Amazon on Thursday confirmed it is working on a project to deploy a network of satellites for high-speed internet service in underserved parts of the world.
Project Kuiper was first reported by tech news website GeekWire, which cited US regulatory filings disclosing the satellite project that could cost billions of dollars to complete.
“Project Kuiper is a new initiative to launch a constellation of low earth orbit satellites that will provide low-latency, high-speed broadband connectivity to unserved and underserved communities around the world,” Amazon said in response to an AFP inquiry.
“This is a long-term project that envisions serving tens of millions of people who lack basic access to broadband internet.”
Named for an astronomer who’s considered “the father of modern planetary science,” Gerard Kuiper, Kuiper Systems is the latest foray into space-based internet networking by a U.S. tech giant.
As private companies look to commercialize space, high-speed internet is among the prospects that offer the highest profits in the short term, while providing necessary services to get online the remaining 3.8 billion people who don’t have access to the internet.
In February, OneWeb, another company that’s expecting to create a network of satellites to provide high-speed internet access, successfully launched its first satellites. The company has raised at least $3 billion, according to Crunchbase, from investors, including Virgin, Coca-Cola and the Bharti Group — and they’re not the only company to raise several billion dollars to develop these services.
SpaceX also has designs on creating a global satellite network – in addition to its leading position as a launch services provider for companies looking to access outer space.
In December, the company set out to raise another $500 million to support its Starlink program, which would create a network of 11,000 satellites to cover the globe with internet connectivity.
To date, the company has launched just two prototype satellites, even though earlier reports stated SpaceX, at one time, projected it would have 400 satellites in orbit by the end of 2018.
Finally, the social networking giant Facebook has been working on satellite capabilities of its own.
In a May report, the IEEE Spectrum laid out how Facebook had set up a small subsidiary called PointView Tech, which was developing a new satellite called “Athena” that could deliver data 10 times faster than SpaceX’s Starlink satellites.
Amazon’s Kuiper satellite service expected to put 3,236 satellites in low orbit at altitudes ranging from 367 miles (590 kilometers) to 391 mile (630-kilometer), according to GeekWire.
The frontier of space is internationally agreed to be 62 miles (100 kilometers) above Earth, known as the Karman Line.
The Seattle-based online powerhouse was looking to partner with like-minded companies on the effort.
There was no indication that Project Kuiper thus far involved Blue Origin, the rocket company owned by Amazon chief executive Jeff Bezos, which blasted off the 10th test flight of its New Shepard rocket early this year.
More test flights lie ahead, but the first flights with passengers on board could start by late 2019.
Amazon would be one of several companies seeking to use satellites to deliver internet to remote areas including Elon Musk’s SpaceX and OneWeb, a venture-backed startup with funding from Japan’s SoftBank, Airbus and chipmaker Qualcomm.
Several companies have been attempting to use space-based internet systems since the 1990s including one backed by Microsoft’s Bill Gates and Saudi royal family investors.
There’s a theory (or perhaps a cautionary tale) among astronomers called the Kessler Syndrome, named for the NASA astrophysicist who proposed it in 1978.
In this scenario, an orbiting satellite or some other piece of material accidentally strikes another and breaks into pieces.
These pieces whirl around the Earth at tens of thousands of miles per hour, destroying everything in their path, including other satellites.
It starts a catastrophic chain reaction that ends in a cloud of millions of pieces of non-functional space debris that orbits the planet indefinitely.
Such an event could make an orbital plane functionally useless, destroying any new satellites sent into it and possibly preventing access to other orbits and even all of space.
So when SpaceX filed a request with the FCC to send 4,425 satellites into low-Earth orbit (LEO) to provide a global high-speed internet network, the FCC was reasonably concerned.
For more than a year, the company responded to questions from the commission and petitions by competitors to deny the application, including filing an “orbital debris mitigation plan” to allay fears of Kesslerian apocalypse. On March 28, the FCC granted SpaceX’s application.
Space junk is not the only thing the FCC is concerned about – and SpaceX isn’t the only entity trying to build the next generation of satellite constellations.
A handful of companies, both new and old, are leveraging new technology, developing new business plans, and petitioning the FCC for access to the parts of the communications spectrum they need to blanket the Earth in fast, reliable internet.
Big names are involved – from Richard Branson to Elon Musk – along with big money. Branson’s OneWeb has raised $1.7 billion so far, and SpaceX president and COO Gwynne Shotwell estimated a $10 billion price tag for that company’s project.
There are big challenges, of course, and a history not exactly favorable to these efforts. Good guys are trying to bridge the digital divide in underserved regions even as bad actors slip illegal satellites onto rocket rideshares.
And it’s all happening as (or really, because) demand for data has skyrocketed:
In 2016, global internet traffic exceeded 1 sextillion bytes, according to Cisco, kicking off the zettabyte era.
If the goal is to provide (good) internet access where previously there was none, satellites are a reasonable way to achieve it.
In fact, companies have been doing this for decades via large geostationary (GSO) satellites that sit in a very high orbit, fixed above a certain point on the Earth.
But aside from a few niche applications, including cargo tracking and providing internet to military bases, this kind of satellite connectivity has not been fast, reliable, or responsive enough to be competitive with modern fiber or cable-based internet.
Non-GSOs include MEOs, which operate in medium-Earth orbit from 1,200 to 22,000 miles above the Earth’s surface, and LEOs (up to about 1,200 miles). If LEOs aren’t all the rage today, at least they’re most of it.
Meanwhile, regulations for non-geostationary satellites are decades old and split between agencies within and beyond the US: NASA, the FCC, DOD, FAA, and even the UN’s International Telecommunication Union all have skin in this game.
There are some big advantages on the technological side, though.
The cost to build a satellite has fallen as gyroscope and battery improvements have trickled down from cell phone guts.
Launching them has gotten cheaper, too, thanks in part to the smaller size of the satellites themselves. Capacity has risen, inter-satellite communication has made systems faster, and large dishes pointing at the sky are on their way out.
On the back of this tech, 11 companies filed applications in the same FCC “processing round” as SpaceX did, each tackling the problem a bit differently.
Elon Musk announced the SpaceX Starlink program in 2015 and opened a Seattle-based division of the company.
He told employees there, “We want to revolutionize the satellite side of things, just as we’ve done with the rocket side of things.”
In 2016, the company filed the FCC application, which called for 1,600 (later reduced to 800) satellites to go up between now and 2021, followed by the rest before 2024.
These will fly between 1,110km and 1,325km above the ground, circling the Earth in 83 distinct orbital planes.
The constellation, as a group of satellites is called, will communicate with one another via onboard optical (laser) interlinks, so that data can be bounced along the sky rather than returning to the ground—tracing a long bridge rather than an upside-down V.
On the ground, customers will mount a new sort of terminal with electronically steered antennas that automatically connect to whichever satellite is currently offering the best signal – similar to the way a cell phone picks towers.
Because LEO satellites move relative to the Earth, the system will switch between them every 10 minutes or so.
And because thousands will be up there, at least 20 will always be available to choose from, according to Patricia Cooper, VP of Satellite Government Affairs for SpaceX.
The ground unit should be cheaper and easier to mount than traditional satellite dishes, which have to be positioned physically to point at the part of the sky where the corresponding GSO satellite lives. SpaceX described the terminal as the size of a pizza box (though it did not note what size pizza).
The communication will happen within two frequency bands: Ka and Ku. Both appear on the radio spectrum, though at much higher frequencies than anything you’d hear on your stereo. Ka-band is the higher of the two, with frequencies between 26.5GHz and 40GHz, while Ku-band inhabits frequencies from 12GHz to 18GHz.
(Starlink has FCC permission to use particular frequencies; typically, uplink from terminal to satellite will be at 14GHz to 14.5 GHz and downlink from 10.7GHz to 12.7GHz, and the others will be used for telemetry, tracking, and control, as well as to connect the satellites to the internet’s terrestrial origin.)
Beyond the FCC filings, SpaceX keeps pretty quiet about its plans.
And it’s hard to tease out technical details, because SpaceX is vertically integrated from the components that go on the satellites to the rockets that get them into the sky.
But for the project to be a success, it will depend on whether the service can, as claimed, offer speeds comparable to or better than fiber at a similar price point, along with a reliable experience and a good user interface.
In February, SpaceX launched its first two prototype Starlink satellites.
Shaped like cylinders with solar panels for wings, Tintin A and B are roughly a meter per side, and Musk confirmed via Twitter that they were successfully communicating.
If the prototypes continue to function, they will be joined in 2019 by hundreds of others.
Once the system is operational, the SpaceX will replace decommissioned satellites (and mitigate space debris) on a rolling basis by instructing them to lower their orbits, whereupon they’ll fall toward Earth and burn up on reentry.
The Wayback (Circa 1996)
Back in the 80s, HughesNet was the satellite technology innovator.
You know the platter-size gray dishes DirecTV mounts on the outside of houses?
Those came from HughesNet, which itself came, circuitously, from aviation pioneer Howard Hughes.
“We invented the technology that allows us to provide interactive communications via satellite,” says EVP Mike Cook.
In those days, then-named Hughes Network Systems owned DirecTV and operated large geostationary satellites that beamed information down to televisions.
Then and now, the company also offered services to businesses, like credit card transactions on gas pumps.
Its first commercial customer was Walmart, which wanted to link employees across the country and its home office in Bentonville.
In the mid-90s, the company built a hybrid internet system called DirecPC: A user’s computer submitted a request via dial-up; it was directed to a web server and completed via a satellite, beaming the requested page down to the user’s dish.
Around the year 2000, Hughes began providing its first two-way interactive system. But keeping the cost of the service – including the consumer equipment – low enough that people would buy it was a challenge.
To do that, the company decided it needed its own satellites, and in 2007, it launched Spaceway.
Though still in use, this satellite was particularly important when it launched, according to Hughes, because it was the first to incorporate onboard packet switching. Its capacity: 10Gbps.
Meanwhile, a company called Viasat spent around a decade in R&D before launching its first satellite in 2008. Called ViaSat-1, the satellite incorporated some new technology, such as spectrum reuse.
This allowed the satellite to choose among different bandwidths so it could pump data down to Earth without interference, even when it neighbored the track of another satellite’s beam, and then reuse that spectrum in connections that were not adjacent.
It was also faster and more powerful. When it went up, its 140Gbps capacity was more than all of the other satellites covering the US combined, according to Viasat President Rick Baldridge.
“The market for satellites had really been the people that had no choice,” Baldridge says. “If you couldn’t get anything else, it was a technology of last resort.
It essentially had a ubiquitous coverage but really, not much data. It had been relegated to things like transactions at gas stations.”
Over the years, HughesNet (now owned by EchoStar) and Viasat put up faster and faster GSOs. HughesNet put up EchoStar XVII (120Gbps) in 2012, EchoStar XIX (200Gbps) in 2017, and plans to launch EchoStar XXIV in 2021, which the company says will offer 100Mbps to consumers.
ViaSat-2 went up in 2017 and now has a capacity of around 260Gbps, and three different ViaSat-3s are planned for 2020 or 2021, each to cover a different part of the globe.
Viasat has said that each of those three ViaSat-3s are projected to have a capacity of a terabit per second each, double the capacity of all other satellites circling Earth combined.
“We have so much capacity in space that it is changing the whole dynamic of providing this traffic. There is no inherent limit in terms of what can be provided,” says DK Sachdev, a satellite and telecom consultant who is doing work for LeoSat, one of the companies launching an LEO constellation. “Today, all the things we thought were disadvantages for satellites, one by one they’re shifting away.”
All this speed has come about, not coincidentally, as internet (two-way communication) has begun replacing television (one-way) as the primary service we demand from our satellites.
“The satellite industry is in a very long-time frenzy, figuring out how it will go from predominantly video, to now and ultimately only predominantly data,” says Ronald van der Breggen, chief compliance officer at LeoSat. “There are a lot of opinions about how to do it, what to do, what market to serve.”
One Problem Remains
There remains one problem: latency. Different from overall speed, latency is the amount of time it takes information from your computer to reach its destination and return.
Say you click on a link to a website; that information has to travel out (in this case, up to a satellite and back down), indicate your request, and return the site.
How long it takes the site to download is based on how much capacity the connection has. How long it takes to ping that server and get it started is latency.
It’s typically measured in milliseconds — not something you’d notice when you’re reading book in your browser but very frustrating when you’re playing with a on line game and your game lags.
Latency on a fiber system varies based on distance, but it’s generally a few microseconds per kilometer.
Latency, when you’re beaming a request to a GSO satellite, is in the neighborhood of 700ms total, according to Baldridge – light travels faster in the vacuum of space than in fiber, but these kinds of satellites are far away, and it just takes time.
In addition to gaming, this is a problem for video conferencing, financial transactions and the stock market, control of the internet of things, and other applications that depend on snappy turnaround.
But how big an issue latency is can be debated.
Much of the bandwidth used around the world is for video; once a video is started and properly buffered, latency becomes a non-issue, and throughput is more important.
Not surprisingly, Viasat and HughesNet tend to minimize the importance of latency for most applications, though both are working to minimize it in their systems, too. (HughesNet uses an algorithm to prioritize traffic based on what users are looking at to optimize data delivery; Viasat announced an MEO constellation to supplement its existing satellites, which should decrease latency and fill in coverage areas including those at high latitude, where equatorial GSOs have a hard time reaching.)
“We’re really focused on high volume and very, very low capital cost to deploy that volume,” says Baldridge. “Is latency as important as the other features for the market we’re supporting?”
But the point remains; an LEO satellite is still much closer to users. So companies such as SpaceX and LeoSat have chosen this route, with their constellations of smaller, closer satellites, anticipating latency of 20 to 30 milliseconds.
“It’s a trade-off that, because they’re in a lower orbit, you get a lower latency from an LEO system, but you have more complexity in the system,” says Cook.
“You have to have at least hundreds of satellites in order to complete the constellation,because they’re orbiting, one’s going over the horizon and disappearing … and you have to have an antenna system which is capable of tracking them.”
Two episodes before this are worth understanding.
In the early 90s, Bill Gates and a few partners invested in a project called Teledesic.
It was to use a constellation of 840 (later reduced to 288) LEO satellites to provide a broadband network to regions that couldn’t afford or would never see fiber connections. Its founders talked about solving the latency problem, and in 1994, applied to the FCC for use of Ka-band spectrum. (Sound familiar?)
Teledesic ate up an estimated $9 billion before it failed, in 2003.
“That idea didn’t work then, but it seems feasible now,” says Larry Press, a professor of information systems at California State University Dominguez Hills who has been tracking LEO systems since Teledesic was new.
“The tech was not there by a long shot.”
Moore’s Law and the trickle-down of battery, sensor, and processor technology from cell phones has given LEO constellations a second chance.
Increased demand makes the economics look tantalizing.
But while the Teledesic saga was playing out, another industry was learning some important lessons about launching communications systems into space.
In the late 90s, Iridium, Globalstar, and Orbcomm collectively launched more than 100 satellites into LEO with the purpose of providing cell phone coverage.
“To get the whole constellation up there takes years, because you need a whole bunch of launches, and it’s really expensive,” says Zac Manchester, an assistant professor of aeronautics and astronautics at Stanford University.
“In the intervening say, five years or so, the ground-based cell tower infrastructure expanded to the point where the coverage was really good, and it covered most of the people.”
All three companies swiftly descended into bankruptcy.
And while each has reinvented itself, offering a smaller range of services for specific applications such as emergency beacons and cargo tracking, none succeeded in supplanting tower-based cell phone service.
(In the last few years, SpaceX has contracted to launch satellites for Iridium.)
“We’ve kind of seen this movie before,” says Manchester. “I don’t see anything inherently different about the current situation.”
SpaceX and the 11 other corporations (and their investors) are betting otherwise. OneWeb is launching satellites this year, with service expected to start next year, and adding several more constellations in 2021 and 2023, with an ultimate goal of 1,000 terabits by 2025. O3b, now a subsidiary of SAS, has a constellation of 16 MEO satellites that has been operational for several years.
Telesat already operates GSO satellites but is planning an LEO system for 2021 that features optical links with 30ms-to-50ms latency.
Upstart Astranis also has a satellite up in geosynchronous orbit and will be placing more in the next few years; though it’s not addressing the latency issue, the company is aiming to bring costs down drastically by working with local ISPs and building smaller and far cheaper satellites.
LeoSat, too, plans to launch a first round of satellites in 2019, with completion in 2022.
These will sail around the earth at 1,400km high, connect to the other satellites in the mesh via optical communication, and beam information up and down in Ku-band.
They have acquired the necessary spectrum internationally, says LeoSat CCO Ronald van der Breggen, and expect to receive FCC approval soon.
The quest for faster satellite internet has largely relied on building bigger, faster satellites that can carry more data, says van der Breggen.
He calls it “the pipe”: the bigger the pipe, the more internet can gush through it.
But companies like his are finding new areas to make improvements by changing the whole system.
“Imagine the smallest type of network – two Cisco routers and a wire in between,” says van der Breggen.
“What everybody in satellites does is to focus on the wire between the two boxes … we’re bringing that whole set of three up in space.”
LeoSat is putting up 78 satellites, each about the size of a large dinner table and weighing about 1,200Kg.
Built by Iridium, they feature four solar panels and four lasers (one on each corner) to connect to their neighbors.
It’s that connection van der Breggen says is most important; historically, satellites would bounce signal in a V shape, from the ground station up to the satellite and then down to the receiver.
Because LEO satellites are lower, they can’t project as far, but what they can do is pass data along very quickly.
To understand how this works, it’s helpful to think of the internet as a thing, with a real physical presence.
It’s not just data; it’s where that data lives, and how it moves.
It’s not just stored in one place; there are servers around the world that hold it, and when you access it, your computer grabs it from the nearest one that happens to have what you’re looking for.
Where it is matters. How far away it is matters. Light (a.k.a. information) travels faster in space than in fiber, almost by half.
And when you bounce that fiber connection around the face of the planet, it has to take a circuitous route from node to node, with detours around mountains and continents.
It winds up taking much longer when the source of the data is far from the consumer, even when you account for the few thousand miles of vertical distance a space-bound signal adds.
Like what van der Breggen describes, the whole industry could be viewed as a progression toward developing a distributed network not unlike the internet itself, just in space. Latency and overall speed are both at play.
While one company’s technology might prove supreme, it’s not entirely a zero-sum game.
Many of these companies are targeting different markets and are even helping one another reach the markets they’re after.
For some it’s ships, planes, or military bases; for others, it’s rural consumers or developing nations.
But ultimately, the companies share a goal: to bring internet where there is none or where it’s insufficient and to do so at a cost that’s low enough to sustain their business model.
“Our view is that this isn’t really a competing technology. We believe that there is a need, in a sense, for both LEO and GEO technology.” says HughesNet’s Cook.
“For certain types of applications, like streaming video, for example, a GEO system is very very cost effective.
However, if you want to have applications which require low latency … then LEO is the way to go.”
To wit, HughesNet has actually partnered with OneWeb to provide the gateway technology that manages traffic and interfaces the system with the internet.
You may have noticed that LeoSat’s proposed constellation is smaller than SpaceX’s by nearly a factor of 10.
That’s okay, says van der Breggen, because LeoSat intends to serve enterprise and governmental clients and therefore needs to light up just a few specific areas. O3b is selling internet to cruise ships, including Royal Caribbean, and it’s working with telecoms in American Samoa and the Solomon Islands, where wired connections are insufficient.
A small startup from Toronto called Kepler Communications is using tiny CubeSats (around the size of a loaf of bread) to provide “delay-tolerant” data – 5GB or more of data in a 10-minute pass, with an emphasis on polar exploration, science, industry, and tourism.
According to Baldridge, one of Viasat’s biggest growth areas is in providing internet to commercial airlines; they’ve inked deals with United, JetBlue, and American, as well as Qantas, SAS, and more.
How, then, does this business-first, for-profit model bridge the “digital divide” and provide internet for developing nations and underserved communities, which may not be able to pay as much for it?
It has to do with the shape of the system.
Because the individual satellites move, an LEO constellation must be evenly distributed around the Earth.
The ones that pass out of view inhabit a different part of the sky and are temporarily a sunk cost.
“My guess is, they will have very different prices for connectivity in different nations, and that will allow them to make it affordable in one place, even though it might be a very poor place,” says Press. “Once the satellite constellation is up there, it’s a fixed cost, and if a satellite is over Cuba, and nobody is using it, then any revenue they can get out of Cuba is positive, is free.”
Wherever it may lie, this consumer market may be the hardest to tap.
In fact, most of the success the industry has had so far has been providing expensive internet for governments and businesses.
But SpaceX and OneWeb particularly have visions of household customers dancing in their business plans.
To access this market, the user interface is going to be important, Sachdev points out. You have to cover the Earth with a system that is easy to use, effective, and cost-effective. “Covering it by itself is not adequate,” says Sachdev.
“What you need is an adequate amount of capacity, but before that, the ability to have consumer equipment that is affordable.”
Who’s in Charge, Anyway?
The two big issues SpaceX had to address for the FCC were how it would share spectrum with existing (and future) satellite communications, and how it would mitigate or prevent space debris.
The first question falls within the purview of the FCC, but the second seems better suited to NASA or the DOD. Both track orbital objects to help prevent collisions, but neither is a regulatory body.
“There isn’t really a good coordinated policy on what we should be doing with regard to space debris,” says Stanford’s Manchester. “Right now, these people aren’t talking to each other effectively, and there’s no coherent policy.”
The issue is further complicated because the LEO satellites pass over many countries. The International Telecommunication Union performs a role somewhat like the FCC, assigning spectrums, but to operate within a country, a company must receive permission from that country.
The important takeaway is that it changes depending on where you are, and so if your satellite is moving like LEO satellites do, it better be capable of adjusting its communication spectrum.
“Do you really want SpaceX to have a monopoly of connectivity in a given region?” says Press. “Do they need to be regulated, and who can regulate them?
They are supernational. The FCC doesn’t have jurisdiction in other countries.”
That doesn’t exactly make the FCC toothless, though. Late last year, a small Silicon Valley startup called Swarm Technologies was denied permission to launch four prototype LEO communications satellites, each smaller than a paperback book.
The FCC’s primary objection was that the tiny satellites might be too difficult to track and thus be unpredictable and dangerous.