₄ can become dominant if Italy enforces open architectures, publishes concentration metrics, requires independent identity and key management, tests exits and directs procurement toward workload-level diversification. H₁ is possible only if sovereignty is measured through operational control rather than public ownership symbolism. H₅ is less probable because the PNRR, PSN and regulatory framework make a full reversal unlikely, although legacy systems will persist. The Bayesian indicators requiring quarterly monitoring include PSN migrations completed; concentration of public expenditure by ultimate cloud provider; percentage of strategic workloads using customer-controlled keys; number of tested exits; subcontractor concentration; number of critical services with isolated recovery environments; foreign support access; incident duration; and effective geographic distribution of active, backup and management systems.

Hypothesis	2026 posterior	2031 outcome	Indicators increasing probability	Indicators decreasing probability
H₁ Sovereign Consolidation	20%	Italy commands strategic public computing despite foreign technology inputs	State-controlled keys, audited step-in rights, successful recovery exercises	Continued provider-controlled administration
H₂ Localized Dependency	30%	Data are in Italy but critical software and control remain external	Italian regions combined with foreign identity, PaaS and support	Independent control planes and tested migration
H₃ Hyperscaler Capture	22%	Two or three ecosystems dominate public and regulated workloads	Committed spend, proprietary AI services, common identity concentration	Enforced diversification and open architecture
H₄ Federated European Autonomy	21%	Workloads distributed by criticality across Italian, EU and global platforms	Interoperable procurement, European alternatives, portable security controls	Procurement based only on price and feature breadth
H₅ Fragmented Reversal	7%	Legacy public systems coexist with incomplete migration	Delays, failed projects, agency resistance and skills shortages	Migration milestones and operational service improvement

The 2026–2031 Sovereignty Timeline

During 2026, the decisive issue is visibility. Italy must establish a single authoritative inventory of public and strategic workloads, their classifications, providers, physical regions, control planes, keys, subcontractors and recovery dependencies. Completion percentages alone are inadequate because migration can increase concentration even while modernizing infrastructure. During 2027, the abolition of switching charges under the EU Data Act will remove one commercial barrier, but Italy must test whether technical and application-level portability actually improves. The key metric should be the number of strategic services successfully reconstructed in an alternative environment within their stated recovery objectives. During 2028, AI adoption will become the principal sovereignty accelerator and risk multiplier. Public authorities will increasingly consume foundation models, vector databases, proprietary accelerators and AI security services through hyperscaler platforms. Prompts, embeddings, model outputs, fine-tuning datasets and audit logs will require the same location and control analysis as conventional data. Italy must prevent a situation in which sovereign datasets remain in PSN while analytical control migrates to foreign AI APIs. During 2029, concentration risk is likely to become visible through synchronized outages, pricing pressure, licence renegotiation or regulatory findings. Public procurement will need portfolio-level limits and common-provider stress testing rather than agency-by-agency assessments. By 2030–2031, the strategic divide will be between states that can continue essential digital operations under provider failure or geopolitical separation and those that merely negotiated favorable cloud contracts during normal conditions. Italy should aim for a national minimum viable sovereign mode: a reduced but functional operating state for identity, emergency communication, health coordination, taxation, social payments, border administration and essential infrastructure control that can operate for an extended period without dependence on a single external cloud, foreign identity system or remote support centre. This does not require duplicating every public workload. It requires identifying the irreducible national digital core and proving that it can survive isolation, legal conflict and correlated technological failure.

Year	Primary sovereignty challenge	Required Italian action	Measurable output
2026	Incomplete dependency visibility	Establish national workload, provider and failure-domain register	100% of strategic and critical workloads mapped
2027	Formal switching rights versus practical portability	Conduct live exit and restoration exercises	Migration time, cost and service-equivalence metrics
2028	AI platform dependency	Classify prompts, models, embeddings and AI control planes	Sovereign AI-processing rules for public data
2029	Provider and software concentration	Apply portfolio-level exposure limits and stress testing	Concentration index by provider and technology stack
2030	Emergency operational command	Implement workload-aware curtailment and cyber-isolation protocol	National exercise covering power and cloud disruption
2031	Strategic continuity under geopolitical separation	Operate minimum viable sovereign digital core	Demonstrated multi-week continuity without single-provider reliance

Required Italian Sovereignty-Control Architecture

The required control system must unite public procurement, cybersecurity, energy security, competition policy and national-security planning. First, Italy should create a confidential National Strategic Computing Register administered jointly by the digital-transformation authorities and ACN, with controlled access for energy, defence, financial and civil-protection bodies. The register should identify ultimate provider, subcontractors, workload classification, physical and logical location, power dependencies, data flows, key ownership, privileged administrators, software licences, recovery sites and exit status. Second, every strategic contract should receive a Sovereignty Control Score covering ownership, jurisdiction, encryption, operational autonomy, interoperability, personnel, geographic resilience and emergency authority. Third, procurement should evaluate total exit cost and systemic concentration, not only annual price and technical features. Fourth, PSN and qualified-cloud services should be subjected to recurring adversarial exercises simulating loss of the primary provider, compromised identity, foreign legal order, sanctions interruption, regional power failure and simultaneous corruption of online backups. Fifth, Italy should maintain domestic teams capable of operating the minimum sovereign stack without immediate vendor support. Sixth, public AI procurement should prohibit uncontrolled export of strategic prompts, embeddings, model logs and fine-tuning data and should require an independent fallback for essential decision-support functions. Seventh, energy-curtailment planning should classify workloads inside data centres so that public necessity, not the commercial power of the tenant, determines emergency priority. Eighth, government should publish aggregated concentration indicators without revealing sensitive architectures, allowing Parliament and the public to understand whether cloud migration is creating or reducing dependency. The ultimate sovereignty metric is not the percentage of workloads “in cloud” or even “in Italy.” It is the percentage of essential state functions that Italy can securely observe, operate, isolate, restore and migrate under hostile conditions.

III. Italy’s Energy-Operator Balance Sheet: Who Can Power the Data-Centre Expansion?

Scope, Evidence Boundary and Analytical Standard

An analysis of “every single energy operator” in Italy must begin by distinguishing the legally complete operator universe from the smaller set of companies whose assets can materially affect the electricity supply available to large data centres. ARERA maintains a downloadable registry covering all entities recorded in its operator-anagraphic system, including generators, electricity sellers, distributors and other regulated entities. The universe contains thousands of legal persons, special-purpose vehicles, municipal companies, renewable-project companies and sellers that own no generation. Many disclose no plant-level cost structure, contracted output, project pipeline, debt covenants, beneficial-owner detail or customer-level supply obligations. An academically defensible assessment therefore cannot claim to reproduce “all existing data” where the data do not exist publicly or remain commercially confidential. It can, however, map the complete institutional perimeter and examine the operators controlling the overwhelming majority of dispatchable generation, renewable portfolios, electricity distribution, retail demand, storage development and grid access. The core operators assessed here are Enel, Eni–Plenitude, Edison, A2A, Sorgenia, EP Produzione, ERG, Iren, Hera, Acea–Areti, Alperia, Dolomiti Energia, Terna, E-Distribuzione, and the principal municipal distribution groups. Foreign-controlled operators already embedded in the Italian system—including EDF/Edison, Czech-controlled EP Produzione, and international capital participating in Plenitude, Sorgenia and listed utilities—are treated as part of the Italian physical system but not automatically as Italian-controlled strategic capacity. The complete legal registry remains available through Ricerca operatori – ARERA – 2026, which permits extraction of the operator population; the national capacity baseline is supplied by Terna’s generation database, while company-specific figures are taken from audited 2025 reports and official industrial plans.

The analysis uses five distinct measures that must not be conflated. Installed capacity, Kᵢ, measures the nameplate megawatts controlled or managed by operator i. Dependable capacity, Dᵢ, measures the power likely to be available during system stress after accounting for forced outages, intermittency, hydrology and fuel constraints. Annual deliverable energy, Eᵢ, measures expected megawatt-hours rather than peak power. Locational deliverability, Lᵢ,z, measures the fraction that can physically reach grid zone z without violating network constraints. Commercially uncommitted capacity, Uᵢ,t, measures output not already absorbed by retail customers, regulated obligations, capacity-market commitments, PPAs or portfolio hedges. A data-centre developer cannot infer supply capability from Kᵢ alone. The relevant quantity is:

DCSPᵢ,z,t = Kᵢ × AFᵢ,t × CFᵢ,t × Lᵢ,z,t × Uᵢ,t × Fᵢ,t

where DCSP is Data-Centre Supportable Power, AF is technical availability, CF is source-dependent utilization or capacity factor, L is locational deliverability, U is the uncommitted commercial share, and F is the fraction contractually firm during stress conditions. All coefficients lie between zero and one. A 1 GW solar portfolio with CF = 0.18 and F approaching zero after sunset cannot independently support a constant 1 GW data-centre load. A 1 GW combined-cycle plant with AF = 0.90 may provide much higher firm capacity, but fuel price, carbon cost and outage risk reduce competitiveness. No forecast formula is “impossible to refute”; the correct academic standard is reproducibility, explicit assumptions, sensitivity testing and falsifiable outputs.

National Operator Architecture

Operator class	Principal entities	Strategic asset controlled	Relevance to data centres	Principal limitation
Transmission and dispatch	Terna	National transmission grid, dispatching, interconnection and connection process	Determines whether generation can reach data-centre clusters	Does not normally sell electricity to end users
National distribution	E-Distribuzione	Largest distribution-network perimeter	Critical for smaller and medium facilities and local reinforcement	Large hyperscale campuses often require transmission-level connections
Municipal and regional distribution	Areti, Unareti, Inrete, Ireti, Deval, Edyna	Urban and regional grids	Essential for Rome, Milan, Emilia-Romagna, Piedmont, Valle d’Aosta and Alto Adige	Constrained by local transformer and feeder capacity
Integrated generation–retail	Enel, Edison, Eni–Plenitude, A2A, Sorgenia	Generation, trading, retail contracts and PPAs	Best positioned to offer bundled power, balancing and renewable procurement	Existing retail obligations reduce uncommitted output
Dispatchable thermal	EP Produzione, Eni, Sorgenia, Edison, A2A	CCGT and residual thermal fleets	Provides firm and flexible electricity	Gas and carbon-price exposure
Renewable specialists	ERG, Plenitude, Enel Green Power, EF Solare/Sorgenia	Wind, solar and storage pipelines	Can contract long-term renewable PPAs	Intermittency requires storage or firming
Hydro-regional operators	Enel, Edison, A2A, Iren, Alperia, Dolomiti Energia	Reservoir, run-of-river and pumped hydro	High strategic value for balancing and rapid response	Hydrology, concessions and environmental constraints
Multiutilities	A2A, Iren, Hera, Acea, Alperia, Dolomiti Energia	Networks, generation, waste-to-energy, heat and retail	Strong local permitting and infrastructure integration	Geographic focus and public-governance complexity
Foreign-controlled generation	Edison/EDF, EP Produzione/EPH	Large Italian power portfolios	Significant physical support capability	Strategic decisions and capital allocation made outside Italy
Infrastructure-capital platforms	Sorgenia/F2i–Sixth Street, Plenitude minority investors	Consolidated generation and retail platforms	Access to capital and acquisition capacity	Potential future stake rotation or foreign exit

The national architecture is already substantially internationalized. Edison operates as an Italian industrial company but is controlled within the French EDF group. EP Produzione belongs to the Czech EPH group and manages major Italian thermal assets. Sorgenia, following its integration with EF Solare Italia, is approximately 62% controlled by the Italian infrastructure manager F2i, while US investment firm Sixth Street holds approximately 38%; Crédit Agricole retains minority interests in parts of the renewable perimeter. Plenitude remains controlled by Eni, but Eni has sold a 20% stake to Ares and a further aggregate 10% to the Energy Infrastructure Partners fund. These transactions do not remove the assets from Italy, but they change governance, return expectations, exit optionality and the probability of future IPOs or secondary share placements. Sorgenia’s December 2025 integration created a group with more than 6 GW of installed capacity and more than one million customers; the company reports €722 million of pro-forma 2025 EBITDA, while its Italian portfolio includes 3.18 GW of wholly controlled CCGT capacity, approximately 4.3 GW including its share of Tirreno Power, about 960 MW of Italian solar, 300 MW of Italian wind and 70 MW of bioenergy.

Operator Comparison: Current Position and Data-Centre Capability

Operator	2025 disclosed operating position	Principal 2030 or plan target	Firm-power capability	Renewable PPA capability	Data-centre support assessment
Enel	~68 GW managed renewable capacity globally; 69 million network users	>80 GW total renewable capacity under 2026–2028 plan; ~€53bn gross investment	High through diversified generation, hydro, storage and retail portfolio	Very high	Strongest integrated candidate, subject to Italy-specific capacity allocation
Eni–Plenitude	5.8 GW renewable capacity; 5.6 TWh renewable production; 18.63 TWh retail electricity sales	15 GW renewables by 2030; EBITDA >€2.5bn	High when combined with Eni gas and thermoelectric portfolio	Very high	Strong corporate-PPA and bundled supply platform
Edison	2.3 GW renewable capacity; €17.7bn revenue; €1.3bn EBITDA	4 GW renewables and >2.5 GW storage by 2030	Very high owing to thermal and hydro portfolio	High	One of the most credible hyperscale counterparties
Sorgenia	>6 GW pro-forma capacity; ~4.3 GW CCGT including Tirreno Power share; 960 MW Italian solar	Renewable and storage integration following EF Solare combination	Very high	High	Strong independent supplier, but ownership partly foreign financial capital
EP Produzione	Portfolio moving toward 6 GW managed Italian capacity	>€500m investment 2025–2029; new Tavazzano and Ostiglia capacity	Very high	Low to medium	Excellent firming provider; carbon and gas exposure
ERG	€540m adjusted EBITDA; ~150 MW installed in 2025–early 2026	Construction of ~230 MW and development of 700 MW during 2026; repowering and storage focus	Low without external firming	High	Strong renewable PPA partner, not standalone baseload supplier
Iren	€1.353bn EBITDA; €925m investment; 74% EBITDA regulated/semi-regulated	€6.4bn investment plan; €1.6bn EBITDA by 2030; solar and wind portfolio to 430 MW	Medium through hydro, CHP, thermal and WTE	Medium	Strong regional solution for northern clusters
Hera	Multiutility with 4.6m energy customers at end-2024 and strong M&A history	€4.6bn 2024–2028 investment; EBITDA target €1.7bn in 2028	Medium through WTE, CHP, procurement and retail	Medium	Better as integrated energy-services and local-infrastructure partner
Acea	€1.420bn pro-forma EBITDA; €1.531bn investment; 96% pro-forma EBITDA regulated	€7.6bn 2024–2028 investment, largely networks and water	Low to medium generation; high distribution relevance in Rome	Medium	Critical for grid and water access in Rome, less suited as principal generator
Alperia	3.8 TWh hydro generation in 2025; 5.6 TWh electricity sales	Renewable diversification beyond hydro	High regionally, hydrology-adjusted	Medium	Valuable Alpine renewable and balancing partner
Terna	National transmission and dispatch monopoly	>€23bn 2025–2034 development plan previously identified	Not a generator	Not applicable	Determines connection feasibility and congestion risk
A2A	Major hydro, thermal, waste-to-energy, networks and retail platform	Long-term multiutility investment and decarbonization plan	High, especially Lombardy	High	Structurally important for Milan data-centre cluster
Dolomiti Energia	Large regional hydro-led portfolio	Expansion constrained by concessions and regional strategy	Medium to high regionally	Medium	Attractive for Alpine renewable contracts; limited hyperscale volume
E-Distribuzione	Largest Italian DSO	Large grid-resilience and digitalization investment	Network rather than generation	Not applicable	Critical enabler outside direct Terna connections

Enel

Enel has the deepest combination of capital access, generation diversification, network competence, renewable-development capability, storage expertise, energy trading and customer aggregation. At the end of 2025 the group reported approximately 68 GW of managed renewable capacity globally and networks serving roughly 69 million end users. Its 2026–2028 strategic plan provides approximately €53 billion of gross investment, including more than €26 billion in networks and roughly €20 billion in renewable generation. Approximately 55% of planned network investment is allocated to Italy, while the group expects to add around 15 GW of renewable capacity globally, taking total installed renewable capacity above 80 GW. [Piano Strategico 2026–2028 – Enel – February 2026]

Enel’s capacity to support Italian data centres should not be measured by its global 68 GW renewable portfolio. The relevant variables are Italian generation available in the required zone, the output already committed to existing customers, regulatory restrictions between generation and distribution activities, and the ability to construct dedicated additional capacity. Enel can provide three distinct products: a conventional retail supply backed by its portfolio and wholesale procurement; a renewable PPA tied to new or existing assets; and a hybrid solution combining renewable production, storage, demand response and residual grid supply. Its competitive advantage is lowest counterparty risk and the ability to internalize balancing across a broad portfolio. Its weakness is that a very large data-centre customer may compete with Enel’s existing industrial and retail load, while the company’s capital plan allocates investment across several countries. Enel is unlikely to dedicate multi-gigawatt Italian capacity without long-duration contracts, creditworthy hyperscaler counterparties and strong locational economics.

Eni and Plenitude

Eni–Plenitude is evolving from an energy retailer with renewable assets into a partially monetized transition platform. At the end of 2025 Plenitude reported 5.8 GW of installed renewable capacity, up from 4.1 GW in 2024, renewable production of 5.6 TWh, electricity sales to end customers of 18.63 TWh, approximately 10 million retail and business customer points, and pro-forma EBITDA close to €1.1 billion. Eni’s 2026–2030 plan targets 15 GW of Plenitude renewable capacity by 2030 and EBITDA above €2.5 billion. The acquisition of Acea Energia is expected to lift the customer portfolio above 11 million.

Plenitude has one of the strongest commercial architectures for supplying data centres because it can combine corporate PPAs, retail load aggregation, renewable development and the wider Eni group’s gas, LNG and thermoelectric capabilities. Eni reported 20.53 TWh of thermoelectric production in 2025 and 27.57 TWh of open-market electricity sales. This creates a portfolio able to manage renewable intermittency more effectively than a pure-play wind or solar generator. The ownership model nevertheless deserves scrutiny. Eni has sold minority stakes in Plenitude to Ares and EIP. Such transactions provide growth capital and establish external valuations but also create shareholder-return expectations, potential future liquidity rights and pressure for an IPO or further sell-down. The Italian state retains strategic influence through its position in Eni, but the economic ownership of future renewable cash flows is becoming more internationally distributed. Plenitude’s five-year growth probability is high because of its 15 GW target, retail scale and access to Eni’s balance sheet; the principal risks are construction costs, power-price compression, integration of Acea Energia and the valuation demands of external shareholders.

Edison

Edison is arguably the most immediately credible operator for a large data-centre supply agreement because it combines hydroelectricity, thermoelectric generation, renewable development, gas sourcing, retail sales and planned storage. In 2025 it generated revenue of €17.7 billion, EBITDA of approximately €1.3 billion, and net profit of €240 million. Renewable investment rose sharply, 200 MW of projects were completed and another 250 MW remained under construction. Edison reported 2.3 GW of installed renewable capacity in 2025 and targets 4 GW by 2030, together with more than 2.5 GW of storage. Its plan envisages annual investment of approximately €1–1.5 billion and EBITDA of roughly €1.7–1.9 billion by 2030.

Edison’s principal advantage is dispatchable diversity. Hydroelectric assets can deliver flexibility and ancillary services; combined-cycle plants can provide firm output; storage can shift renewable production and participate in balancing. This makes Edison more capable than a pure renewable developer of providing an hourly matched product. Its principal sovereignty issue is foreign control through EDF. Physical assets, employees, taxes and operating licences remain Italian, but strategic capital allocation ultimately sits within a French state-controlled group. A hypothetical disposal of Italian assets, shift in dividend policy, internal prioritization of French demand or cross-border restructuring would therefore require Italian authorities to examine the transaction under competition, energy-security and golden-power rules. Edison itself is not a likely hostile-takeover target in the normal listed-market sense because control is already concentrated. The more realistic transaction risk concerns individual power plants, renewable portfolios, minority stakes, joint ventures or internal group reorganization.

Sorgenia

Following the integration with EF Solare Italia, Sorgenia has become one of the most strategically significant independent platforms in Italy. It reports more than 6 GW of installed capacity on a pro-forma basis, approximately one million customers and €722 million of 2025 pro-forma EBITDA. The portfolio combines 3.18 GW of directly controlled CCGT capacity, approximately 4.3 GW when including its share of Tirreno Power, 960 MW of Italian photovoltaic generation, 300 MW of Italian wind and 70 MW of bioenergy.

For data centres, Sorgenia offers a strong physical hedge: solar assets can supply lower-cost daytime electricity while efficient CCGT units cover residual demand and system stress. It has also begun integrating BESS into thermal sites, including the Termoli project, and has signed Energy Release arrangements requiring the construction of new renewable capacity. Ownership is the central strategic variable. F2i’s approximately 62% stake anchors control in an Italian infrastructure-fund structure, but Sixth Street’s approximately 38% interest is substantial and could eventually be sold, refinanced or used in a wider consolidation. A full foreign takeover is legally possible only subject to shareholder agreement, financing, antitrust review and potential golden-power intervention. A partial participation sale is more probable than outright loss of control. Sorgenia’s five-year growth outlook is positive because it can cross-sell renewable and firm power, but its carbon exposure remains significant. If gas and EU ETS prices rise materially, the cost of 24/7 supply will rise even while the strategic value of its thermal plants increases.

EP Produzione

EP Produzione is one of the most important operators for adequacy rather than decarbonized branding. The company, controlled by the Czech EPH group, expects managed Italian capacity to reach approximately 6 GW following new units at Tavazzano e Montanaso and Ostiglia. It plans more than €500 million of investment between 2025 and 2029, in addition to over €200 million invested during 2024. Its new combined-cycle assets are designed for efficiency of approximately 62%, and the company previously secured capacity-market contracts for new and existing capacity.

EP Produzione can support data centres indirectly by maintaining system adequacy and directly through bilateral industrial supply via EP Energia Italia. Its portfolio is particularly relevant to northern Italy, where data-centre connection requests are concentrated. The commercial formula for a CCGT-backed supply price can be represented as:

Pₑ = HR × Pgas + EF × PCO₂ + VOM + FC + RM

where HR is heat rate in MWh thermal per MWh electrical, Pgas is delivered gas price, EF is carbon-emission factor, PCO₂ is allowance price, VOM is variable operating cost, FC is allocated fixed-cost recovery and RM is risk margin. At 62% efficiency, HR is approximately 1.61 MWh thermal per MWh electrical before auxiliary losses. This makes EP Produzione’s competitiveness highly sensitive to gas and carbon prices. Its five-year physical importance is high, but its long-term valuation is exposed to decreasing annual running hours as renewables and storage grow. Foreign control does not make the operator unreliable, but it places strategic decisions—asset rotation, capital expenditure and potential sale—outside Italy.

ERG

ERG is a renewable specialist with strong capability in wind development, repowering and PPAs, but it cannot independently guarantee a constant hyperscale load without storage or external balancing. In 2025 it generated adjusted EBITDA of €540 million and adjusted net profit of €155 million. Around 150 MW of new capacity entered operation between 2025 and early 2026. During 2026 the company plans approximately 230 MW under construction and development of a further 700 MW, divided between wind repowering and storage. It secured long-term Italian FER X remuneration for 141 MW of repowering projects and signed eight PPAs covering 8.7 TWh.

ERG’s “value over volume” discipline requires investments to create a return 200–400 basis points above WACC. This is a useful benchmark for data-centre PPAs: a hyperscaler can reduce financing cost through a long-duration contract, but the contracted price must still cover development, construction, degradation, imbalance, curtailment, financing and required equity return. ERG is well positioned to sign additional corporate PPAs in southern Italy and transfer the environmental attributes to northern data-centre customers. It is less capable of solving the physical locational problem unless the PPA is paired with grid reinforcement, storage or a firming supplier. The company’s five-year outlook is moderate growth with asset rotation rather than maximum capacity expansion. Acquisition risk exists because ERG is listed and internationally active, but concentrated family influence and Italian strategic review reduce the probability of an unsolicited full takeover. Portfolio-level sales and foreign joint ventures remain more likely.

A2A

A2A is strategically indispensable to the Milan–Lombardy data-centre ecosystem because it combines electricity distribution through Unareti, hydroelectric generation, thermal plants, waste-to-energy, district heating, retail supply and municipal relationships. Its significance exceeds its generation statistics: a data-centre campus in or around Milan may depend on A2A-group distribution infrastructure, local substations, heat-recovery opportunities, water relationships, land planning and municipal authorization. The shareholder structure is protected by the combined positions of the municipalities of Milan and Brescia, making a foreign acquisition of control politically and legally difficult even though international institutional investors can accumulate minority positions. CONSOB’s current ownership disclosures identify the significant holdings and voting structure, while the company’s public-governance architecture gives the two municipalities decisive influence.

A2A’s competitive advantage is therefore regional integration rather than only low-cost generation. It can potentially offer data centres grid connection, renewable contracts, district-heat integration, demand response, cooling-water coordination and municipal planning within one industrial relationship. Its limitation is the same conflict identified in the physical-balance chapter: Milan’s digital demand competes with urban electrification, transport, hospitals and Lombardy industry. A2A should not be evaluated solely as a supplier seeking new sales; its regulated and municipal responsibilities require it to protect network reliability. Over five years the company is likely to grow through regulated investment, environmental infrastructure, storage and selective generation. A foreign takeover is low probability; minority financial participation, project-level partnerships and acquisitions by A2A are more plausible.

Iren

Iren reported 2025 EBITDA of €1.353 billion, net profit of €301 million and technical investment of €925 million. Approximately 74% of EBITDA came from regulated or semi-regulated activities. The 2025–2030 plan provides €6.4 billion of gross technical investment and targets EBITDA of approximately €1.6 billion in 2030. The strategy places greater weight on regulated networks and disciplined capital allocation, while slowing part of the previously envisioned renewable expansion. Its solar and wind portfolio is expected to reach around 430 MW by 2030; Iren reported 210 MWp of photovoltaic capacity and about 7 MW of wind in the disclosed earlier operating perimeter.

Iren is not the largest generation counterparty, but it has important hydroelectric, cogeneration, district-heating, waste-to-energy and distribution assets across Piedmont, Liguria and Emilia. It can support regional or edge data centres, particularly where waste heat can be integrated into district-heating networks. Its regulated cash flows reduce counterparty risk, although the shift toward a more focused multiutility means it may reject capital-intensive renewable projects that fail its return thresholds. Iren is municipality-influenced and therefore difficult to acquire outright. Consolidation with another Italian multiutility is more plausible than foreign control, although foreign infrastructure funds could participate in individual network, renewable or environmental assets.

Hera

Hera is primarily a multiutility and commercial-energy platform rather than a large standalone generator. Its strategic value lies in energy retail, waste-to-energy, cogeneration, distribution, environmental services, water, digital infrastructure and a proven M&A model. Hera reported 4.6 million energy customers at the end of 2024 and has described external growth as responsible for roughly 40% of its historical expansion. Its 2024–2028 plan includes approximately €4.6 billion of investment, with around €1.3 billion dedicated to innovation and digitalization, and targets EBITDA of approximately €1.7 billion in 2028.

Hera can support data centres through aggregated supply, local network coordination, waste-heat use, water and environmental permitting, particularly in Emilia-Romagna and the north-east. It is less likely to offer hundreds of megawatts of proprietary firm generation without wholesale procurement or partnerships. The company is structurally designed to acquire and integrate municipal and regional utilities. Its own foreign-takeover probability is low because of public-shareholder governance and territorial sensitivity. The higher-probability transaction path is continued acquisition of smaller operators, customer portfolios, environmental assets and distribution interests.

Acea and Areti

Acea is central to any data-centre strategy in Rome because its Areti subsidiary controls the local electricity-distribution infrastructure, while the group also manages water, environmental and public-lighting systems. Acea reported pro-forma 2025 EBITDA of €1.420 billion, investments of €1.531 billion, and approximately 96% of pro-forma EBITDA from regulated businesses after treating Acea Energia as a discontinued operation. Its 2024–2028 plan contains €7.6 billion of investment, targets RAB of €10.5 billion and EBITDA of approximately €1.8 billion in 2028.

The sale of Acea Energia to Eni Plenitude demonstrates the group’s strategic refocusing from competitive retail toward regulated infrastructure. The transaction covers 100% of Acea Energia, including its 50% interest in Umbria Energy, while excluding selected efficiency, mobility, circular-economy and energy-management activities. For data centres, Acea is consequently more important as gatekeeper of electrical and water infrastructure than as energy merchant. Rome-area projects may require Areti reinforcement and Acea water analysis even if electricity is purchased from Enel, Plenitude or Edison. Rome municipality remains the controlling shareholder, while Suez held a disclosed 19.333% interest in March 2026, including a non-voting portion. A foreign takeover of Acea is improbable without municipal consent, but foreign minority participation already exists and could be restructured.

Alperia and Dolomiti Energia

Alperia produced approximately 3.8 TWh of hydroelectric electricity in 2025, plus smaller quantities from wind, solar and cogeneration, while selling 5.6 TWh to customers. Its distribution network delivered 2.7 TWh through around 243,000 withdrawal points and extended approximately 9,505 kilometres. Alperia’s hydro portfolio can provide highly valuable renewable balancing, but drought exposure and regional electricity needs limit the quantity available for external data-centre contracts. A northern Italian campus could obtain a long-term renewable supply agreement, but physical delivery remains subject to the national grid.

Dolomiti Energia has similar strategic characteristics: hydroelectric strength, regional public ownership, retail activity and access to Alpine generation. Both companies are protected by territorial governance and concession structures. Their most likely role is not to power a multi-gigawatt national data-centre expansion independently, but to supply high-quality renewable blocks, balancing products and regional infrastructure. Their acquisition risk is low at parent-company level and higher at project or minority-participation level.

Competitiveness and Cost Structure

A data-centre electricity contract must be evaluated through total delivered cost rather than nominal generation cost. For operator i, zone z and contract year t:

TDCᵢ,z,t = LCOEᵢ,t + Cbalᵢ,t + Cgrid_z,t + Ccapᵢ,t + Cfirmᵢ,t + Ccertᵢ,t + Ctax_t + Criskᵢ,t

Where:

• LCOE is levelized generation cost;
• Cbal is imbalance and profile-management cost;
• Cgrid is network and congestion cost;
• Ccap is capacity or adequacy cost;
• Cfirm is the cost of ensuring supply when the contracted renewable source is unavailable;
• Ccert is the cost of guarantees of origin or equivalent attributes;
• Ctax includes applicable fiscal and parafiscal charges;
• Crisk prices credit, regulatory, curtailment and volume uncertainty.

Technology and operator type	Cost advantage	Hidden cost	Suitability for 24/7 data centre
Existing hydro	Low marginal cost and rapid response	Concession renewal, hydrology and opportunity cost	Excellent but limited volume
New solar	Low daytime generation cost	Night-time firming, land, curtailment and grid cost	Strong component, not standalone solution
New onshore wind	Competitive annual production	Forecast error and low-wind periods	Strong with diversified balancing
CCGT	Dispatchable and locationally useful	Gas, EU ETS and lower future running hours	High physical reliability, volatile price
Waste-to-energy	Programmable and locally embedded	Limited scalable volume and environmental constraints	Useful regional baseload
Batteries	Fast response and peak shifting	Duration, degradation and charging-source dependence	Essential for flexibility, not primary energy source
Retail portfolio	Can aggregate multiple sources	Supplier margin and hidden wholesale exposure	Attractive for customer simplicity
Direct wire or behind-the-meter	May reduce network exposure	Permitting, redundancy and stranded-asset risk	Valuable only at appropriate sites

The strongest likely counterparties for a hyperscale campus are therefore Enel, Edison, Plenitude, Sorgenia and potentially A2A, because each can combine multiple technologies and market functions. ERG is a high-quality renewable originator but needs a firming partner. EP Produzione is an excellent firming counterparty but offers a more carbon-intensive product. Iren, Hera, Acea, Alperia and Dolomiti Energia are stronger for regional, infrastructure-integrated or smaller-scale projects.

Acquisition, Participation and Foreign-Control Risk

Operator	Current control characteristic	Full foreign-acquisition probability, 2026–2031	More probable transaction
Enel	Italian state retains strategic minority influence	Very low	Asset rotation, minority stakes, project JVs
Eni–Plenitude	Eni control with foreign minority investors	Low for Eni; medium for further Plenitude sell-down	IPO, additional minority placement, renewable asset partnerships
Edison	Already controlled by French EDF	Not applicable as new foreign takeover	Internal restructuring or asset sale
EP Produzione	Already Czech-controlled	Not applicable	Plant sale, portfolio restructuring or new foreign investment
Sorgenia	F2i majority, Sixth Street large minority	Medium	Sixth Street exit, IPO, strategic stake sale or merger
ERG	Listed with stable Italian reference ownership	Low to medium	Asset rotation, minority investment or friendly strategic transaction
A2A	Municipal control	Very low	Project JVs and acquisitions by A2A
Iren	Municipal-shareholder structure	Very low	Italian multiutility consolidation or asset partnerships
Hera	Municipal and territorial governance	Very low	Continued acquisition of smaller operators
Acea	Rome municipality control; Suez minority	Very low	Minority reshaping and regulated-asset partnerships
Alperia	Regional public ownership	Very low	Project-level financing or concessions partnership
Terna	CDP Reti holds 29.851%; regulated national monopoly	Extremely low	Institutional share movements below control threshold
Snam/Italgas	Strategic regulated infrastructure	Extremely low	Foreign institutional minority accumulation

Italy’s strategic-energy regime makes foreign acquisition of control over critical assets legally and politically difficult, but it does not prohibit all foreign participation. The meaningful pathways are minority purchases, fund investments, joint ventures, project-finance structures, holding-company mergers, customer-portfolio sales and asset rotations. Terna’s ownership illustrates the model: CDP Reti holds 29.851%, while institutional investors can hold significant minority positions; CONSOB reported BlackRock moving to 4.711% in June 2026 and identifies other relevant investors. Such ownership does not transfer operational control but exposes valuation and governance to global capital markets.

The most transaction-sensitive company is Sorgenia because its ownership already combines Italian infrastructure capital and a large US financial investor. Plenitude is also transaction-active by design. Eni’s “satellite” model monetizes minority stakes while preserving industrial control, allowing outside investors to fund growth. This can accelerate renewable construction, but it means that returns from Italian customer contracts and potentially data-centre PPAs are shared with foreign funds. Edison and EP Produzione demonstrate that substantial Italian generating capacity can be foreign-controlled without leaving the national regulatory perimeter. The correct national-security question is therefore not “Is the shareholder foreign?” but “Can the asset be operated, financed, maintained and dispatched in Italy during a conflict between shareholder interests and national-system requirements?”

Five-Year Growth and Decline Outlook

Operator	Growth probability	Decline probability	Central 2031 thesis
Enel	80%	10%	Growth in networks, renewables and storage; selective rather than indiscriminate Italian generation expansion
Plenitude	85%	8%	Rapid renewable and customer growth, supported by acquisitions and minority capital
Edison	82%	10%	Strong growth in renewables and storage while retaining dispatchable portfolio
Sorgenia	78%	15%	Integration-led growth; potential ownership transaction by 2031
EP Produzione	55%	35%	Firm-capacity importance rises while annual thermal utilization eventually declines
ERG	70%	18%	Moderate selective renewable growth, repowering and storage; asset rotation remains central
A2A	78%	10%	Strong regulated, environmental, storage and Lombardy infrastructure growth
Iren	75%	10%	Regulated and semi-regulated growth; slower renewable-volume expansion
Hera	76%	8%	Continued M&A and regulated growth, limited direct hyperscale-generation role
Acea	72%	10%	Infrastructure-led expansion; reduced competitive energy exposure
Alperia	55%	20%	Stable hydro value, but climate and concession risks constrain output growth
Dolomiti Energia	55%	20%	Regional stability with selective renewables and services growth
Terna	95%	2%	Structural investment expansion driven by electrification and connection demand

These are judgmental probabilities rather than market-price forecasts. They are derived from official capital plans, regulated-revenue exposure, asset age, technology mix, leverage, ownership stability and identifiable project pipelines. The growth function can be formalized as:

Gᵢ = w₁Pᵢ + w₂Rᵢ + w₃Fᵢ + w₄Nᵢ + w₅Mᵢ − w₆Lᵢ − w₇Cᵢ − w₈Xᵢ

where P is permitted project pipeline, R is regulated or contracted revenue share, F is financing capacity, N is network or locational advantage, M is M&A execution ability, L is leverage pressure, C is commodity and carbon exposure, and X is execution and authorization risk. The weights must be disclosed and calibrated rather than presented as universal constants.

The most probable system outcome by 2031 is consolidation around five commercially powerful electricity suppliers—Enel, Plenitude, Edison, A2A and Sorgenia—supported by specialist generators and regional multiutilities. Data-centre operators will favor counterparties able to combine renewable PPAs, balancing, capacity, guarantees of origin and credit strength. Pure renewable operators will increasingly sell output through structured agreements rather than raw electricity. Thermal operators will experience fewer average operating hours but higher scarcity value during constrained periods. Distribution companies will gain regulated investment but face political resistance if data-centre reinforcement appears to socialize private infrastructure costs.

Operator-Specific Ability to Support New Data Centres

Operator	Renewable block	24/7 firming	Storage	Grid or municipal integration	Overall suitability
Enel	5/5	5/5	5/5	5/5	Very high
Edison	4/5	5/5	5/5 planned	3/5	Very high
Plenitude/Eni	5/5	4/5	3/5	3/5	Very high
A2A	4/5	4/5	4/5	5/5 in Lombardy	Very high regionally
Sorgenia	4/5	5/5	3/5	3/5	High
EP Produzione	1/5	5/5	2/5	2/5	High for firming
ERG	5/5	1/5	3/5 developing	1/5	High as PPA originator
Iren	3/5	3/5	3/5	5/5 regionally	Medium–high
Hera	2/5	3/5	2/5	5/5 regionally	Medium–high for integrated services
Acea	2/5	1/5	2/5	5/5 in Rome	High as infrastructure gatekeeper
Alperia	4/5	4/5 hydrology-adjusted	3/5	4/5 regionally	Medium–high
Dolomiti Energia	4/5	4/5 hydrology-adjusted	3/5	4/5 regionally	Medium–high

No operator can guarantee that a new data centre will receive power merely because it owns sufficient annual generation. The final support test must satisfy simultaneously:

• Energy sufficiency: annual contracted MWh cover expected consumption.
• Capacity sufficiency: firm MW cover peak demand.
• Hourly matching: generation, storage and balancing cover the load profile.
• Locational feasibility: transmission and distribution networks can deliver power.
• Additionality: new demand is matched by new generation or demonstrable surplus.
• Water and cooling feasibility: local water and heat-rejection constraints are acceptable.
• Emergency flexibility: non-critical computing can be curtailed.
• Ownership resilience: provider control or financial change does not terminate supply.
• Contractual portability: the data centre can change supplier without losing grid rights.
• System cost allocation: dedicated reinforcement is paid by the beneficiary where causally attributable.

Strategic Judgment

Italy possesses enough sophisticated energy operators to support a significant data-centre industry, but not enough unconstrained, firm, low-carbon and geographically matched electricity to support the full connection queue. Enel, Edison, Plenitude, A2A and Sorgenia can structure credible multi-year products. EP Produzione can provide firmness but not decarbonized supply alone. ERG can produce additional renewable blocks but requires firming. Iren, Hera, Acea, Alperia and Dolomiti Energia can unlock regional solutions, especially when data centres integrate heat recovery, water reuse and grid services. Terna and the DSOs remain the final arbiters of physical feasibility regardless of the commercial contract.

The acquisition question is equally important. Italy is not facing a single imminent foreign takeover of the energy system; it is experiencing gradual internationalization through existing foreign control, minority capital, asset rotation, infrastructure funds, PPAs and project-level partnerships. The most probable five-year transactions are additional Plenitude monetization, a liquidity event or ownership adjustment at Sorgenia, renewable portfolio trades by ERG and other developers, municipal-utility consolidation, and further acquisitions of retail customer books. Full foreign acquisition of Enel, Eni, Terna, A2A, Iren, Hera, Acea or Alperia is low probability because of state, municipal or regional control and the strategic-asset regime. Foreign ownership of individual plants, renewable projects and minority stakes will continue to rise.

The final policy principle is mathematically simple but institutionally demanding:

Approve a data-centre load only where NVAᵈᶜ > OCCᵉⁿᵉʳᵍʸ + SRCᵍʳⁱᵈ + WRCʷᵃᵗᵉʳ + SDRˢᵒᵛ

where NVAᵈᶜ is the net national value added by the computing investment, OCCᵉⁿᵉʳᵍʸ is the opportunity cost of electricity diverted from alternative uses, SRCᵍʳⁱᵈ is the socialized reinforcement cost, WRCʷᵃᵗᵉʳ is water-resource cost, and SDRˢᵒᵛ is the strategic-dependency risk generated by foreign ownership or technological control. The variables cannot be reduced to a single universally correct number, but each can be disclosed, scored and stress-tested. Without this test, the strongest energy operators will rationally sell electricity to the highest-credit customer while the Italian system absorbs congestion and opportunity costs. With it, operator growth, foreign capital and data-centre construction can be directed toward additional generation, storage and national computing value rather than simple extraction of scarce grid capacity.

IV. The 2026–2031 Allocation Conflict: Who Receives Europe’s Next Megawatt?

Electricity Allocation Is Becoming Industrial Policy

Between 2026 and 2031, Italy’s central electricity problem will not be a simple shortage of annual kilowatt-hours. It will be an increasingly explicit conflict over the allocation of firm power, high-voltage connection capacity, network reinforcement, storage, dispatchable reserves, cooling water, land near substations and public financial support. Data centres enter this contest with structural advantages: their developers are often globally capitalized; their customers can sign long-duration contracts; their facilities can be built in standardized phases; their credit quality may be superior to that of energy-intensive manufacturers; and political authorities associate artificial intelligence and cloud computing with productivity, sovereignty and international investment. Manufacturing, rail transport, urban mobility, ports, hospitals and household electrification nevertheless create broader domestic employment, public-service and decarbonization value that is not always reflected in the price a project can offer for electricity. A market-only allocation can therefore deliver a privately efficient but nationally suboptimal result. The operator willing to pay the highest price for a connection may not be the operator producing the highest Italian value per constrained megawatt. Italy’s national electricity demand reached 311.9 TWh in 2024, with net production of 263.2 TWh, imports of 55.9 TWh, exports of 4.9 TWh and gross efficient generating capacity of 137.6 GW. These figures show why installed capacity cannot be used as evidence of abundant supply: Italy simultaneously possessed 137.6 GW of nameplate capacity and remained a major net importer because solar and wind output vary, hydro depends on hydrology, plants undergo outages, thermal generation faces fuel and carbon costs, and network constraints prevent all generation from serving all loads at all times. Statistical Publications – Terna – 2025 — Official electricity statistics.

The allocation conflict can be stated through a constrained national-welfare function. Let xⱼ,z,t represent the firm megawatts allocated to project j in electrical zone z during period t. Italy should maximize:

NW = Σⱼ [VAⱼ + SEⱼ + DSⱼ + CO₂ⱼ + LRⱼ − GCⱼ − WCⱼ − CRⱼ − DRⱼ] × xⱼ,z,t

subject to:

Σⱼ xⱼ,z,t ≤ Kfirm,z,t

Σⱼ Eⱼ,z,t ≤ Gz,t + Iz,t + Dz,t − Xz,t − Rz,t

Wⱼ,b,t ≤ Wavailable,b,t

N−1z,t ≥ N−1minimum

Here, VA is domestic value added, SE is supply-chain and employment value, DS is digital or strategic sovereignty, CO₂ is avoided emissions, LR is locational relief supplied to the network, GC is grid-reinforcement cost, WC is water and territorial cost, CR is concentration risk and DR is dependency risk. Kfirm is dependable capacity at the relevant node, G is domestic generation, I imports, D discharging storage, X exports, and R operating reserve. The formulation does not prescribe a single political answer; it demonstrates why a connection queue ordered only by application date cannot optimize national welfare. A hospital extension and an AI-training campus are not equivalent because they request the same electrical product. A steel plant and a cloud facility are not equivalent because both can sign a PPA. The allocation system must compare their additionality, flexibility, locational impact and consequences if Italy refuses the project.

Allocation claimant	Electricity characteristic	Public-interest value	Flexibility potential	Principal consequence if delayed
Data-centre AI training	High, continuous, highly concentrated	Strategic compute, investment and possible productivity spillovers	Potentially high if workloads can shift geographically or temporally	Compute is imported; investment may relocate
Cloud inference and sovereign computing	High and latency-sensitive	Public services, industrial AI and national data control	Medium; essential functions may be inflexible	Dependence on foreign regions and reduced service resilience
Steel and metallurgy	Very high, process-concentrated	Employment, defence supply chains and industrial autonomy	Medium, depending on furnace and production cycle	Plant closure, import dependence and carbon leakage
Chemicals and pharmaceuticals	High and often continuous	Strategic materials, medicines and high-value exports	Low to medium	Production relocation and supply-chain vulnerability
Rail traction	Predictable but spatially constrained	Public mobility, freight decarbonization and national productivity	Limited within operating schedules	Capacity bottlenecks and continued road dependence
Ports and shore power	High at vessel calls	Logistics competitiveness and local air-quality gains	Medium through scheduled berthing	Continued marine-fuel use and urban pollution
Electric mobility	Distributed and increasingly significant	Oil-import reduction and industrial transition	High with managed charging	Slower decarbonization and automotive-market decline
Hospitals and water systems	Continuous and mission-critical	Direct public safety	Very low	Unacceptable service and health risk
Heat pumps and urban electrification	Seasonal and weather-correlated	Gas-demand reduction and household decarbonization	Medium with thermal storage	Continued fossil dependence and energy poverty exposure
Electrolysers	Large and potentially flexible	Industrial hydrogen and energy-system balancing	High if economically dispatched	Slower industrial decarbonization

Italy’s Queue Creates a De Facto Priority System Before Parliament Chooses One

Italy’s allocation conflict is already underway because reservation of connection capacity creates economic rights long before electricity is consumed. Terna recorded data-centre applications of approximately 30 GW in December 2024, 69.05 GW through 411 applications at the end of 2025, and 78.79 GW through 449 applications by the end of January 2026. The increase represents far more than an informational list. Each credible application can initiate technical studies, identify reinforcement requirements, influence substation planning and affect whether subsequent projects receive an immediate connection, a delayed date or a conditional offer. Da rifiuto a risorsa: la nuova sfida della transizione ecologica e digitale – Terna – February 2026 — Official Terna analysis. Terna’s 2025 Grid Development Plan had already reported approximately 30 GW of data-centre requests and stated that new infrastructure or upgrades to existing network elements would be required. The same plan anticipates more than 65 GW of additional renewable capacity by 2030 relative to 2023, yet renewable expansion and demand connections cannot be evaluated separately: solar and wind projects also wait for connection, and their geographic distribution does not necessarily correspond to data-centre concentration around Milan. Grid Development Plan – Terna – 2025 — Official development-plan portal.

At the end of 2025, Terna reported accepted minimum technical connection solutions for new renewable plants totalling approximately 328.98 GW through 5,921 applications. The largest requested volume was in southern mainland Italy at 139.01 GW, followed by Sicily at 74.43 GW, Sardinia at 47.52 GW, the North at 37.49 GW, and the Centre at 30.54 GW. These numbers are also queues rather than construction forecasts, but their geography exposes the allocation paradox. Much of the proposed renewable power is located in southern regions and islands, while the most commercially desirable data-centre cluster remains in northern Italy. A northern data centre purchasing a southern renewable PPA does not eliminate the need for transmission capacity, balancing and residual generation. Introducing Terna – Terna – December 2025 data — Official renewable connection figures. Italy can therefore face simultaneous excess claims for generation connection in the South and insufficient demand-connection capacity in the North. The binding asset is the transport corridor between them.

Italian queue or system indicator	Official value	Allocation implication
National electricity demand, 2024	311.9 TWh	Baseline against which all electrification demand accumulates
Net domestic production, 2024	263.2 TWh	Italy remains structurally reliant on imports and balancing resources
Electricity imports, 2024	55.9 TWh	Additional demand increases exposure to neighboring systems
Gross efficient generating capacity, 2024	137.6 GW	Nameplate capacity greatly exceeds dependable and locationally deliverable capacity
Data-centre connection requests, January 2026	78.79 GW	Physically impossible as a full build-out, but capable of crowding queues
Number of data-centre applications	449	Implies average request near 175 MW
Accepted renewable connection solutions, end-2025	328.98 GW	Generation pipeline is also heavily oversubscribed
Renewable applications in South	139.01 GW	Strong geographical mismatch with northern digital demand
Renewable applications in Sicily	74.43 GW	Requires island reinforcement, storage and interconnection
Additional renewable target by 2030	>65 GW	Sufficient annual energy is possible only if projects and grids are delivered

Manufacturing Cannot Compete Only Through Willingness to Pay

Italian industrial policy officially subsidizes simultaneous digitalization and energy efficiency, demonstrating that the state expects manufacturing to consume electricity more intelligently rather than disappear from the system. The Transizione 5.0 framework originally provided a total envelope of €6.3 billion under REPowerEU and supported advanced machinery, digital systems, employee training and renewable self-generation where projects reduced energy consumption by at least 3% at facility level or 5% for the affected production process. The wider 2024–2025 transition-policy framework made €12.7 billion available across the relevant complementary measures. Demand was strong enough for the Ministry to announce exhaustion of the original resources in November 2025, while a new 2026 platform later became operational for advanced productive assets and renewable self-generation. Piano Transizione 5.0 – Ministry of Enterprises and Made in Italy – June 2026 — Official incentive framework. The PNRR investment target includes a final-energy saving of 0.16 Mtoe during 2024–2026, confirming that efficiency is being treated as additional national capacity. PNRR – Transizione 5.0 – Ministry of Enterprises and Made in Italy – May 2026 — Official PNRR measure.

The weakness in the current architecture is that industrial efficiency support and data-centre connection allocation operate through different policy channels. A manufacturer may receive an incentive to reduce its consumption by 5%, while a nearby data centre requests hundreds of megawatts of new load. The manufacturer’s efficiency saving lowers national demand but may not preserve its future right to expand, electrify heat or install a new production line. This creates a perverse dynamic: legacy industry finances or implements efficiency improvements, freeing local network headroom that can then be absorbed by higher-credit digital infrastructure. Italy needs a mechanism that recognizes saved megawatts as industrial development rights, at least for a defined period. Where a plant invests in verified efficiency, flexibility or self-generation, part of the released connection capacity could remain attached to its decarbonization plan instead of being immediately allocated elsewhere. This would transform efficiency from a one-time reduction into an instrument of reindustrialization.

A data-centre project should be ranked against manufacturing through at least six ratios:

VA/MW = Italian gross value added divided by dependable megawatts consumed

EMP/MW = direct and indirect Italian employment divided by dependable megawatts

EXP/MWh = net exports or avoided imports divided by annual electricity use

CAPEXadd/MW = additional Italian productive investment per requested megawatt

FLEX = verified interruptible share of peak demand

SOV = domestic control over technology, data, contracts and critical supply chains

Data centres can score strongly on capital expenditure and digital sovereignty where they host Italian strategic computing, but weakly on operating employment and domestic income retention when the facility functions primarily as foreign-owned colocation. Manufacturing may score strongly on employment and exports but poorly on flexibility or carbon intensity. The correct decision requires disclosure rather than categorical preference.

Transport Electrification Is a Distributed Claim That Is Easy to Underestimate

Transport electrification does not normally submit a single 500 MW connection application, which makes it less visible than a hyperscale campus. Its load emerges through thousands of charging hubs, railway traction substations, bus depots, port terminals, logistics facilities and household chargers. This distributed character can make transport appear less threatening to system planning while increasing the complexity of distribution-grid reinforcement. Italy’s official infrastructure strategy includes a national plan for charging electric vehicles intended to ensure coherent development according to local requirements. Piano nazionale infrastrutturale per la ricarica – Ministry of Infrastructure and Transport – current official programme — Official programme portal. PNRR transport investments include electric-charging infrastructure, renewal of buses and trains, rapid mass transit, digital public mobility, hydrogen experimentation and development of sustainable-mobility industrial supply chains. ItaliaDomani: il PNRR – Ministry of Infrastructure and Transport – official implementation framework — Official MIT PNRR portal.

Rail infrastructure represents a particularly important allocation claimant because it converts electricity into both passenger mobility and freight capacity. RFI’s July 2026 commercial plan describes infrastructure projects with financed implementation phases and activations predominantly scheduled between 2026 and 2030, enabling operators and territorial authorities to plan future passenger and freight services. Piano Commerciale RFI, July 2026 Edition – Rete Ferroviaria Italiana – July 2026 — Official RFI plan. RFI also launched a national framework for maintenance, renewal and upgrading of electric-traction systems during 2026–2028, showing that rail electrification creates continuing demand for electrical infrastructure rather than a one-off procurement cycle. DAC.0044.2026 – Rete Ferroviaria Italiana – July 2026 — Official procurement record.

The conflict is not necessarily zero-sum. Managed vehicle charging can absorb solar production, buses can charge outside service peaks, and railway demand is partly predictable. Data centres can also become flexible loads. The problem emerges when both are granted inflexible connection assumptions. A 500 MW AI campus operating at 70% average load consumes approximately 3.07 TWh annually. The same annual electricity could support a very large portfolio of electric vehicles or substantial rail operations, although the exact substitution depends on vehicle efficiency, distance and charging losses. Allocation cannot therefore rely only on annual energy. Transport’s load may be distributed and time-flexible, while data-centre demand is spatially concentrated and continuous. This means a megawatt assigned to transport may require more distribution investment but less firm generation, whereas a megawatt assigned to a data centre may require a smaller number of high-voltage connections but greater continuous system firmness.

Public Services Have Absolute Criticality but Weak Market Power

Hospitals, water-treatment plants, emergency communications, public transport, municipal networks and public-administration computing occupy a different position in the allocation hierarchy. Their public value is extremely high, but their ability to pay scarcity prices is constrained by regulated budgets. If network capacity is allocated primarily through commercial readiness and connection fees, essential services may receive no formal priority until a crisis occurs. Italy should therefore distinguish commercial connection ranking from emergency service ranking. A hospital expansion should not necessarily move ahead of every industrial investment in normal planning, but its critical load must receive higher reliability and restoration priority after connection. Similarly, a data centre hosting strategic health systems may deserve protected operation for specific racks, while the same campus’s commercial AI-training load should remain curtailable.

The correct architecture separates three decisions:

• Investment allocation: which new projects receive connection capacity and who pays reinforcement costs.
• Operational dispatch: which loads respond to prices, congestion or balancing instructions during normal conditions.
• Emergency priority: which services remain energized or are restored first during a severe adequacy or cyber event.

Conflating these levels creates strategic arbitrage. A data-centre owner can argue that the campus is critical because some tenants serve public institutions, thereby seeking priority for the entire building. An industrial producer can claim strategic status because its products enter defence or health supply chains, even when individual production lines can be scheduled. Public authorities need workload- and process-level classifications, verified through metering and technical separation.

Priority class	Examples	Normal-market treatment	Scarcity treatment	Emergency treatment
A₁ Irreducible public safety	Hospitals, emergency systems, water and essential communications	Standard connection planning with resilience premium	Protected except under extreme system collapse	Highest restoration priority
A₂ Strategic state and infrastructure	Sovereign cloud, grid control, rail signalling, public identity	Priority based on verified national function	Limited degradation only	Protected essential component
A₃ Strategic production	Defence inputs, medicines, selected chemicals and metallurgy	Ranked by value added and substitutability	Contracted flexibility where technically possible	Protected minimum operational level
A₄ Productive flexible demand	EV charging, electrolysers, batch industry, AI training	Competitive access with locational incentives	Curtailed or shifted in response to scarcity	First major curtailment block
A₅ Ordinary commercial demand	Non-essential digital, retail and commercial services	Standard market access	Price-responsive reduction	Curtailment before critical loads
A₆ Speculative capacity	Projects without land, finance or credible commissioning	No durable reservation	Capacity released	No priority

France Uses Surplus Low-Carbon Electricity to Avoid a Direct Zero-Sum Choice

France enters the allocation conflict from the strongest physical position among the four countries examined. RTE’s 2025–2035 outlook states that France has a favorable position for electrification because short-term consumption remains relatively stable while low-carbon production has recovered and expanded. The system can use this surplus to reduce fossil-fuel imports, reindustrialize and electrify transport and heating. Les bilans prévisionnels – RTE – 2025 — Official RTE planning portal. RTE’s earlier long-term analysis places total French electricity consumption between approximately 580 and 640 TWh in 2035 in a pathway combining rapid decarbonization and reindustrialization, compared with roughly 460 TWh at the reference point. The increase is driven by industry, mobility and data centres. Bilan Prévisionnel 2023–2035 – RTE – September 2023 — Official scenario analysis.

RTE projects data-centre consumption of approximately 15–20 TWh in 2030 and 23–28 TWh in 2035, equal to around 4% of French electricity consumption in the latter year. It also observes that recently connected data centres initially consume, on average, only about 20% of the power requested, because deployments ramp gradually and developers reserve safety margins. Les data centers en chiffres clés – RTE – 2026 — Official RTE data-centre analysis. This creates both an advantage and a governance challenge. France has sufficient low-carbon generation to accommodate meaningful digital growth without immediately sacrificing industry, but unused reserved capacity can still block other connections locally.

France’s strategic response is increasingly based on planned integration rather than passive acceptance. RTE, Data4 and Schneider Electric launched the DCFlex demonstrator at Marcoussis to test the interaction between data centres and the electricity system and assess whether facilities can provide balancing and resilience services. Lancement du premier démonstrateur européen dédié à l’interaction entre data centers et système électrique – RTE – 2026 — Official project description. The French model therefore rests on four pillars: nuclear and renewable generation surplus; identification of suitable high-voltage sites; progressive rather than nominal accounting of data-centre demand; and technical integration of data-centre flexibility. France’s principal 2031 risk is not insufficient national energy but excessive concentration at selected nodes and the possibility that privileged digital access absorbs capacity intended for reindustrialization.

Germany Treats Data Centres as Industrial Infrastructure but Couples Growth with Queue Reform

Germany approaches the conflict from almost the opposite physical position. It possesses Europe’s largest industrial electricity demand, a large existing data-centre sector, significant renewable generation and persistent concerns about electricity prices, grid congestion and dispatchable adequacy. The German federal government adopted its first national data-centre strategy in 2026 with the explicit objective of at least doubling data-centre capacity to more than 6 GW by 2030 and quadrupling AI computing capacity. The strategy contains 28 measures across energy and sustainability, sites and land, and digital sovereignty. It emphasizes reliable, affordable and increasingly climate-compatible electricity; pragmatic waste-heat requirements; and removal of speculative connection applications that lack genuine construction intent. Municipalities hosting facilities are also intended to participate more directly in business-tax revenues. Bundesdigitalminister zur Rechenzentrumsstrategie – German Federal Government – April 2026 — Official federal statement.

Germany’s model recognizes data centres as part of national industrial policy rather than merely commercial property, but it cannot rely on a French-style electricity surplus. The government must balance computing growth against electrification of chemicals, steel, automotive production, heat and transport. The key innovation is its willingness to filter speculative grid applications. A credible queue is itself an industrial asset because it allows transmission operators to plan against projects with land, financing and construction intent rather than inflated optionality. Germany is also attempting to make heat reuse more practical: rigid obligations are weakened where no realistic heat customer exists, reducing the danger that nominal sustainability conditions prevent investment without delivering useful heat.

The 2026–2031 German allocation conflict will be dominated by price competitiveness. If industrial electricity remains costly, hyperscalers may still proceed because cloud customers can absorb higher margins, whereas electro-intensive factories may reduce output or relocate. This means equal electricity prices do not produce equal economic consequences. Germany is likely to use long-term contracts, state support for industrial transformation, grid acceleration and location policy to prevent computing from crowding out manufacturing. Its comparative advantage is strong industrial integration: data centres can be connected to research, automotive, engineering and advanced manufacturing ecosystems. Its disadvantage is that every additional firm megawatt competes with a decarbonization programme already requiring vast amounts of electricity.

The United Kingdom Is Creating an Explicit Political Priority for AI

The United Kingdom has adopted the most interventionist pro-data-centre allocation model. Its 2026 Compute Roadmap forecasts a need for at least 6 GW of AI-capable data-centre capacity by 2030, approximately three times current capacity, with demand potentially exceeding the baseline if AI capabilities and adoption accelerate. The government aims to establish nationally significant sites capable of serving at least 500 MW each, with at least one AI Growth Zone scaling beyond 1 GW by 2030. It also distinguishes geographically flexible AI training from inference workloads that benefit from proximity to users and data sources. UK Compute Roadmap – UK Government – April 2026 — Official roadmap.

The AI Growth Zone criteria require access to at least 500 MW by 2030, sufficient water and discharge capacity, and at least 100 acres of developable land by 2028. Applicants must demonstrate a secured connection or credible behind-the-meter solution and provide written water-supplier confirmation. AI Growth Zones: Open for Applications – UK Government – June 2025 — Official eligibility criteria. This is a fundamentally different model from waiting for individual projects to appear in congested metropolitan queues. The state identifies zones, bundles land, power and planning, and then directs investment into them.

The November 2025 delivery framework makes the political prioritization explicit. The government intends to remove speculative demand, reserve and reallocate connection capacity for AI Growth Zones and other strategically important projects, permit developers to build high-voltage infrastructure, and reduce operating costs where data-centre demand relieves regional grid constraints. A 500 MW campus could receive support of up to £24/MWh in Scotland, £16/MWh in Cumbria, or £14/MWh in north-east England. The government estimates that its measures could reduce time to power by up to five years and save a 500 MW facility as much as £80 million annually, while unlocking up to £100 billion of investment. Delivering AI Growth Zones – UK Department for Science, Innovation and Technology – November 2025 — Official policy paper.

This policy solves one problem by creating another. Locating data centres where wind generation is curtailed can reduce system costs and support regional development. However, reserving capacity for AI is an overt political choice that can displace hydrogen, batteries, manufacturing or transport projects seeking the same northern connection points. The United Kingdom is therefore not avoiding allocation; it is ranking AI highly and attempting to choose locations where the opportunity cost is lowest. Italy has not yet articulated an equivalent doctrine.

The European Union Is Building Rules but Has Not Established a Common Priority Hierarchy

The wider European Union is constructing the regulatory foundation for allocation without deciding which sector receives priority. The Commission states that the EU aims to triple data-centre capacity by 2035. Data centres above 500 kW are subject to mandatory reporting under the Energy Efficiency Directive framework, and the Commission is developing an EU sustainability-rating scheme and minimum performance standards. The reporting system captures energy use, installed IT capacity, water, renewable energy, heat reuse and related indicators. Energy Performance of Data Centres – European Commission – 2026 — Official policy framework. Minimum Performance Standards for EU Data Centres – European Commission – 2026 — Official preparatory study.

The EU’s grid problem is substantially larger than data centres. The Commission estimates investment needs of approximately €730 billion for distribution networks and €477 billion for transmission by 2040. Its guidance on anticipatory investment seeks to permit networks to build ahead of existing connection requests while protecting affordability and industrial competitiveness. EU Guidance on Ensuring Electricity Grids Are Fit for the Future – European Commission – June 2025 — Official guidance. The earlier Commission estimate for grid investment required by the end of 2030 was €584 billion, underscoring the scale of the infrastructure gap even before later 2040 calculations. Commission Collects Views in Preparation of the European Grids Package – European Commission – May 2025 — Official consultation announcement.

The Affordable Energy Action Plan identifies high and volatile electricity prices, fossil-fuel dependence, incomplete integration and insufficient grids as threats to European industry and households. It promotes long-term contracts, lower taxation, better network charges, storage, flexibility and interconnection. The Commission states that Italy remains among the member states with relatively weak interconnection in relation to national energy requirements. Affordable Energy – European Commission – 2026 — Official strategy page. The Union is therefore moving toward three allocation principles: build networks anticipatorily; make users respond to locational and temporal conditions; and impose efficiency and transparency on large digital loads. It has not yet created a binding European taxonomy ranking data centres against steel, transport or public services. National governments retain that responsibility.

Jurisdiction	Physical starting point	Data-centre objective	Allocation mechanism	Principal protected interest	Principal 2031 risk
Italy	Import dependence, north–south mismatch, large speculative queues	No single published national capacity ceiling	Conventional connection process plus evolving grid plan	Mixed digital growth and general electrification	Queue capture and northern industrial crowding-out
France	Low-carbon generation surplus and nuclear flexibility	15–20 TWh data-centre consumption by 2030	Planned high-voltage sites and flexibility pilots	Reindustrialization plus export reduction	Local concentration despite national surplus
Germany	Large industrial demand, high price sensitivity and congested grids	>6 GW data-centre capacity by 2030	National strategy, speculative-queue filtering and site policy	AI capacity and manufacturing competitiveness	Digital load outbids electro-intensive industry
United Kingdom	Strong wind resource but severe connection queues and locational constraints	≥6 GW AI-capable capacity by 2030	AI Growth Zones, reserved capacity and regional electricity support	AI leadership and regional investment	Public subsidy privileges compute over alternative demand
European Union	Fragmented national systems and major grid-investment gap	Triple data-centre capacity by 2035	Reporting, ratings, minimum standards, grid and tariff reform	Competitiveness, decarbonization and digital sovereignty	Member-state subsidy competition and cross-border cost shifting

The Shadow Allocation Conflict: Capital, Contracts and Liquidity

Physical megawatts are only one layer. The allocation outcome is also determined by access to capital and long-duration electricity contracts. Hyperscalers can sign twenty-year PPAs, finance dedicated substations and accept take-or-pay obligations. Many Italian manufacturers—especially smaller and medium-sized companies—cannot commit to the same volumes or durations because their revenues are cyclical, their credit ratings are weaker and their future electricity demand depends on uncertain decarbonization investments. The electricity seller therefore rationally prefers the data-centre customer, even where national value added is lower. This is contractual crowding-out.

A comparison must calculate the certainty-equivalent value of each load:

CEVⱼ = Σₜ [Pⱼ,t × Qⱼ,t × CPⱼ,t − Cserveⱼ,t − ELⱼ,t] / (1 + WACC)ᵗ

where P is contracted price, Q volume, CP counterparty survival probability, Cserve the cost of supply and balancing, and EL expected loss. A hyperscaler’s high CP and stable Q can make its contract more valuable than a higher-price industrial contract. Public policy can correct this through PPA guarantees, aggregation platforms and tripartite agreements involving government, producers and industrial buyers. The Commission’s affordable-energy programme opened work on removing PPA barriers and facilitating tripartite agreements, while its April 2026 recommendation addresses access, guarantees, accounting and guarantees of origin. Actions Supporting Affordable Energy – European Commission – April 2026 update — Official implementation tracker.

The second shadow dimension is liquidity. A large data-centre announcement raises the value of land near substations and fibre routes. Developers can accumulate options before final investment, while manufacturers typically require sites adjacent to supply chains and employees and cannot relocate as easily. The third is public subsidy arbitrage. Technology investors can compare Italy, France, Germany and the United Kingdom and demand tax relief, accelerated permits or infrastructure contributions. A factory also compares jurisdictions, but governments may perceive cloud investment as politically modern and manufacturing support as defensive. The fourth is cyber and sovereignty externality: a data centre hosting foreign commercial workloads may consume Italian electricity while generating little Italian strategic capability; conversely, a sovereign computing project can reduce national dependency. The fifth is cross-border electricity externality. Italy can support new digital load through imports, but this may tighten neighboring markets during regional scarcity and transfer security risk into interconnectors.

A Defensible Italian Allocation Formula

Italy should not adopt a categorical rule that data centres always rank below industry or above it. It should establish a transparent weighted score for all large demand connections, with ministerial weights approved through public procedure and technically assessed by Terna, distribution operators, ARERA and sector ministries. A proposed Strategic Connection Allocation Score, SCAS, is:

SCASⱼ = 0.20Vⱼ + 0.15Fⱼ + 0.15Lⱼ + 0.12Aⱼ + 0.10Sⱼ + 0.10Cⱼ + 0.08Eⱼ + 0.05Wⱼ + 0.05Rⱼ

Each component is normalized from zero to 100:

• V: Italian value added and fiscal retention per dependable megawatt;
• F: verified demand flexibility and curtailment capability;
• L: locational benefit or congestion penalty;
• A: electricity additionality through new generation, storage or dedicated reinforcement;
• S: sovereignty and strategic-security contribution;
• C: counterparty and construction credibility;
• E: energy and water efficiency;
• W: employment, skills and supply-chain development;
• R: reversibility, including release of unused capacity.

A project would receive no final ranking until it passes mandatory gates for land, financing, planning, water, ultimate ownership and commissioning milestones. The score would not replace engineering feasibility; it would rank technically feasible projects when capacity is scarce.

Score component	Data-centre evidence	Manufacturing evidence	Transport/public-service evidence
Italian value added	Domestic cloud revenue, research, taxes and procurement	Output, exports, domestic supply chain and wages	Public productivity and avoided external costs
Flexibility	Workload shifting, battery dispatch and rack-level curtailment	Demand response and production rescheduling	Managed charging and operational scheduling
Locational impact	Siting near surplus generation or constrained node	Existing cluster and industrial-network compatibility	Depot, rail or port connection requirements
Additionality	Dedicated renewable, storage and grid investment	Self-generation, process efficiency and flexibility	Smart charging and distributed storage
Sovereignty	Public compute, Italian keys and domestic control	Strategic materials and defence or health supply chains	Essential service and infrastructure resilience
Credibility	Tenant, financing, equipment and phased deployment	Approved investment and customer contracts	Funded public programme and procurement
Efficiency	PUE, WUE and heat reuse	Output per MWh and emissions reduction	Passenger- or tonne-kilometres per MWh
Employment	Permanent skilled jobs and AI ecosystem	Direct, indirect and induced employment	Service access and territorial cohesion
Reversibility	Capacity-release milestones	Reallocation of unused industrial capacity	Modular expansion and controllable charging

Bayesian Outlook, 2026–2031

Five competing hypotheses define the next phase. H₁ — Market-Led Digital Priority assumes hyperscalers secure the best northern connections because of superior credit, speed and willingness to finance infrastructure. H₂ — Strategic Rebalancing assumes Italy creates formal ranking criteria that protect manufacturing, public services and transport while approving high-additionality digital projects. H₃ — Infrastructure Expansion Resolves the Conflict assumes grids, renewables and storage grow fast enough that allocation becomes less politically contentious. H₄ — Industrial Crowding-Out assumes data-centre expansion, high electricity costs and connection delays materially reduce manufacturing electrification. H₅ — Queue Correction and Project Attrition assumes most announced data-centre capacity disappears, leaving only a manageable group of projects.

The 2026 posterior judgment is:

Hypothesis	Probability	Core supporting evidence	Principal disconfirming indicator
H₁ Market-led digital priority	26%	Hyperscaler credit strength, rapid queue expansion and political attraction of AI investment	Formal national sector-ranking rules
H₂ Strategic rebalancing	24%	European queue reforms and growing recognition of industrial competitiveness	Continued first-come allocation without public scoring
H₃ Infrastructure resolves conflict	15%	Terna investment and >65 GW renewable target	Persistent permitting and transformer delays
H₄ Industrial crowding-out	20%	Northern concentration, import dependence and industrial price sensitivity	Strong manufacturing connection protection and PPA support
H₅ Queue correction	15%	French evidence of low initial utilization and obvious overstatement of Italian queue	Rapid final investment decisions and hardware procurement

These probabilities are analytical judgments, not official forecasts. H₁ and H₂ together represent the most likely political contest: either allocation continues through the economic power of applicants or the state creates an explicit hierarchy. H₃ is possible but insufficient as a standalone strategy because infrastructure cannot be delivered instantaneously and new capacity will attract additional demand. H₄ remains a serious tail because industrial closure can occur before new grids are commissioned. H₅ will certainly occur to some degree, but queue attrition does not eliminate congestion if the surviving projects remain concentrated.

Year	Primary allocation event	Italy	France	Germany	United Kingdom	EU-wide implication
2026	Queue validation	Pressure to distinguish mature from speculative projects	Fast-track sites and DCFlex development	National strategy begins filtering applications	Growth Zone capacity reservation	Data-centre rating and grid reforms advance
2027	Contract competition	Hyperscalers and industry compete for PPAs and firming	Surplus supports electrification	Industrial electricity cost remains decisive	Regional tariff support begins influencing siting	Switching, PPA and network-charge reforms deepen
2028	First concentrated bottlenecks	Milan and selected nodes face visible sequencing conflicts	Local rather than national congestion	Heat reuse and site acceptance tested	Large zones move from planning to construction	Cross-border equipment shortages and subsidy competition
2029	Operational flexibility becomes valuable	Curtailment and storage obligations enter contracts	Data-centre grid services expand	Flexible computing becomes part of adequacy strategy	Zones tested against network constraints	Common minimum performance standards mature
2030	Strategic hierarchy becomes unavoidable	Industry, transport and compute need formal ranking	Data-centre use reaches 15–20 TWh pathway	>6 GW target tested against industrial demand	≥6 GW AI-capable target and ≥1 GW zone	Capacity tripling trajectory accelerates
2031	Systemic outcome visible	Either managed digital-industrial integration or northern crowding-out	France retains comparative power advantage	Germany’s competitiveness depends on price reform	UK has strongest AI priority but highest policy concentration	EU faces divergent national priority systems

Strategic Judgment

Italy cannot copy France because it does not possess the same low-carbon generation surplus. It cannot copy Germany without accepting a far more explicit national industrial strategy and aggressive queue discipline. It cannot copy the United Kingdom without deciding politically that AI infrastructure deserves reserved capacity, accelerated planning and potentially differentiated electricity costs. It cannot rely solely on the European Union because EU rules improve reporting, grids and efficiency but leave the core allocation decision to national governments.

The optimum Italian model is a conditional strategic allocation regime. Data centres should receive accelerated access where they satisfy five conditions simultaneously: they add new electricity generation or storage; locate where they reduce rather than intensify network congestion; provide technically verifiable demand flexibility; retain material computing, fiscal and knowledge value in Italy; and finance attributable reinforcement and water infrastructure. Manufacturing should receive protected development capacity where it demonstrates decarbonization investment, export value, strategic supply-chain importance and efficiency improvement. Transport and public services should receive long-term network reservations based on approved infrastructure plans rather than having to compete project by project with commercial applicants. Unused capacity should expire through staged milestones.

The allocation rule can be summarized as:

Priorityⱼ = Technical feasibility × Strategic additionality × Locational compatibility × Delivery probability

A project with zero technical feasibility receives zero priority regardless of political importance. A project with no strategic additionality should not receive subsidized scarcity. A project in the wrong location should relocate or finance the network needed to correct its impact. A speculative project should not reserve capacity merely because its application was filed first.

By 2031, electricity will function as a form of productive sovereignty. The countries that allocate it only through queues and wholesale prices will discover that the most creditworthy global firms obtain the best infrastructure while nationally embedded sectors absorb delays and system costs. The countries that politicize every connection will suppress investment and technological change. Italy’s strategic task is to create a rule-based intermediate system in which price, engineering credibility and national value are evaluated together. The conflict cannot be eliminated. It can only be made transparent, measurable and consistent with the country’s industrial future.

---

Copyright of debuglies.com - Even partial reproduction of the contents is not permitted without prior authorization – Reproduction reserved