Executive Summary

West Asia is acquiring computing capacity faster than it is acquiring digital sovereignty.
Server localization reduces latency and some transfer risks but does not neutralize foreign jurisdiction.
The decisive control points are corporate governance, encryption keys, identity systems, software updates, AI models and contractual exit rights.
Saudi Arabia and the UAE are becoming major compute hubs through investments involving US hyperscalers.
Those partnerships accelerate capability while embedding external platforms more deeply into critical national systems.
The most probable 2030 outcome is neither full dependence nor technological autonomy, but a segmented dual-stack architecture.
Critical government workloads will increasingly demand sovereign controls; commercial systems will remain hyperscaler-led.
Sovereignty will depend less on where servers stand than on whether states can operate, audit, migrate and legally defend workloads without foreign permission.

West Asia’s Cloud Boom Has a Sovereignty Problem

West Asia is building the infrastructure of an artificial-intelligence economy at historic speed. Data centres, national clouds and advanced computing clusters promise productivity, strategic diversification and new influence over global technology flows. Yet the decisive question is no longer where a server stands. It is who controls the encryption keys, identity system, software updates, AI models and legal entity operating it. A country may localise public data while remaining dependent on foreign platforms whose parent companies, intellectual property and compliance obligations sit abroad. Between 2026 and 2030, this distinction will separate nominal data residency from credible digital sovereignty. The region’s challenge is not to expel global technology companies, but to convert investment into enforceable authority, operational resilience and a genuine capacity to continue governing when foreign access, support or political consent is withdrawn.

The French Test

The sovereignty gap became unusually visible in Paris on 10 June 2025. During an inquiry into public procurement, French senator Dany Wattebled asked Anton Carniaux, Microsoft France’s director of public and legal affairs, whether he could guarantee that data belonging to French citizens would never be transmitted to the United States without the French authorities’ explicit consent. His answer was unequivocal: he could not provide that guarantee. The exchange is recorded in the French Senate’s official hearing transcript and reproduced in its final report on public procurement.

The significance extends far beyond Microsoft or France. The Senate established, in institutional language, that territorial storage does not necessarily confer exclusive national control. Servers may operate in France, Riyadh, Abu Dhabi or Doha while the service remains governed through a foreign corporate group, remotely administered software, proprietary identity architecture and legal duties attaching to the provider. The physical asset is local; decisive authority can remain extraterritorial.

This is the central strategic error in many localisation policies. Data residency answers where information is stored. Sovereignty requires answers to harder questions: who can decrypt it, authenticate administrators, alter the platform, inspect metadata, discontinue service, compel disclosure and restore operations without the original supplier?

Capital Without Control

The Gulf has acquired genuine infrastructural weight. The World Bank counted 39 data centres in the United Arab Emirates and 33 in Saudi Arabia as of June 2025. The corresponding average was 81.34 in high-income economies, or 53.7 when the United States was excluded. The figures place the two Gulf states far ahead of most neighbouring economies, but they also reveal the distance separating regional ambition from the mature capacity of established digital powers. The assessment appears in The Gulf’s Digital Transformation – World Bank – December 2025.

Investment is accelerating. On 30 October 2024, Saudi Arabia’s Public Investment Fund and Google Cloud announced an advanced AI hub near Dammam. A preliminary study commissioned by the partners estimated that the initiative could add $71 billion to Saudi GDP over eight years. The project envisages Arabic-language models, large-scale computing infrastructure and training programmes reaching millions of students and professionals. It is an industrial-policy instrument aligned with economic diversification—but also a long-term relationship with a foreign platform owner. The terms and projected economic contribution were published in PIF and Google Cloud to Create Advanced AI Hub in Saudi Arabia – Public Investment Fund – October 2024.

In the UAE, the scale is larger still. On 3 November 2025, Microsoft disclosed plans to invest $15.2 billion in the country between 2023 and 2029. It reported $7.3 billion committed or invested during 2023–2025: $1.5 billion for an equity stake in G42, more than $4.6 billion in advanced cloud and AI data-centre capital expenditure, and over $1.2 billion in local operating expenditure and cost of goods sold. A further $7.9 billion was planned for 2026–2029, including more than $5.5 billion in infrastructure capital expenditure. Microsoft president Brad Smith presented the programme to Sheikh Khaled bin Mohamed bin Zayed Al Nahyan, Crown Prince of Abu Dhabi, alongside Khaldoon Khalifa Al Mubarak and G42 chief executive Peng Xiao. The figures are detailed in Microsoft’s UAE Investment Plan – Abu Dhabi Media Office – November 2025.

These investments are economically substantial. They do not, by themselves, prove sovereign capability. Capital can localise hardware while control remains concentrated in software, licences and remote administration.

The Control Plane

A cloud service is not merely a warehouse of machines. It is a hierarchy of dependencies. At the bottom are land, power, cooling systems, fibre routes and servers. Above them sit virtualisation software, orchestration tools, security monitoring, application interfaces and databases. Higher still are encryption-key management, identity and access controls, proprietary AI models, model-safety systems and vendor support. The top layer consists of contracts, export licences, intellectual-property rights and the laws governing the provider.

The upper layers dominate the lower ones. A government may own the building yet lack access to the source code. It may hold a contractual right to its information yet depend on the vendor’s authentication service to reach it. It may possess encryption keys while relying on foreign firmware, security patches or specialised personnel. It may train a national model whose inference engine, accelerators and development framework remain controlled elsewhere.

Identity is especially critical. Whoever governs privileged credentials can determine who enters the system, which actions are logged and whose access is revoked. Encryption is equally decisive: customer-held keys improve protection only if backup procedures, hardware-security modules and administrative identities cannot be overridden through another dependency. AI adds a further asymmetry because national data may generate value inside models whose weights, training methods and deployment restrictions remain proprietary.

The correct unit of analysis is therefore not the data centre. It is the complete control plane from electricity supply to legal compulsion.

Jurisdiction Travels

The United States’ CLOUD Act illustrates why geography alone cannot settle authority. The US Department of Justice explains that a communications or remote-computing provider subject to American jurisdiction may be required, through valid legal process, to produce information within its “possession, custody, or control” regardless of whether the information is stored inside or outside the United States. The statute did not dispense with judicial standards or authorise indiscriminate collection; it clarified the territorial reach of lawful orders directed at qualifying providers. That distinction is legally important but strategically uncomfortable for states that equate domestic storage with immunity from foreign process. See The Purpose and Impact of the CLOUD Act – US Department of Justice – April 2019.

The underlying principle is not uniquely American: jurisdiction can follow corporate control. The practical exposure depends on ownership, corporate structure, possession of the data and the provider’s technical ability to retrieve it. A locally incorporated subsidiary does not automatically sever those connections. Nor does a sovereign-cloud label guarantee that foreign parent-company personnel, intellectual property or support systems have ceased to matter.

West Asian governments must therefore treat legal architecture as part of critical infrastructure. Procurement teams require conflict-of-laws assessments before deployment, not after a disclosure demand, sanctions decision or diplomatic rupture. Contracts must identify every entity with administrative access, every jurisdiction touching the service and every circumstance in which the customer may be notified—or legally prevented from being notified.

The Exit Test

The strongest measure of sovereignty is not localisation but substitutability. Can the state transfer workloads, identities, logs, keys and trained models to another environment within a defined period and at a tolerable cost? Can essential services continue if licences are revoked, updates cease or foreign specialists become unavailable? If not, the system is resident but captive.

Qatar’s official Cloud First Policy already recognises this problem. It requires government entities to prepare exit strategies, retain ownership of their data, favour standard formats and test the continuing viability of the chosen cloud arrangement. It also distinguishes highly sensitive national-security information from ordinary workloads. These provisions make Cloud First Policy – State of Qatar – 2022 more than an adoption directive: they treat reversibility as a governance obligation.

Saudi Arabia’s data regime similarly places substantive conditions on cross-border transfers. Its Personal Data Protection Law states that transfers must not prejudice national security or vital interests and must provide an appropriate level of protection. The importance of the Personal Data Protection Law – Saudi Data and Artificial Intelligence Authority – April 2023 lies in linking data policy to sovereign risk. Yet transfer restrictions cannot eliminate dependency if the domestic platform still requires foreign-controlled software, maintenance or computing components.

Europe’s Regulatory Signal

Europe offers a useful benchmark, not a complete solution. The EU Data Act entered into force on 11 January 2024 and has applied since 12 September 2025. It establishes requirements concerning switching between data-processing services, interoperability and safeguards against unlawful third-country governmental access to non-personal data. Its strategic contribution is to redefine cloud competition around portability and control rather than server location alone. The operative text is Regulation (EU) 2023/2854 – European Union – December 2023.

For West Asia, the lesson is not to reproduce European regulation mechanically. It is to require measurable exit performance. Public tenders should specify maximum migration periods, machine-readable export formats, transfer of operational documentation, portability of security logs, continuity of identity systems and escrow arrangements for critical software. Penalties should attach to failed exit tests as directly as they attach to downtime. Sovereignty must become a service-level obligation that auditors can verify.

Energy Becomes Strategy

Digital autonomy also has a physical balance sheet. The International Energy Agency projects global data-centre electricity consumption to more than double to approximately 945 terawatt-hours by 2030, with demand rising by around 15 per cent annually from 2024. Electricity used by accelerated servers, the machines central to AI workloads, is projected to grow by about 30 per cent annually. These projections appear in Energy and AI – International Energy Agency – April 2025.

The regional constraint is sharper because computing will compete with cooling, desalination and industrial expansion. The IEA expects electricity demand across the Middle East and North Africa to rise by 50 per cent by 2035, with cooling and desalination accounting for roughly 40 per cent of the increase. See The Future of Electricity in the Middle East and North Africa – International Energy Agency – January 2025.

A sovereign-computing strategy that ignores generation, grids, water efficiency and backup power is therefore incomplete. The competitive asset will not be the data-centre announcement but the dependable megawatt: low-cost electricity delivered through resilient networks under conditions of extreme heat and geopolitical stress.

The 2030 Settlement

Between 2026 and 2030, total technological autarky is neither realistic nor economically desirable. The probable settlement is managed dependence: global platforms for scale, domestic or jointly governed systems for strategic workloads, and interoperable infrastructure linking the two. The decisive question is whether “dual stack” means genuine redundancy or merely two foreign dependencies operating in parallel.

Credible sovereignty requires national control over root identities and critical encryption keys; locally accountable security operations; access to operational telemetry; tested portability of applications and models; diversified accelerators and software frameworks; domestic engineering depth; and contractual protection against unilateral discontinuation. The most sensitive public, defence, energy, health and identity systems require an independently operable stack, even if less critical workloads remain on commercial hyperscale platforms.

West Asia possesses capital, energy resources, state coordination and rapidly expanding demand. Its weakness lies in the layers where cumulative intellectual property and global scale create lock-in. The region should use its purchasing power to demand joint control, open interfaces, skills transfer and enforceable reversibility—not simply additional server capacity.

The sovereignty gap will not be closed by flags on data-centre façades. It will close when a government can demonstrate, under audit and during crisis, that no foreign corporation or jurisdiction can unilaterally read, alter, suspend or immobilise its essential digital systems. Until that standard is met, localisation remains an address. Sovereignty is authority.

Saudi Arabia’s War Pivot: Deterrence or Entrapment?

---

Navigational Index

1. The sovereignty gap — Why local data residency does not establish exclusive national authority.
2. The dependency architecture — Platforms, keys, identity planes, AI models, legal compulsion and vendor exit.
3. The 2026–2030 contest — Managed dependence, dual-stack systems and credible sovereign capability.

---

Master Abstract

The central analytical finding is that digital sovereignty is a control system, not a storage policy. The French Senate hearing of 10 June 2025 provides unusually direct evidence. Questioned about whether Microsoft could guarantee that data hosted in France would never be transmitted to a foreign authority without French governmental authorization, Microsoft France’s legal representative declined to provide such an absolute guarantee. The hearing was not itself proof that French-hosted information had been improperly transferred; it established the narrower—but strategically decisive—fact that territorial hosting did not eliminate every jurisdictional pathway capable of reaching the provider. The Senate subsequently characterized the use of Microsoft Azure for France’s health-data platform as exposure to American extraterritorial legislation and described the choice as a serious sovereignty failure. CE Commande publique: audition de Microsoft France – Sénat – June 2025 — verified official hearing record. L’urgence d’agir pour éviter la sortie de route: piloter la commande publique au service de la souveraineté – Sénat – July 2025 — verified official report. The underlying jurisdictional mechanism is equally explicit. US law permits providers subject to American jurisdiction to be compelled through valid legal process to produce information within their possession, custody or control, irrespective of the storage location. The statute does not create unrestricted intelligence access, eliminate judicial process or make every foreign dataset automatically obtainable; it nevertheless means that data localization alone cannot remove the provider’s American legal nexus. The Purpose and Impact of the CLOUD Act – US Department of Justice – April 2019 — verified official explanation. For West Asian governments, the relevant sovereignty test must therefore examine at least seven layers: physical infrastructure, corporate control, encryption-key custody, privileged administration, identity management, software and model dependence, and the enforceability of migration or service-continuity rights.

The region’s investment trajectory magnifies this distinction. Saudi Arabia and the United Arab Emirates are not passive technology markets: both are deploying capital, energy, land, sovereign investment vehicles and regulatory authority to become globally relevant compute jurisdictions. Yet a substantial portion of their new capacity is being built through firms whose foundational platforms, intellectual property, security updates and ultimate corporate obligations remain external. AWS states that its Saudi region is planned with an investment of US$5.3 billion, alongside existing Middle East regions in Bahrain and the UAE. How AWS Can Help Partners Grow in the Middle East – Amazon Web Services – March 2025 — verified corporate source. Google Cloud and Saudi Arabia’s Public Investment Fund announced a US$10 billion joint commitment to an AI hub launched with HUMAIN; the same announcement estimates benefits to American companies and employment, illustrating that localization simultaneously advances Saudi capacity and the commercial-strategic position of the foreign technology supplier. Google Cloud and PIF Advance AI Hub in Saudi Arabia – Google Cloud and PIF – May 2025 — verified corporate announcement. Microsoft has confirmed that its Saudi Arabia East region, comprising three availability zones, is scheduled to accept customer workloads from the fourth quarter of 2026. Microsoft Confirms Saudi Arabia Datacenter Region Available from Q4 2026 – Microsoft – February 2026 — verified corporate announcement. In the UAE, Microsoft reports planned expenditure of US$15.2 billion between 2023 and 2029, including a US$1.5 billion equity investment in G42, more than US$4.6 billion in AI and cloud capital expenditure through 2025, and more than US$5.5 billion in additional infrastructure capital expenditure during 2026–2029. Microsoft’s US$15.2 Billion Investment in the UAE – Microsoft – November 2025 — verified corporate disclosure. These figures demonstrate real localization of capital and capacity; they do not, by themselves, demonstrate autonomous control of the complete technology stack.

The Dual-Frontier Strategy: Assessing the Geopolitical and Technological Calculus of Saudi Arabia’s Investment in European Defense Capabilities – The Leonardo case

The five-year outlook is best represented through five competing hypotheses. H₁ — sovereign substitution assumes that regional states convert current infrastructure investment into domestically controlled platforms, models, cryptographic systems and operational talent. H₂ — managed hyperscaler dependence anticipates continued foreign-platform dominance, moderated by localization requirements, contractual protections and national regulators. H₃ — dual-stack sovereignty predicts separation between highly sensitive government or defense workloads and hyperscaler-operated commercial environments. H₄ — coercive platform exposure assumes that geopolitical confrontation, sanctions, export controls, corporate policy enforcement or legal orders reveal concentrated external leverage. H₅ — multipolar diversification anticipates that American, Chinese, European and domestic providers create sufficient redundancy to reduce any single state’s control. Based on the verified investment structure, current supplier relationships and the difference between residency and operational independence, the initial Bayesian assessment assigns the highest posterior probability to H₃ at 42%, followed by H₂ at 29%, H₅ at 14%, H₄ at 10% and H₁ at 5%. These are structured analytical estimates—not reported facts—and must be updated when evidence emerges concerning sovereign key custody, source-code access, domestic chip supply, platform interoperability, procurement rules and tested exit capability. A preliminary Monte Carlo design for Chapter 1 will vary foreign-platform concentration, extraterritorial exposure, domestic technical depth, workload portability, sanctions intensity and regional conflict shocks across correlated distributions. The decisive indicator will not be installed megawatts or the number of data centers. It will be continuity under denial: whether a state can keep critical systems operating, secure identities and keys, update software, maintain AI inference, investigate abuse and transfer workloads when a foreign provider, jurisdiction or supply chain becomes unavailable. Regional connectivity continues to expand—the ITU places internet use in the Arab States at 70% in 2025—but widespread consumption of digital services does not establish control over their underlying architecture. Measuring Digital Development: Facts and Figures 2025 – International Telecommunication Union – November 2025 — verified official report.

The Sovereignty Gap: When Local Clouds Remain Foreign-Controlled

Territory is not authority

The foundational error in contemporary cloud policy is the assumption that the physical location of a server determines the political location of the authority governing it. That assumption collapses at the moment a foreign-controlled provider retains possession, custody, administrative access or legally cognizable control over the hosted information. The French Senate established this distinction with unusual precision during its 10 June 2025 examination of Microsoft France. Asked under oath whether Microsoft could guarantee that French citizens’ data entrusted to the company would never be transmitted following a United States governmental order without explicit French approval, Anton Carniaux answered: “Non, je ne peux pas le garantir.” The statement did not prove that Microsoft routinely exports locally hosted data, nor that American authorities can obtain it without legal procedure. It proved something more structurally important: a supplier’s promise of local residency cannot eliminate every obligation generated by the jurisdiction governing the supplier. CE Commande publique: compte rendu de la semaine du 9 juin 2025 – Sénat – June 2025 — official verified hearing record. The Senate’s subsequent report concluded that Microsoft remained unable to guarantee immunity from foreign-government access and documented how France’s supposedly temporary reliance on Azure for its health-data platform developed into lasting dependence because the planned reversibility strategy was never implemented. Piloter la commande publique au service de la souveraineté – Sénat – July 2025 — official verified report. The relevant American rule is equally explicit but frequently overstated. The CLOUD Act does not authorize arbitrary, universal or warrantless collection of every dataset held by a US provider. It clarifies that a communications-service provider subject to US jurisdiction may, following valid legal process, be required to produce responsive information under its control regardless of where it stores that information. The Purpose and Impact of the CLOUD Act – US Department of Justice – April 2019 — official verified white paper. Consequently, residency answers “where are the bytes?” while sovereignty must answer “who can compel, decrypt, administer, interrupt, modify, monetize or relocate them?”

The seven-layer control structure

Exclusive national authority exists only when the state, or entities exclusively accountable to it, can control all material layers required to operate the digital service. The first layer is facility control: ownership of the land, building, power supply, cooling systems and physical access mechanisms. The second is hardware control, including servers, accelerators, network interfaces, firmware and replacement inventory. The third is the virtualization and orchestration plane, where the provider creates virtual machines, containers, storage accounts and policy objects. The fourth is the identity and privileged-access plane, which determines who can authenticate administrators, issue credentials, reset accounts, recover tenants and change access policies. The fifth is cryptographic authority: who generates, stores, rotates, backs up and can ultimately retrieve encryption keys. The sixth is the software and AI dependency layer, encompassing source code, proprietary APIs, security patches, model weights, safety filters, inference services, telemetry and licensing. The seventh is legal-corporate control, including the parent company’s jurisdiction, contractual governing law, sanctions exposure, export-control obligations, subcontractors and capacity to challenge governmental demands. A data center can therefore be entirely local at layer one while remaining externally dependent at layers three through seven. “Bring your own key” arrangements reduce exposure only if the customer exclusively controls the root material, the provider cannot silently redirect decryption, backups do not introduce alternative keys, support personnel cannot elevate privileges, and essential services remain functional when the external key-management service is unavailable. Confidential-computing environments similarly reduce some categories of provider access but cannot independently solve licensing, update, identity, availability or legal-compulsion risks. Even technically strong protection can be undermined by metadata: tenant identifiers, access logs, billing records, IP histories, administrator activity, model prompts, abuse-monitoring outputs and service telemetry may reside in systems distinct from the principal customer database. Sovereignty must therefore be measured as a composite capability rather than certified through a binary statement that information “remains in-country.”

Control layer	What localization can establish	What localization cannot establish	Decisive sovereignty test
Physical facility	Domestic server location and physical inspection	Control of software, keys or foreign parent company	Can national authorities operate the facility independently?
Hardware	Local possession of servers and accelerators	Firmware transparency or replacement supply	Can components be maintained during export denial?
Orchestration	Regional workload execution	Independence from proprietary control planes	Can workloads run if the global control plane is disconnected?
Identity	Local authentication endpoints	Exclusive control of privileged recovery	Can the foreign supplier create or restore administrator access?
Cryptography	Encryption at rest and in transit	Exclusive control of all keys and backups	Can anyone outside national jurisdiction decrypt the data?
AI and software	Low-latency access to models and tools	Ownership of weights, patches, APIs or safety controls	Can the service be maintained without the original vendor?
Legal-corporate	Contractual commitments and notification clauses	Immunity from mandatory foreign law	Which court or authority can ultimately compel the provider?
Exit capability	Nominal contractual portability	Functional migration at operational scale	Has a full migration been tested within a fixed deadline?

West Asia’s legal firewall—and its limits

Saudi Arabia’s regulatory architecture illustrates both the seriousness of regional data-sovereignty efforts and the boundary between legal protection and technological control. The Saudi Personal Data Protection Law applies to processing undertaken inside the Kingdom and, through its territorial scope, to certain processing of the personal data of individuals residing in Saudi Arabia by parties located abroad. The official guidance expressly uses the example of a US-hosted cloud solution serving Saudi users to demonstrate that offshore storage does not remove the processing from the Saudi law’s scope. Personal Data Protection Law Compliance Guide – Saudi Data and AI Authority – 2024/2026 edition available on the official platform — official verified guidance. Saudi implementing rules also require controllers transferring personal data outside the Kingdom to assess protection levels, minimize the information transferred, consider effects on national security and vital interests, apply safeguards, and stop transfers when prescribed high-risk conditions arise. Implementing Regulation of the Personal Data Protection Law and Regulation on Personal Data Transfer Outside the Kingdom – Saudi Data and AI Authority – September 2023 — official verified regulatory text. The UAE’s federal framework likewise regulates cross-border transfers and processing safeguards. Data Protection Laws – Official Platform of the UAE Government – December 2025 — official verified government overview. These rules create national claims over processing, disclosure and transfer, but they cannot automatically nullify contradictory obligations imposed on a foreign supplier by its home jurisdiction. If Saudi or Emirati law prohibits disclosure while another jurisdiction orders the parent company to produce data under its control, the result is a conflict-of-laws problem. Contractual assurances may allocate liability and require notice; they cannot guarantee that a foreign court will accept the contract as a defense against mandatory public law.

Regional control instrument	Immediate benefit	Residual gap	Required technical counterpart
Domestic storage mandate	Reduces routine offshore replication	Foreign provider may retain legal or administrative control	Independent privileged-access controls
Cross-border transfer assessment	Forces documentation and risk review	Undisclosed legal demands may remain possible	Tamper-evident access and disclosure logs
National-security limitation	Creates a domestic legal prohibition	Does not automatically bind a foreign court	Customer-exclusive encryption keys
Adequacy or safeguards mechanism	Structures lawful international transfers	Protection depends on enforcement and recipient behavior	Continuous compliance verification
Local corporate subsidiary	Creates a domestic contracting entity	Parent-company influence may remain decisive	Operational and governance separation
Government cloud designation	Separates sensitive workloads	May retain proprietary foreign software dependencies	Tested offline and degraded-mode operation
Exit clause	Creates a contractual migration right	Migration may be slow, costly or technically incomplete	Regular, measured portability exercises

The Chinese, Russian and European comparison

The multilingual legal comparison shows that localization is neither a uniquely West Asian response nor a uniform doctrine. China’s Data Security Law defines data processing broadly to include collection, storage, use, processing, transmission, provision and disclosure; it also asserts legal responsibility for certain processing activities conducted outside China when they harm Chinese national security, public interests or the lawful interests of Chinese persons and organizations. 中华人民共和国数据安全法 [Data Security Law of the People’s Republic of China] – Cyberspace Administration of China/National People’s Congress – June 2021 — official verified Chinese text. China’s framework therefore combines localization and transfer governance with its own extraterritorial claim. Official Chinese interpretation of the Personal Information Protection Law identifies security assessment, protection certification and standard contractual mechanisms as alternative pathways for cross-border personal-information transfers. 专家解读｜个人信息保护法的深远意义：中国与世界 [Expert Interpretation of the Far-Reaching Significance of the PIPL] – Cyberspace Administration of China – August 2021 — official verified Chinese source. Russia’s Federal Law No. 242-FZ requires relevant operators to localize databases containing Russian citizens’ personal data, while Roskomnadzor’s official implementation material confirms that the obligation extends to databases containing categories such as medical information. Федеральный закон от 21.07.2014 № 242-ФЗ [Federal Law No. 242-FZ] – Official Publication of Russian Legal Acts – July 2014 — official verified Russian legislation. The European Union has adopted a different combination: Article 32 of the Data Act requires providers of data-processing services to take technical, organizational and legal measures against unlawful third-country governmental access to or transfer of non-personal data held in the Union. The same regulation contains provisions intended to reduce switching obstacles and improve interoperability. Regulation (EU) 2023/2854 on Harmonised Rules on Fair Access to and Use of Data – European Parliament and Council – December 2023 — official verified EU legal text. These regimes differ politically, but all implicitly recognize that storage location is only one element of authority; each supplements geography with jurisdiction, authorization, provider obligations or technical control.

The hidden planes of access

The sovereignty gap is widest in systems where attention is concentrated on database location while the identity plane, management plane and software-supply chain remain transnational. A national ministry may encrypt a locally hosted database and still rely on the foreign provider’s directory service for user authentication, certificate service for machine identity, global telemetry system for anomaly detection, proprietary repository for software packages and external support organization for emergency recovery. If the provider can disable a tenant, revoke a license, block an API, change an authentication policy or withdraw a critical security update, the state lacks exclusive operational authority even if no data leaves the country. This is particularly consequential for AI. A government may retain its documents locally while submitting prompts, embeddings, retrieval queries or model-evaluation records to a managed inference environment. The sensitive asset is then not merely the original document: it includes the derived vector representation, system prompt, retrieval graph, fine-tuning corpus, model adapters, output filters and usage telemetry. Control over the model gateway can shape what the system answers, which requests it refuses, how activity is logged and whether the service remains available. A sovereign deployment must consequently distinguish at least three cryptographic states: data at rest, data in transit and data in use. Conventional encryption protects the first two more effectively than the third because computation ordinarily requires plaintext or an accessible execution state. Confidential-computing technologies can reduce this exposure, but their value depends on attestation integrity, processor supply, firmware trust and whether the surrounding orchestration system remains externally controlled. The French Senate record is important precisely because it cuts through the vocabulary of “trusted,” “regional” and “sovereign-ready” services. A cloud may be technically hardened and commercially reliable while still failing the strongest sovereignty test: whether an outside actor retains a legally or operationally effective route to compel, alter, suspend or observe the service.

Sovereignty object	Frequently overlooked exposure	Evidence required for assurance	Failure indicator
Human identities	Foreign-controlled directory or recovery service	National root administrator and independent audit trail	External account suspension disables operations
Machine identities	Vendor-managed certificates and secrets	Domestic certificate authority and key lifecycle	Services cannot authenticate during disconnection
Encryption	Provider-generated or escrowed keys	Customer-generated roots in nationally controlled hardware	Provider can recover plaintext independently
AI inference	Prompts, embeddings and telemetry	Local processing map and retention controls	Derived data enters global monitoring systems
Model operation	Proprietary weights and external safety layers	Deployable weights or guaranteed offline runtime	API withdrawal terminates critical functions
Updates	Foreign signing infrastructure	Verified local repository and rollback capability	Security patches require external authorization
Backups	Replicas outside the declared region	Complete topology and deletion verification	Residual copies survive national deletion order
Incident response	Foreign support and privileged engineers	Cleared domestic operations team	Critical recovery depends on external personnel

Lock-in, liquidity and the contractor shadow

The economic dimension converts technical dependence into durable political leverage. Cloud migration rarely consists of moving interchangeable virtual machines between equivalent facilities. Organizations progressively adopt proprietary databases, identity services, security tooling, serverless functions, data warehouses, AI development environments and observability systems. Each adoption generates a new migration cost, and the combined cost increases non-linearly because applications become interdependent. The customer’s expenditure therefore creates supplier-specific intangible capital: staff certifications, operating procedures, automation scripts, data schemas, security rules and procurement frameworks optimized for one ecosystem. Microsoft’s audited fiscal-year 2025 filing reported that Azure exceeded US$75 billion in annual revenue, while the company identified continuing investment in cloud and AI engineering and infrastructure as central to its strategy. Microsoft Corporation Form 10-K for Fiscal Year 2025 – Microsoft/US Securities and Exchange Commission – July 2025 — official verified filing. The significance for West Asia is not simply that foreign providers are large. It is that hyperscale economics allow them to spread model development, cybersecurity, accelerator procurement and software engineering across a global customer base, whereas a national alternative must finance similar capabilities from a smaller domestic market. Sovereign funds can close part of the capital gap, but money alone does not immediately reproduce mature developer ecosystems, vulnerability-response capacity or accumulated operational knowledge. The contractor “shadow” deepens this dependency. Systems formally owned by governments may be designed, configured, audited and maintained by multinational consultancies and subcontractors whose personnel, remote-access procedures and software dependencies span several jurisdictions. This is the digital equivalent of reliance on external security contractors: formal sovereignty remains with the state, while practical readiness resides in a distributed commercial network. Liquidity flows reinforce the structure because public investment, foreign direct investment, cloud credits and venture funding encourage local startups to build directly on hyperscaler services. The resulting ecosystem can generate genuine domestic economic value while simultaneously increasing the cost of strategic separation.

Competing hypotheses and Bayesian update

An Analysis of Competing Hypotheses produces five credible trajectories rather than a single deterministic claim. H₁, “residency becomes sovereignty,” predicts that local regions, national regulations and domestic subsidiaries will progressively eliminate material foreign leverage. H₂, “managed dependence,” predicts that West Asian governments will accept continuing hyperscaler control because economic and operational benefits exceed the perceived jurisdictional risk. H₃, “dual-stack sovereignty,” predicts that defense, intelligence, health, identity and core government systems will migrate toward nationally controlled or strongly isolated environments while ordinary commercial workloads remain on foreign platforms. H₄, “fragmented multipolarity,” predicts diversification across American, Chinese, European and domestic stacks, reducing single-provider concentration but increasing integration and cyber-governance complexity. H₅, “coercive exposure,” predicts that sanctions, export controls, conflict, foreign legal orders or provider-policy enforcement will reveal that apparently local infrastructure remains externally contestable. The initial priors used here were deliberately broad: H₁ 15%, H₂ 30%, H₃ 25%, H₄ 20%, H₅ 10%. The French Senate evidence decreases H₁ because it directly demonstrates that localization cannot generate an absolute jurisdictional guarantee. Saudi transfer regulation and EU Article 32 increase the probability of institutional countermeasures, supporting H₃. China’s and Russia’s stronger localization and authorization regimes show that states can increase national control, but they also demonstrate that sovereignty requires legal barriers beyond local hosting, again weakening the simple H₁ formulation. The concentration of technical capability and the high cost of replicating complete cloud platforms support H₂, while geopolitical bifurcation and expanding sovereign-cloud procurement support H₃ and H₄. After qualitative likelihood weighting, the posterior assessment is H₁ 6%, H₂ 28%, H₃ 41%, H₄ 15% and H₅ 10%. These values are analytic judgments, not frequencies extracted from an official dataset. Their principal purpose is to expose which observations would change the assessment.

Hypothesis	Core proposition	Posterior	Confirming indicators through 2030	Disconfirming indicators
H₁	Local regions evolve into complete sovereign control	6%	Domestic operation of every critical layer; tested independence	Continued foreign identity, update and model control
H₂	Governments institutionalize managed hyperscaler dependence	28%	Long contracts; deeper proprietary-service adoption	Mandatory portability and independent key custody
H₃	Critical and commercial workloads split into two stacks	41%	Classified sovereign tiers; domestic operations corps	Uniform migration to one public-cloud model
H₄	Providers diversify across geopolitical technology blocs	15%	Multi-cloud procurement; Chinese, European and local alternatives	Export restrictions prevent interoperable deployments
H₅	A geopolitical shock exposes coercive external control	10%	Service denial, sanctions, compulsory disclosure conflict	Durable continuity through repeated crisis tests

Monte Carlo design and the 2026–2030 outlook

The five-year model treats the sovereignty gap as a weighted composite of six variables: foreign platform concentration, extraterritorial legal exposure, external control of cryptographic and identity systems, workload immobility, domestic operational weakness and geopolitical disruption. The baseline model assigns the largest weights to legal exposure and control-plane dependence because server ownership becomes strategically secondary if a foreign entity can administer the tenant or if the local operator cannot maintain it independently. The simulation architecture uses correlated scenario ranges rather than pretending these variables are independent. For example, higher proprietary-service adoption tends to reduce portability; increased geopolitical tension tends to raise both export-control risk and incentives for domestic investment; stronger customer-held encryption may reduce disclosure exposure while increasing operational complexity. A full research implementation would run at least 100,000 iterations, draw values from bounded beta or triangular distributions, apply positive correlations between concentration and immobility, and test tail events including sanctions, undersea-cable interruption, accelerator-export denial, software-signing disruption and conflict-of-laws orders. The illustrative central paths used in Figure 1 begin from a sovereignty-gap index of 68 in 2026. Under managed dependence, incremental regulatory and technical controls lower the index to 60 by 2030, but the foreign platform remains structurally indispensable. Under sovereign-control reform, national key custody, portable architectures, isolated identity planes and domestic operating capability reduce the index to 34. Under concentration stress, deeper proprietary integration combined with geopolitical or supply constraints raises it to 85. These are scenario outputs, not observed measurements. The critical forecast is that the period from 2026 through 2028 will be dominated by infrastructure deployment and migration; 2028–2029 will expose the difference between nominal exit clauses and executable portability; and by 2030 regional systems will stratify according to sensitivity. States able to operate a sovereign tier without abandoning commercial hyperscale benefits will achieve the most credible balance between resilience, innovation and cost.

Year	Dominant test	Expected structural development	Leading indicator	Strategic warning
2026	Residency versus control	New regional capacity accelerates migration	Percentage of workloads hosted domestically	Residency reported without control-plane disclosure
2027	Key and identity authority	Procurement begins separating sensitive workloads	Customer-exclusive keys and national identity roots	Provider retains emergency recovery authority
2028	Portability	Governments test multi-cloud and exit commitments	Time required to migrate a critical application	Data exports work but applications cannot run elsewhere
2029	Supply continuity	Accelerator, firmware and software dependencies become visible	Domestic spare capacity and offline update capability	Hardware is local but maintenance remains external
2030	Full-stack resilience	Dual-stack architectures become the probable equilibrium	Continuity under simulated external-service denial	“Sovereign cloud” remains a contractual label

The operational threshold of sovereignty

The correct policy response is not autarky and not the indiscriminate exclusion of foreign technology. Full national replication of every cloud and AI layer may be economically inefficient, technologically inferior and less secure if it fragments scarce expertise. The appropriate objective is credible control over critical functions combined with deliberate interdependence elsewhere. A state should classify workloads by consequences of disclosure, manipulation and denial; require complete data-flow maps rather than location statements; separate domestic key custody from provider administration; ensure the national government controls its root identities; maintain immutable logs accessible to national auditors; specify which corporate entities and jurisdictions can access each layer; prohibit silent subcontractor substitution; require notice and challenge procedures for governmental demands to the maximum extent legally permitted; and perform recurring live migration exercises. Reversibility must be treated as an operational capability comparable to disaster recovery, not as a paragraph in a procurement contract. A credible test would require the supplier and customer to migrate a representative critical workload—including databases, identities, policies, logs, model dependencies and application interfaces—within a predetermined recovery window while measuring functionality, data integrity, cost and external assistance. The strongest sovereignty standard should also assess continuity when the provider’s global identity, licensing, telemetry and update services are deliberately disconnected. If the environment ceases to operate, it is locally hosted but not autonomously controlled. The final analytical conclusion is therefore categorical but bounded: local data residency is necessary for some sovereignty objectives, yet it is never sufficient to establish exclusive national authority. Exclusive authority requires convergence between territorial possession, domestic legal supremacy, independent cryptographic control, operational competence, software maintainability, supply resilience and tested portability. West Asia’s central strategic risk is not that its multibillion-dollar infrastructure investments are unreal; it is that visible national facilities may conceal invisible foreign dependencies. Its opportunity is to use current investment leverage to procure control rights, domestic skills and architectural exit capacity before proprietary integration makes those conditions prohibitively expensive.

The Dependency Architecture: How Platforms Convert Infrastructure into Control

The platform is the operating constitution

Cloud dependence begins where physical infrastructure ends. A conventional data-center inventory counts buildings, racks, servers, processors, storage arrays, power capacity and network connections; a sovereignty audit must instead identify every authority capable of determining how those assets operate. In a hyperscale environment, the decisive component is the control plane: the integrated system through which tenants are created, identities authenticated, resources allocated, software deployed, policies enforced, events logged, keys connected, bills calculated and services suspended. A government may own the underlying land, mandate domestic data storage and negotiate a locally incorporated supplier while still relying on a globally managed control plane whose source code, signing keys, update channels, support procedures and administrative escalation mechanisms remain under foreign corporate authority. This produces dependency at three levels. Technical dependence arises when applications require proprietary databases, serverless runtimes, messaging systems, security products or AI interfaces. Operational dependence arises when local personnel cannot restore, patch or scale the environment without the supplier. Strategic dependence arises when loss of the provider would interrupt functions whose absence affects public administration, finance, healthcare, energy or national security. The scale of hyperscaler investment intensifies this asymmetry because global providers can amortize advanced engineering over very large markets. Amazon reported US$128.3 billion in cash capital expenditure during 2025, primarily reflecting technology infrastructure—most of it supporting AWS growth—and fulfilment capacity; the company expected further growth in 2026. Amazon.com, Inc. Form 10-K for the Year Ended 31 December 2025 – Amazon/US Securities and Exchange Commission – February 2026 — official verified filing. This does not prove that a particular West Asian deployment is insecure or politically controlled. It demonstrates the economic barrier confronting any state seeking to reproduce the surrounding ecosystem rather than merely purchase servers.

Platform dependency layer	Foreign-provider control point	Apparent local control	Residual strategic exposure	Evidence needed for sovereignty
Physical infrastructure	Facility standards and approved equipment	Domestic land, power and physical security	Foreign-designed operating dependencies	Domestic operation during provider disconnection
Hypervisor and container layer	Proprietary orchestration and scheduler	Workloads execute on local hardware	Provider controls resource creation and recovery	Independently operable orchestration environment
Managed databases	Engine, replication and backup formats	Primary records remain in-country	Schemas and recovery procedures may be non-portable	Restored database on an alternative platform
Serverless services	Runtime, triggers and proprietary event model	Code executes in local region	Application cannot run without provider APIs	Functionally equivalent deployment elsewhere
Observability	Logs, telemetry, tracing and anomaly detection	Local dashboard access	Global analytics or support systems may receive metadata	Complete telemetry map and independent logging
Software supply	Update repositories and signing infrastructure	Local installation	Security and functionality depend on foreign updates	Domestic verified repository and rollback capability
Billing and entitlement	Subscription, licensing and quota control	National customer pays locally	License or account action can disable services	Continuity without external entitlement server
Administrative support	Global escalation and privileged engineers	Local service desk	Emergency access may cross jurisdictions	Named access boundaries and tamper-evident records

Keys are sovereignty only when authority is exclusive

Encryption is frequently treated as the definitive answer to extraterritorial access, yet the phrase “customer-managed encryption” can describe materially different arrangements. The sovereignty question is not simply whether encryption is enabled; it is who controls the complete cryptographic lifecycle. That lifecycle includes entropy generation, root-key creation, wrapping keys, hardware-security modules, access policies, activation, rotation, backup, recovery, revocation, archival, destruction and audit. NIST divides key management into states and phases precisely because a key can be secure during routine use while remaining exposed through recovery, replication or administration. Recommendation for Key Management: Part 1, Revision 5 – National Institute of Standards and Technology – May 2020 — official verified publication. Provider-managed keys offer operational simplicity but ordinarily leave the supplier with substantial authority. Bring-your-own-key arrangements improve control if the customer generates the material and can revoke access, yet the provider may still operate the key-management service, retain wrapped copies, control identity policies or determine whether an encrypted workload remains available. Hold-your-own-key or external-key-store designs can move the root outside the provider’s environment, but they create a new dependency: if connectivity to the external key authority fails, the national service may become unavailable. Double-key or split-key models reduce unilateral access only when independent parties control genuinely separate cryptographic components and the application cannot route around the intended process. The audit must also examine data in use. Encryption at rest protects stored objects; encryption in transit protects movement; neither automatically prevents exposure while information is being processed. Confidential-computing mechanisms can reduce provider visibility through hardware-backed isolation and attestation, but their trust chain still includes processor architecture, microcode, firmware, attestation services, orchestration and vulnerability remediation. A nationally hosted database is therefore not sovereign merely because its storage volume is encrypted. Sovereignty requires that no foreign supplier can independently decrypt it, silently replace the key pathway, obtain equivalent plaintext through logs or memory, or render the national service inoperable through control of the cryptographic infrastructure.

Key-control model	Root-key generator	Routine access authority	Recovery authority	Sovereignty strength	Principal weakness
Provider-managed key	Provider	Provider-controlled service	Provider	Low	Provider can administer the complete lifecycle
Customer-managed key in provider service	Customer or provider-assisted process	Customer policy through provider platform	Often shared or provider-mediated	Medium	Control plane and recovery remain provider-dependent
Bring your own key	Customer	Provider consumes imported key	Contract-specific	Medium–high	Copies, wrapping and identity paths require verification
External key store	National/customer-controlled service	Provider requests key operations externally	National/customer authority	High if correctly isolated	Availability depends on external key connectivity
Split or double key	Separate national and provider authorities	Joint authorization	Distributed	High	Complexity and misconfiguration can defeat separation
Customer-controlled confidential computing	Customer plus hardware trust chain	Attested workload	Architecture-dependent	Potentially high	Processor, firmware and attestation dependencies persist
Air-gapped sovereign key authority	National authority	Restricted domestic services	Domestic controlled procedure	Very high for key custody	Scalability, integration and operational burden

Identity is the hidden root of every other control

The identity plane is often more powerful than the encryption layer because an actor capable of creating or recovering a privileged identity can potentially change key policies, access storage, reconfigure networks, alter logs or deploy new code. NIST’s zero-trust architecture rejects implicit trust based on network location and focuses authorization on users, devices, assets and resources. Zero Trust Architecture, Special Publication 800-207 – National Institute of Standards and Technology – August 2020 — official verified publication. This principle has a direct sovereignty implication: locating an identity server inside the country is insufficient if the authoritative directory, federation root, certificate chain, recovery channel or policy decision point remains controlled through a foreign platform. The audit must distinguish workforce identities, citizen identities, service accounts, machine certificates, application secrets, automated agents and emergency administrator accounts. Each category has a separate lifecycle and a different capacity to propagate control. A government may federate its national directory with a hyperscaler while retaining nominal ownership of user accounts; however, the supplier may still operate authentication endpoints, evaluate conditional-access policies, deliver multifactor authentication, identify compromised credentials or control the platform-level account to which the national tenant belongs. Emergency recovery is particularly important. If a foreign support organization can re-establish access when national administrators are locked out, it possesses a latent pathway into the system. If it cannot, the government must prove it has sufficient domestic capability to recover independently. Machine identity creates an additional risk because modern cloud applications contain thousands or millions of short-lived service credentials, certificates and tokens. Exporting human-user records while leaving these trust relationships behind does not produce a functioning replacement environment. A credible sovereign design therefore requires nationally controlled root identities, domestic privileged-access management, hardware-backed administrator authentication, separation of duties, immutable logs, independent certificate authorities where justified, and periodic tests in which foreign identity services are made unavailable. The decisive question is not whether the government owns the usernames; it is whether the government alone can establish who or what is trusted.

Identity object	Control capability	Dependency pathway	Failure consequence	Required exit artifact
Root tenant administrator	Creates or revokes all subordinate authority	Provider tenant hierarchy	Total administrative compromise or lockout	Independent national root and recovery procedure
Workforce identity	Grants staff access	Managed directory and multifactor services	Government workforce loses system access	Exportable identities, roles and authentication factors
Machine identity	Authenticates services and workloads	Provider certificates and token services	Applications cannot communicate	Certificates, trust chains and issuance automation
Service accounts	Operate databases and automation	Proprietary secret stores	Silent application failure	Complete account and secret inventory
Citizen identity	Connects public services to individuals	Federated identity gateway	Public-service interruption or profiling exposure	National federation standard and continuity plan
AI-agent identity	Authorizes autonomous tools	Model platform and workflow engine	Automated activity cannot be attributed or constrained	Agent registry, credentials and policy mapping
Emergency identity	Restores privileged access	Global provider support	Foreign entity retains latent control	Domestic break-glass mechanism with audit controls

AI converts platform dependence into cognitive dependence

AI infrastructure introduces dependencies that are more difficult to identify and migrate than conventional compute workloads. A government application built on a foreign foundation model may keep its source records locally yet depend externally on model weights, tokenizer, embedding model, vector format, inference runtime, safety policies, content filters, evaluation framework, fine-tuning method and accelerator architecture. The NIST AI Risk Management Framework directs organizations to govern, map, measure and manage risks across the AI system lifecycle, explicitly including cloud-based services and acquisition. Artificial Intelligence Risk Management Framework 1.0 – National Institute of Standards and Technology – January 2023 — official verified publication. For sovereignty analysis, the most important shift is from data ownership to behavioral control. An external provider may not possess the government’s original records, yet it can still influence output through model updates, safety tuning, system-level instructions, retrieval interfaces, tool permissions, inference throttling or withdrawal of a model version. A locally stored retrieval corpus does not make the system locally controllable if the embedding model is discontinued or the hosted model changes its behavior. Model substitution is rarely exact: replacement alters tokenization, context management, factual performance, language capabilities, refusal boundaries, latency, cost and downstream agent behavior. Arabic-language performance adds a regional dependency because governments may require dialect coverage, culturally specific evaluation and domain terminology not reproduced uniformly across models. A sovereign AI inventory must therefore record model origin, version, weights availability, training-data disclosures where legally obtainable, evaluation results, update rights, rollback rights, prompt retention, location of inference, telemetry paths, human-review requirements and hardware dependencies. It must also classify derived artefacts—embeddings, adapters, synthetic training data, system prompts and evaluation traces—as potentially strategic data. The critical test is reproducibility: can the institution recreate an acceptably equivalent service using another model, runtime and accelerator without losing legal compliance, critical functionality or accumulated institutional knowledge? If not, the AI layer constitutes a distinct dependency even when the underlying database is portable.

AI dependency object	Why data export is insufficient	Sovereignty test	Lock-in severity
Foundation-model weights	Weights determine core capability and behavior	Can the model run under national operational control?	Critical
Tokenizer and prompt format	Replacement changes inputs, costs and outputs	Can prompts be translated without material degradation?	Medium
Embedding model	Existing vectors may be incompatible with replacement	Can the corpus be re-embedded within the exit window?	High
Vector database schema	Index structures and metadata may be proprietary	Can retrieval quality be reproduced elsewhere?	High
Fine-tuning adapters	Adapter may depend on a specific base model	Are weights exportable and legally reusable?	High
Safety and policy layer	Provider can alter refusals and content handling	Can national policy controls be independently operated?	Critical for government use
Evaluation history	Performance evidence may be platform-specific	Can benchmarks be rerun on substitute models?	Medium–high
Agent tooling	Tools, memory and permissions bind models to workflows	Can the complete agent graph be reconstructed?	Critical
Accelerator runtime	Kernels and drivers may depend on particular hardware	Can inference continue under supply restrictions?	High
Inference telemetry	Prompts and outputs create sensitive derivative records	Is collection, retention and access independently auditable?	Critical

Legal compulsion operates through corporate control

Legal exposure cannot be reduced to a slogan that all data under an American provider is automatically available to the United States government. The actual mechanism is narrower and therefore analytically more useful. The US CLOUD Act clarified that providers subject to US jurisdiction may be required, through valid legal process, to disclose responsive information within their possession, custody or control regardless of storage location. The Department of Justice states that the Act did not create a new category of warrant, does not permit indiscriminate bulk collection under ordinary Stored Communications Act warrants and preserves traditional jurisdictional analysis concerning corporate entities and control. The Purpose and Impact of the CLOUD Act – US Department of Justice – April 2019 — official verified white paper. These limitations matter because an accurate sovereignty assessment must avoid treating possibility as certainty. Nevertheless, a foreign state cannot establish exclusive authority if another jurisdiction maintains a legally effective route to compel the provider. Exposure depends on five factual questions: which corporate entity contracts with the customer; which entity possesses or controls the relevant data; which employees or systems can access it; which parent-subsidiary relationships create control; and whether usable plaintext or relevant metadata exists. Encryption may reduce the practical value of compelled production if the provider lacks the key, but it does not necessarily eliminate production of ciphertext, subscriber data, access records, billing information or associated communications metadata. Contractual notification clauses may permit customers to challenge an order, but law can restrict notice. A supplier’s commitment to contest overbroad demands reduces risk without eliminating jurisdiction. The European Union’s Data Act addresses the inverse problem by requiring data-processing providers to implement technical, organizational and legal measures against unlawful third-country governmental access to non-personal data held in the Union. Regulation (EU) 2023/2854 on Harmonised Rules on Fair Access to and Use of Data – European Parliament and Council – December 2023 — official verified legal text. The coexistence of these regimes demonstrates that the cloud is governed by overlapping legal claims rather than a single digital border.

Legal-compulsion variable	Low-exposure condition	High-exposure condition	Evidence required
Contracting entity	National entity with operational independence	Local reseller controlled by foreign parent	Corporate ownership and control analysis
Possession or custody	Provider cannot access customer data	Provider stores accessible plaintext or metadata	Technical access map
Corporate control	Strong legal and technical separation	Parent can direct subsidiary systems	Governance documents and operational evidence
Encryption	National customer holds exclusive keys	Provider manages or can recover keys	Key architecture and recovery test
Notification	Mandatory notice unless lawfully prohibited	Broad secrecy restrictions	Contract terms and transparency procedure
Challenge process	Provider must contest defective demands	Challenge is discretionary	Binding legal and contractual language
Data categories	Minimal metadata and encrypted content	Extensive telemetry, prompts and support records	Complete data inventory
Jurisdictional conflict	Defined escalation and judicial remedy	Contradictory laws with no resolution process	Conflict-of-laws protocol

Vendor exit is a systems-engineering operation, not a data download

The final dependency layer becomes visible only when an organization attempts to leave. A cloud contract may promise data portability while omitting application portability, identity reconstruction, metadata export, policy translation, log preservation, model replacement and continuity during transition. The EU Data Act establishes obligations intended to facilitate switching between data-processing providers and progressively remove switching charges, including relevant data-egress charges. Rules on Fair Access to and Use of Data – EUR-Lex – September 2024 — official verified EU summary. Regulation improves the legal environment, but successful exit still depends on architecture. ENISA’s 2025 technical guidance advises organizations to avoid long-term commitments or vendor lock-in without clear exit strategies and emphasizes that entities remain accountable for confidentiality, integrity and availability when suppliers provide the service. Technical Implementation Guidance on Cybersecurity Risk-Management Measures – European Union Agency for Cybersecurity – June 2025 — official verified guidance. Qatar’s government Cloud First Policy provides an especially relevant regional benchmark: government agencies must prepare an exit strategy before using cloud services, retain ownership of their data, prefer standard formats, identify data and metadata that must be extracted, and integrate cloud exit into risk, continuity and disaster-recovery planning. Cloud First Policy, Version 1.0.0 – State of Qatar Ministry of Communications and Information Technology – 2023 — official verified policy. The policy’s insistence that the provider acts as custodian while government retains ownership is necessary but still incomplete: ownership does not recreate executable capability. A genuine exit package must contain data, schemas, configuration, infrastructure definitions, identities, secrets, certificates, access policies, logs, container images, source code, dependency manifests, AI artefacts, licensing information, operating procedures and trained personnel. It must also establish deletion verification at the old provider and a rollback path if migration fails.

Exit component	Minimum deliverable	Frequent hidden omission	Validation method
Data	Complete export in documented format	Deleted, archived or cold-tier records	Record counts, hashes and reconciliation
Metadata	Schemas, labels, retention and lineage	Provider-generated classifications	Rebuild catalog on target platform
Applications	Source, binaries and deployment definitions	Proprietary serverless dependencies	Functional test in alternate environment
Infrastructure	Machine-readable architecture	Undocumented console configuration	Recreate environment automatically
Identity	Users, roles, services and trust relationships	Recovery and machine identities	Authentication test without old provider
Cryptography	Keys, certificates and rotation records	Wrapped backups and historic keys	Restore and decrypt representative archives
Security	Policies, alerts, detections and cases	Proprietary threat analytics	Execute incident-response scenario
Logs	Complete audit and operational history	Global support and telemetry records	Timestamp and event reconciliation
AI	Models, embeddings, prompts, adapters and evaluations	Provider-owned base weights or filters	Benchmark replacement system
Operations	Runbooks, inventory and trained staff	Tacit knowledge held by contractors	Independent disaster-recovery exercise
Deletion	Verified removal of primary and secondary copies	Snapshots, caches and support artefacts	Auditable deletion certification
Continuity	Parallel-running and rollback plan	Licensing expires before migration completes	Timed live cutover

The dependency graph is multiplicative

The principal analytical mistake is to score each dependency independently and then assume that reducing one control produces an equivalent reduction in total exposure. The architecture is multiplicative because control layers can defeat one another. National key custody has limited value if a foreign identity plane can authorize key use. A sovereign identity directory has limited value if applications require a proprietary platform that cannot run elsewhere. Portable applications remain dependent if their AI models, safety layers or embeddings cannot migrate. A comprehensive technical exit remains vulnerable if a foreign legal order can compel accessible data before migration. Conversely, separating one high-centrality node can reduce several risks simultaneously. Moving root identities and cryptographic authority under genuinely independent national control limits provider access, reduces the usable value of compelled disclosure and makes migration more credible. Standardizing application interfaces and deployment artefacts reduces platform lock-in and AI substitution costs. This suggests a dependency centrality model rather than a checklist. For the baseline architecture used in Figure 2, identity and key authority receive the highest centrality because they connect platform operation, legal exposure and exit capability. Vendor exit receives the highest migration difficulty because every upstream dependency converges there. AI-model dependence ranks close to platform dependence because model substitution affects behavior, not merely infrastructure. These values are analytical indices rather than empirical measurements of any named provider. A formal implementation would create a directed graph in which nodes represent control authorities and edges represent necessary operational relationships. It would then calculate weighted in-degree, betweenness centrality, single-point-of-failure exposure and recovery-time contribution. The resulting score should be tested under denial scenarios rather than accepted from documentation. If disabling one foreign identity service causes national keys, applications and administrative access to fail, identity is the functional sovereignty root regardless of what the contract states.

Control domain	Illustrative dependency index	Network centrality	Migration difficulty	Primary propagation channel
Platform control plane	82/100	Very high	High	Orchestration, APIs and managed services
Identity plane	89/100	Critical	High	Privilege, recovery and machine trust
Cryptographic authority	91/100	Critical	High	Decryption, signing and service availability
AI model layer	86/100	High	Very high	Model behavior, embeddings and runtime
Legal compulsion	77/100	Medium–high	Not directly migratory	Corporate jurisdiction and accessible records
Vendor exit	88/100	Convergence node	Critical	All technical and organizational dependencies
Contractor ecosystem	74/100	Medium	Medium–high	Tacit knowledge and privileged support
Hardware supply	71/100	Medium	High under restriction	Accelerators, firmware and replacement capacity

ACH assessment and five-year outlook

The updated Analysis of Competing Hypotheses evaluates five possible evolutions from 2026 to 2030. H1 holds that commercial “sovereign cloud” offerings will sufficiently isolate legal and operational control without fundamental architectural change. H2 predicts continued managed dependence because hyperscaler performance, security and economics outweigh sovereignty concerns. H3 predicts a dual-stack system in which sensitive government workloads use nationally controlled identity, keys and operations while commercial workloads remain deeply integrated with foreign providers. H4 anticipates multi-provider diversification without true portability, replacing single-vendor dependence with a more complex web of incompatible dependencies. H5 predicts that regulation, open standards and repeated exit exercises produce substantive portability across the regional market. The evidence raises H3 above the alternatives. NIST’s identity and key-management frameworks demonstrate that sovereignty requires lifecycle control; the EU Data Act and ENISA guidance establish that switching and exit must be engineered; Qatar’s policy shows that a West Asian government has already made exit planning, standard formats and lifecycle ownership explicit requirements; and the enormous capital deployed by global providers indicates that full regional replication will remain difficult. The posterior allocation is therefore H1 8%, H2 27%, H3 39%, H4 16% and H5 10%. Through 2027, the decisive procurement shift will be from region availability to control-plane disclosure. During 2028, early migrations will reveal that database export is easier than identity, security-policy and AI reconstruction. In 2029, states will increasingly demand national key roots, domestic privileged operations and deployable model alternatives for sensitive workloads. By 2030, the most advanced systems will not be fully autarkic: they will be deliberately modular, maintaining hyperscaler access for elasticity while reserving nationally controlled recovery, identity, cryptography and degraded-mode operation. The strategic winner will be the state that can interrupt foreign dependencies selectively without interrupting itself.

Hypothesis	2026–2030 trajectory	Posterior	Decisive confirming indicator
H1	Commercial sovereign-cloud wrappers resolve the control problem	8%	Verified independence of control plane and legal governance
H2	Hyperscaler integration deepens under managed safeguards	27%	Long-term adoption of proprietary databases, identity and AI
H3	Sensitive and commercial systems split into distinct stacks	39%	National key and identity roots for critical workloads
H4	Multi-cloud procurement increases complexity without portability	16%	Multiple vendors but no tested cross-platform recovery
H5	Standards and mandatory exit testing create real substitutability	10%	Repeated full-service migrations within defined time limits

The sovereign procurement threshold

A government should not describe a cloud or AI environment as sovereign until it can prove six operational conditions. First, platform continuity: critical applications continue operating in a defined degraded mode when the provider’s global management, licensing and support services are unavailable. Second, exclusive cryptographic authority: nationally accountable entities control root keys, recovery, rotation and destruction, and the supplier cannot obtain equivalent plaintext through another pathway. Third, identity independence: national authorities control privileged administrators, machine trust, certificate issuance and emergency recovery. Fourth, AI substitutability: the institution can replace the model, embedding system and inference runtime while maintaining a defined minimum performance and compliance level. Fifth, legal containment: corporate structure, encryption and access minimization reduce the provider’s possession or control of usable information, while contracts require notice and challenge wherever legally possible. Sixth, executable exit: the complete system—not merely the data—has been transferred or reconstructed during a timed exercise. The UAE Information Assurance Regulation already instructs critical entities to consider legal restrictions on external processing, document cloud security requirements, remain aware of where information is stored or transmitted, reserve audit rights where possible and maintain migration plans for service termination. UAE Information Assurance Regulation, Version 1.1 – Telecommunications and Digital Government Regulatory Authority – 2020 — official verified regulation. The next policy step is to turn these principles into measurable acceptance tests. Procurement scoring should assign material weight to recovery time without the provider, percentage of services using portable standards, proportion of keys under exclusive national control, number of foreign privileged-access paths, time required to rebuild machine identities, model-replacement performance loss and completeness of verified deletion. Sovereignty then becomes auditable engineering rather than marketing language. The architecture remains internationally connected, but every connection is mapped, bounded, reversible and subordinated to a nationally controlled continuity plan.

The 2026–2030 Contest: From Rented Compute to Credible Sovereignty

The contest is over control, not construction

The strategic contest between 2026 and 2030 will not be decided by which West Asian state announces the largest data center, secures the greatest number of accelerators or hosts the widest selection of international cloud regions. Those indicators measure installed capacity and commercial attractiveness; they do not measure the ability to exercise autonomous authority over critical digital functions. The World Bank’s December 2025 assessment counted 39 data centers in the UAE and 33 in Saudi Arabia as of June 2025, compared with a high-income-country average of 81.34, or 53.7 when the United States is excluded. It simultaneously identified Saudi Arabia and the UAE as the Gulf’s leading investors in data centers, cloud computing and high-performance AI infrastructure. The Gulf’s Digital Transformation: A Powerful Engine for Economic Diversification – World Bank – December 2025 — official verified report. These figures establish the regional hierarchy of physical capability but reveal nothing by themselves about domestic control of orchestration, encryption, privileged identities, model weights, software updates, export permissions or emergency recovery. The most useful strategic distinction is therefore between capacity sovereignty, meaning physical access to computing resources; operational sovereignty, meaning the ability to administer and maintain them; legal sovereignty, meaning effective authority over disclosure and use; and cognitive sovereignty, meaning the ability to determine how AI systems process information and produce decisions. A state may rank highly on the first dimension while remaining weak on the other three. Between 2026 and 2030, policy will move from maximizing regional availability toward testing whether the infrastructure remains functional when foreign control planes, support organizations, chip supplies or legal permissions are constrained. The states that build the most capacity will become regional compute hubs. The states that additionally control the critical dependency layers will become sovereign compute powers.

Strategic dimension	What can be counted publicly	What must be proven operationally	False-positive sovereignty indicator
Physical capacity	Facilities, racks, megawatts and accelerators	Availability during grid, cable and provider disruption	Large domestic data-center footprint
Platform capacity	Cloud regions and service catalogues	Independent orchestration and recovery	Local region marketed as “sovereign-ready”
Data authority	Domestic storage and transfer rules	Exclusive control of usable data and metadata	Data-residency contractual clause
Cryptographic authority	Hardware-security modules and encryption adoption	National control of roots, recovery and revocation	Provider-managed encryption at rest
Identity authority	Local directories and digital-ID systems	Independent privileged and machine-identity operation	Domestic identity endpoint
AI capability	Model access, GPU inventory and applications	Model substitutability, evaluation and runtime control	Local inference through foreign API
Legal resilience	National legislation and contractual safeguards	Effective response to conflicting foreign orders	Local subsidiary as contracting party
Exit capability	Export tools and contractual termination rights	Timed reconstruction of the complete service	Downloadable customer database

Managed dependence will remain economically rational—but strategically conditional

Managed dependence is not policy failure by definition. It is a deliberate arrangement in which governments accept substantial reliance on foreign platforms because hyperscalers provide superior scale, security tooling, developer ecosystems, reliability and rapid access to new AI capabilities. For ordinary commercial applications, low-sensitivity government services and globally connected businesses, attempting to replace every foreign component could increase cost, delay innovation and produce weaker cybersecurity. The problem arises when the same architecture is extended without differentiation into identity, healthcare, energy, finance, defense or central-government functions. Saudi Arabia’s partnership between the Public Investment Fund and Google Cloud illustrates the economic attraction. PIF announced an AI hub near Dammam intended to expand cloud infrastructure, support Arabic-language models, train large numbers of students and professionals and contribute—according to preliminary research commissioned by Google Cloud—an estimated cumulative US$71 billion to Saudi GDP over eight years. PIF and Google Cloud to Create Advanced AI Hub in Saudi Arabia – Public Investment Fund – October 2024 — official verified announcement. The economic estimate is not a guaranteed outcome and should not be treated as an audited forecast; the important structural fact is that Saudi capital, domestic power and nationally generated data are being combined with foreign silicon, cloud platforms and model technology. The UAE has followed a similarly large but differently structured path. Abu Dhabi’s official media office reported Microsoft plans to invest US$15.2 billion in the UAE between 2023 and 2029, encompassing AI and cloud infrastructure, skills and research activity. Khaled bin Mohamed bin Zayed Reviews Microsoft’s Plans to Invest US$15.2 Billion – Abu Dhabi Media Office – November 2025 — official verified government source. These partnerships create real national assets and expertise, but they also produce path dependence if local workloads become inseparable from externally controlled technology.

Managed-dependence benefit	Immediate national gain	Embedded dependency	Required containment
Hyperscale capital	Faster infrastructure deployment	Foreign investment priorities affect capacity	National minimum-capacity obligations
Advanced accelerators	Immediate AI training and inference	Export licensing and replacement dependence	Approved inventories and alternative supply planning
Managed security	Global threat detection and rapid patching	Foreign telemetry and response authority	National log access and domestic response teams
Model access	High-performance AI without full training cost	Provider controls weights, versions and policies	Multi-model architecture and rollback rights
Developer ecosystem	Faster application creation	Proprietary APIs increase switching cost	Portability standards and dependency budgets
Global connectivity	Low-latency international service	Reliance on foreign backbone and control systems	Diverse cables, carriers and domestic routing
Commercial credibility	Attracts multinational customers	Reputation linked to provider continuity	National service-continuity guarantees
Training programs	Expands local technical workforce	Certifications may be vendor-specific	Platform-neutral engineering curriculum

Dual stack is the most probable equilibrium

A dual-stack system separates workloads according to the consequences of disclosure, manipulation and denial. The commercial stack uses global hyperscalers for elasticity, software breadth and access to frontier models. The sovereign stack hosts workloads whose interruption or compromise would affect state continuity, public identity, military activity, health systems, monetary functions, energy operations or strategically important industrial data. The two stacks are connected through controlled gateways, shared standards and carefully defined data flows, but they do not share an unrestricted root of trust. The sovereign tier requires nationally controlled encryption roots, privileged identities, logs, network policy, recovery systems and domestic operating personnel. It must also function in a degraded mode without continuous access to the provider’s global control plane. This architecture is more realistic than total autarky because it concentrates expensive sovereign measures where the consequence of failure justifies them. Qatar’s official Cloud First Policy already contains elements of this logic: government agencies must classify their information before migration, exclude the highest C4 national-security/confidential category from the policy’s ordinary cloud pathway, retain data ownership, use standard formats where possible, develop exit strategies before using cloud services and test those strategies according to business criticality and anticipated risk. Cloud First Policy, Version 1.0.0 – State of Qatar Ministry of Communications and Information Technology – 2023 — official verified policy. The dual-stack challenge is not merely technical separation. Governments must prevent low-sensitivity services from becoming hidden dependencies of the sovereign tier. A classified application may run on nationally operated infrastructure while relying on a commercial email system, external time service, remote software repository, foreign certificate authority or globally managed identity provider. Sovereign architecture therefore requires dependency tracing across every upstream and downstream service. By 2030, the leading regional systems will be those that separate not only databases but also roots of trust, management planes, update channels and operational personnel.

Dual-stack function	Commercial/global stack	Sovereign/critical stack	Controlled interface
Compute	Elastic global cloud services	Nationally operated reserved capacity	Approved workload-transfer gateway
Identity	Enterprise cloud directory	National root identities and machine trust	One-way or tightly scoped federation
Encryption	Provider-integrated key services	Nationally controlled hardware key authority	Audited cryptographic request channel
AI inference	Frontier hosted models	Deployable national or isolated models	Sanitized retrieval and policy gateway
Data storage	Commercial and public-service information	Classified, health, identity and critical-sector data	Classification-enforced exchange
Security analytics	Global threat intelligence	Domestic security operations and immutable logs	Curated indicator sharing
Updates	Continuous provider delivery	Tested national repository and staged deployment	Signature and provenance validation
Disaster recovery	Multi-region provider redundancy	Independent domestic recovery environment	Pre-authorized continuity procedures
Internet connectivity	Global routes and public services	Protected government and critical networks	Filtered national exchange points
Operations	Vendor and local partners	Cleared domestic operators	Controlled escalation with recorded access

Energy becomes part of the sovereignty stack

Compute sovereignty cannot exceed energy sovereignty because modern AI infrastructure is an electricity-conversion system as much as a digital platform. The International Energy Agency projects global data-center electricity consumption to increase from approximately 415 TWh in 2024 to about 945 TWh by 2030, with AI as the principal driver. It estimates average data-center electricity growth of around 15% annually from 2024 to 2030, while accelerated-server consumption grows by approximately 30% annually in its base case. Energy and AI: Energy Demand from AI – International Energy Agency – April 2025 — official verified analysis. This growth creates an opportunity for Gulf producers with land, capital and comparatively abundant energy, but it also introduces new constraints. High ambient temperatures increase cooling requirements; water scarcity can make water-intensive cooling politically and environmentally costly; concentrated data-center loads require transmission investment, grid balancing and high-quality continuous power; and sovereign AI clusters may compete with desalination, industry and urban demand. The IEA’s assessment of the Middle East and North Africa projects regional electricity demand to rise by another 50% by 2035, with cooling, desalination, industrial expansion, urban growth and new digital infrastructure all contributing. The Future of Electricity in the Middle East and North Africa – International Energy Agency – 2025 — official verified analysis. The sovereign-capability test must consequently include contracted generation, grid redundancy, fuel availability, cooling resilience, water use and black-start capability. A nationally controlled model that cannot obtain power during peak stress is not operationally sovereign. Conversely, the Gulf can convert an energy advantage into geopolitical compute leverage if it integrates data centers with dedicated low-cost generation, diverse grids, efficient cooling and assured access to replacement hardware. Between 2026 and 2030, the strategic unit will cease to be the data center alone; it will become the compute-energy-water-network complex.

Infrastructure input	2026–2030 pressure	Sovereignty relevance	Required national metric
Firm electricity	AI loads require continuous high-density supply	Foreign cloud capacity fails without domestic power resilience	Firm MW available after single-grid failure
Grid connection	Large clusters can exceed local network capacity	Construction does not equal usable compute	Energized capacity versus announced capacity
Cooling	Gulf temperatures increase heat-removal requirements	Cooling failure can disable national services	Maximum-temperature operating capability
Water	Some cooling systems compete with scarce water resources	Resource dependence constrains expansion	Litres consumed per unit of compute
Natural gas	Supports dispatchable generation	Links compute resilience to fuel availability	Days of assured supply under disruption
Renewables	Can reduce operating cost and carbon exposure	Variable output needs storage or firm backup	Hourly matched clean-energy share
Network	AI clusters require high-capacity low-latency links	Cable or routing concentration creates denial risk	Independent domestic and international routes
Spare hardware	Accelerators and network components have long lead times	Export or logistics interruption affects continuity	Months of critical spares and replacement capacity

Chips turn partnerships into geopolitical licenses

The Gulf’s emerging AI infrastructure sits inside a hardware order still shaped by external export controls, semiconductor manufacturing concentration and supplier-approved end use. Access to advanced accelerators is not simply a commercial purchase: it can depend on bilateral political confidence, diversion controls, data-center security and the exporting state’s assessment of the recipient’s technology relationships. The US Bureau of Industry and Security reported that, consistent with the US–UAE artificial-intelligence cooperation framework signed in May 2025, the Department of Commerce was approving the UAE government and specified companies to receive advanced computing items, including AI chips and servers, under license-free arrangements, while the UAE reaffirmed commitments including matching investment in US AI infrastructure. News and Updates on US–UAE Advanced-Technology Cooperation – Bureau of Industry and Security – 2025/2026 — official verified US government source. This arrangement strengthens UAE access while demonstrating the underlying dependence: the availability of critical compute is connected to an American regulatory and strategic framework. China provides the counter-model. Its National Development and Reform Commission’s data-center action plan links national compute deployment with energy efficiency, renewable-energy use, geographic coordination and the “East Data, West Computing” architecture; by 2030 it aims for internationally advanced data-center efficiency and carbon performance. 数据中心绿色低碳发展专项行动计划 [Special Action Plan for Green and Low-Carbon Data-Centre Development] – National Development and Reform Commission and Chinese Government Departments – July 2024 — official verified Chinese policy. China’s approach illustrates what a more integrated sovereign stack requires: domestic planning of compute, energy, networks and industrial policy. West Asian states do not need to reproduce the Chinese model, but they must recognize that accelerator supply, firmware, high-speed interconnects and replacement parts constitute a strategic dependency. A credible Gulf dual stack will require diversified legally compliant hardware procurement, domestic maintenance skills, spare capacity, workload schedulers capable of using heterogeneous chips and model architectures optimized for several accelerator families rather than one proprietary runtime.

Hardware-control vector	Managed-dependence posture	Dual-stack posture	Credible-sovereignty posture
Accelerator procurement	Purchase best available foreign chips	Reserve approved capacity for critical workloads	Diversified suppliers and heterogeneous scheduling
Export authorization	Supplier manages licensing	Government-to-government framework	Institutional capability to anticipate policy changes
Firmware	Vendor-controlled updates	Domestic validation before deployment	Reproducible update, rollback and audit process
Interconnect	Proprietary high-speed networking	Segmented sovereign clusters	Multiple interoperable network technologies
Maintenance	Foreign field support	Cleared domestic first-line teams	National repair, replacement and forensic capability
Spares	Just-in-time procurement	Reserved critical inventory	Multi-year continuity inventory for strategic systems
Software runtime	Single accelerator ecosystem	Portable container and abstraction layer	Tested execution across different hardware families
Model efficiency	Scale through more hardware	Optimize critical models for constrained operation	National capability in compression and efficient inference

Regulation will determine whether investment creates leverage or rent

The 2026–2030 contest will be shaped as much by procurement and regulatory architecture as by engineering. Saudi Arabia’s Personal Data Protection Law requires that cross-border transfer or disclosure not prejudice national security or the Kingdom’s vital interests and applies adequacy and protection conditions to relevant transfers. Personal Data Protection Law – Saudi Data and AI Authority – April 2023 — official verified English text. Such rules strengthen state authority over data movement, but they do not automatically produce technical independence from a provider. The next regulatory stage must govern control rather than location: disclosure of corporate jurisdiction, customer-exclusive key custody, identity separation, provider personnel access, operational telemetry, AI-model changes, export-control exposure, subcontracting, deletion, portability and continuity during termination. The European Data Act offers a relevant external benchmark because it imposes switching-related obligations on data-processing services, addresses interoperability and progressively restricts switching charges. Regulation (EU) 2023/2854 on Harmonised Rules on Fair Access to and Use of Data – European Parliament and Council – December 2023 — official verified regulation. ENISA’s technical guidance additionally warns organizations against long-term commitments or supplier lock-in without clear exit strategies and maintains that the regulated entity remains accountable for services performed by suppliers. Technical Implementation Guidance on Cybersecurity Risk-Management Measures – European Union Agency for Cybersecurity – June 2025 — official verified guidance. West Asian governments can move beyond these baselines by requiring live exit tests before contract renewal, placing measurable ceilings on proprietary dependencies, and imposing a sovereignty budget similar to a cybersecurity-risk budget. Each project would declare how many critical functions depend on a single foreign provider and how long each could continue if the dependency disappeared. Regulation would then reward reduced concentration without forcing inefficient duplication across every workload.

Procurement control	Conventional requirement	Sovereignty-grade requirement	Verification event
Data residency	Primary data stored domestically	All replicas, metadata and support pathways mapped	Independent data-flow audit
Key management	Encryption enabled	National exclusive control of root and recovery keys	Provider-exclusion decryption test
Identity	Government owns accounts	National control of privileged and machine identities	Global-directory disconnection test
AI service	Model available in local region	Substitute model meets minimum benchmark	Model-switch exercise
Portability	Data export supported	Complete application reconstructed elsewhere	Timed full-stack migration
Continuity	Provider multi-zone resilience	Operation without provider global services	Isolation exercise
Legal process	Provider promises compliance	Jurisdiction, notification and challenge duties disclosed	Conflict-of-laws tabletop exercise
Hardware	Capacity committed	Replacement and export-control plan documented	Supply-interruption simulation
Personnel	Local support offered	Cleared domestic team can operate independently	Provider-free recovery drill
Deletion	Contractual deletion promise	Primary, replica, cache and support copies verified	Auditable deletion evidence

The shadow contest: talent, contractors and liquidity

The visible contest over facilities conceals a deeper contest for technical labor, operating knowledge and capital allocation. A sovereign data center without engineers capable of debugging distributed storage, managing cryptographic infrastructure, validating firmware, operating high-speed networks, evaluating Arabic-language models and conducting incident response remains dependent on contractors. This contractor layer can become a functional “mercenary” market in the limited analytical sense that scarce specialists move between governments, providers and integrators according to compensation, access and project cycles. The risk is not that contractors are inherently disloyal; it is that strategically important knowledge remains tacit, externally employed and difficult to retain inside national institutions. The World Bank’s Gulf assessment finds strong digital foundations and significant investment in tertiary AI skills, while also emphasizing the importance of attracting and retaining talent capable of sustaining digital infrastructure. The Gulf’s Digital Transformation: A Powerful Engine for Economic Diversification – World Bank – December 2025 — official verified report. Liquidity flows create another shadow dependency. Sovereign capital can finance data centers and take equity positions in technology ventures, yet if local startups receive cloud credits, train workers only on proprietary services and design products around inaccessible foreign models, national investment expands the foreign platform’s ecosystem. The relevant financial metric is therefore not total digital investment but the percentage that creates transferable capability: domestic intellectual property, platform-neutral skills, open interfaces, nationally controlled datasets, model-evaluation infrastructure and operating teams. By 2030, countries that treat training as certification procurement will have many qualified users of foreign platforms; countries that fund deep systems engineering, cryptography, compiler development, model optimization and hardware operations will possess a sovereign capability base. The difference will appear during the first prolonged disruption, when certificates matter less than the ability to restore and redesign systems.

Capability investment	Superficial output	Sovereign output	2030 measurement
Workforce training	Number of vendor certificates	Independent platform and systems expertise	Provider-free operating exercises passed
Research funding	Publications and demonstrations	Deployable models, tools and security components	Production systems under national control
Startup finance	Cloud-based application growth	Portable products and domestic IP	Share of revenue independent of one platform
University programs	General AI enrolment	Cryptography, compilers, distributed systems and chips	Graduates in critical technical disciplines
Government contracting	Delivery of functioning platform	Transfer of knowledge and operational responsibility	Domestic staff performing critical roles
Sovereign investment	Capital committed to foreign partnerships	Negotiated access, governance and local capability	Enforceable control rights and domestic assets
Data programs	Volume of collected data	Governed, high-quality national training resources	Reusable datasets with legal provenance
Cybersecurity spending	Imported tools	Domestic detection, response and forensic competence	Mean recovery time without foreign support

Bayesian update: the dual stack remains dominant

The five competing hypotheses can now be updated using evidence available through August 2026. H1, managed dependence consolidates, assumes governments continue accepting foreign platform control while improving contracts, localization and regulatory oversight. H2, dual-stack transition, assumes states segment critical and commercial systems, controlling keys, identities and recovery for the former while retaining hyperscaler integration for the latter. H3, credible sovereign capability, assumes one or more Gulf states achieve broad independent control across compute, energy, operations, AI models, hardware resilience and legal architecture. H4, fragmented multi-cloud, assumes governments diversify suppliers but fail to create portability, increasing complexity without materially reducing dependence. H5, external shock reveals latent control, assumes export restrictions, legal disputes, geopolitical conflict, cable disruption or provider action exposes operational vulnerability. The previous-stage evidence favored dual stack; the new infrastructure, energy and export-control evidence strengthens that conclusion. Large foreign-linked investments make H1 economically plausible, but stronger classification, exit and data-transfer rules reduce the probability of unrestricted dependence. Hardware licensing and the extraordinary capital intensity of frontier infrastructure constrain H3. Qatar’s operational exit requirements and the EU’s switching framework demonstrate mechanisms that could support H2, while the growing number of platforms makes H4 a material risk. The posterior assessment is H1 25%, H2 43%, H3 12%, H4 12% and H5 8%. These are structured judgments, not official statistics. The main change signal for H3 would be successful national operation of critical AI and cloud services through repeated foreign-service denial exercises. The main change signal for H5 would be abrupt tightening of accelerator access, support or licensing affecting installed regional capacity. Until either occurs, H2 remains the most coherent forecast.

Hypothesis	Prior	Updated posterior	Evidence driving update	Key disconfirming observation
H1 — Managed dependence	27%	25%	Large foreign-linked infrastructure investments	Critical workloads systematically removed from foreign control
H2 — Dual-stack transition	39%	43%	Classification, exit planning and national-control requirements	Governments retain one undifferentiated hyperscaler architecture
H3 — Credible broad sovereignty	10%	12%	Rising capital, skills and national AI ambition	Continued inability to operate without foreign models or hardware
H4 — Fragmented multi-cloud	16%	12%	Supplier proliferation but weak portability	Tested interoperability becomes routine
H5 — External shock	8%	8%	Export-control and geopolitical exposure persists	Durable access and continuity through repeated crises

Monte Carlo outlook and annual decision points

The scenario model underlying Figure 3 treats sovereign capability as an aggregate of eight variables: domestic key custody, identity-plane control, workload portability, AI-model substitutability, domestic operational depth, hardware-supply resilience, legal insulation and tested service continuity. The simulation concept applies uncertainty bands because implementation outcomes are not independent. Greater national key control may initially reduce availability; stronger portability may slow application delivery; tighter hardware controls can increase incentives for efficiency and alternative architectures; foreign investment can simultaneously expand domestic capacity and deepen platform dependence. The managed-dependence path rises only from an illustrative 31 in 2026 to 38 in 2030, reflecting improved contracts and localization without structural control. The dual-stack path rises to 69, assuming progressive segregation of sensitive workloads and national control of high-centrality layers. The credible-sovereignty path reaches 87, but requires unusually successful execution across energy, workforce, hardware, models, law and operations. A full Monte Carlo implementation would run at least 100,000 iterations, use correlated bounded distributions, and apply discontinuous shocks for chip restrictions, model withdrawal, severe cyber incidents, cable outages and regional power stress. For 2026, the critical decision is classification: governments must identify which functions require sovereignty rather than residency. In 2027, procurement must move from promises to architectures, especially national key and identity roots. In 2028, governments must conduct the first full-stack exit and isolation exercises. In 2029, sovereign AI alternatives and heterogeneous accelerator operation become decisive. In 2030, credibility will depend on demonstrated continuity rather than capital announcements. Failure at any preceding stage compounds later because applications, staff and data accumulate around the incumbent platform.

Year	Primary contest	Required deliverable	Success indicator	Failure indicator
2026	Classification and dependency discovery	National critical-workload register	All high-impact systems mapped to control layers	“Sovereign” designation based only on location
2027	Roots of trust	National key, identity and audit architecture	Provider cannot unilaterally recover privileged access	Emergency administration still depends on global support
2028	Portability and isolation	Full-stack migration and disconnection exercises	Critical service restored within defined recovery time	Data exports but application cannot operate
2029	AI and hardware resilience	Substitute models and heterogeneous compute runtime	Minimum national benchmark met without primary provider	Model or accelerator loss terminates function
2030	Credibility under stress	Multi-sector continuity certification	Repeated successful foreign-service denial tests	Compliance documents substitute for operational proof

Credible sovereignty is the power to choose dependence

The strongest end-state is not technological isolation. It is the ability to choose, limit and terminate dependencies without losing control of essential national functions. Credible sovereignty therefore has four defining properties. It is selective, because the state does not spend sovereign resources on every low-risk workload. It is modular, because applications, identities, keys, models and infrastructure can be separated or substituted. It is tested, because continuity and exit are demonstrated through exercises rather than assumed from contractual language. It is economically integrated, because global platforms remain available where they create value without becoming irreplaceable national control points. The central 2030 metric should be the proportion of critical services capable of meeting a nationally specified operating threshold for a defined period after the loss of foreign control-plane access, software support and new hardware deliveries. Supporting metrics should include the share of critical keys held exclusively by national authorities; time required to rebuild privileged and machine identities; percentage of applications deployable on an alternate platform; percentage of AI functions meeting minimum benchmarks with substitute models; number of foreign personnel possessing potential privileged access; and verified deletion completeness following migration. A government that reaches these thresholds may continue purchasing foreign cloud and AI services extensively, but the relationship changes: the supplier provides capability without possessing decisive leverage over continuity. The 2026–2030 contest is therefore not between openness and isolation, nor between Western, Chinese and domestic technologies in the abstract. It is between architectures that accumulate invisible veto points and architectures that convert international interdependence into bounded, reversible and auditable relationships. West Asia has sufficient capital, energy potential and strategic demand to build the second model. Whether it does so will depend less on the size of announced investments than on procurement discipline, systems engineering, national talent and willingness to test what happens when the external platform is no longer available.

---

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