Startups race to deploy orbital data centers for AI computing despite massive technical hurdles

The Orbital Data Center Hype Machine Is Already in Orbit

Why the stars -- and the math -- won't align for space compute anytime soon "The lowest-cost place to put AI will be in space, and that will be true within two years, maybe three at the latest," SpaceX founder Elon Musk told the World Economic Forum in Davos this past January, as his company was preparing to go public. Later that month, SpaceX filed an application with the Federal Communications Commission for an orbital data center constellation of up to 1 million satellites in low Earth orbit, 500 to 2,000 kilometers above Earth. And just three days before the IPO, he discussed some initial design specifications for a new AI-1 satellite data center in a video interview. Musk is prone to hyperbole when it comes to timelines. Full self-driving cars by 2017. First human mission to Mars in 2024. Ten thousand Optimus humanoid robots by the end of 2025. Et cetera. For orbital data centers, which he says will be a cost-effective alternative to terrestrial data centers within three years, the math won't make sense for several years, if ever. Consider this: There are roughly 14,500 active satellites in orbit. Musk's Starlink constellation accounts for about two thirds of those. Both the launch cadences and satellite-manufacturing capacity would have to scale up astronomically to deploy a million orbital data center satellites. For context, there have been roughly 7,000 orbital launches in all of human history. To loft 1 million satellites into low Earth orbit on SpaceX's Starship, which is designed to carry up to 60 satellites per vehicle, would require 16,666 launches exclusively devoted to satellite deployments. Considering that SpaceX launched a record 165 orbital missions in 2025, even at 10 times that cadence, it would take a decade. And how long would it take to build 1 million satellites, given Starlink's current pace of around 4,000 per year and a generous tenfold increase in capacity? Short of a manufacturing revolution, try 25 years. The reality is that the vision of massive constellations of orbital data centers is nowhere close to being realized. As this month's cover story, "Why Orbital Data Centers Are So Hard" by Andrew Cavalier of ABI Research, makes clear, the reality is that the vision of massive constellations of orbital data centers is nowhere close to being realized. Dina Genkina, IEEE Spectrum's computing and hardware editor, put the idea into perspective: "Starcloud (a startup that has applied to the FCC for an 88,000 orbital data center satellite constellation) sent one Nvidia H100 GPU in space so far. Their radiator was too weak to let the chip run at full power." As Cavalier shows, cooling even a single Nvidia H100 GPU in space is difficult: It draws 700 watts, which will require 1.4 square meters of radiator at 60 °C. A 40-kilowatt rack of servers will need an 80-m² radiator; a 100-megawatt data center will require 2,500 of those radiators. Some astronomers are understandably concerned that a million satellites with giant radiative wings would blot out the stars. So if the economics doesn't make sense, if the chips are at the mercy of the radiative ravages of space, and if humanity will lose its view of the stars, not to mention increasing the risk of triggering the Kessler syndrome, why are the hyperscalers hyping orbital data centers? Genkina offered the obvious answer: sweet, sweet moolah. "The Elon Musk part of it is honestly genius because he's got xAI building the data centers, SpaceX sending them to space, and Tesla building solar panels," Genkina says. "It's almost like he's paying himself."

US startup proposes 100,000-satellite data center constellation

The next major battleground for the survival of artificial intelligence is being fought roughly 300 miles above our heads. And now, to challenge the tech giants, a newly launched, five-month-old startup has set a big, ambitious goal. Los Angeles-based startup Orbital Compute, Inc. recently laid its cards on the table. It has filed an audacious plan with the Federal Communications Commission (FCC) to launch up to 100,000 AI-focused data-center satellites into low Earth orbit. At full scale, the constellation promises to deliver 10 gigawatts of pure computing power. In comparison, the figure matches the total new electricity capacity added to the entire United States power grid last year. In the recent years, various companies have been moving server infrastructure entirely into space. The key reasons are to combat severe strain on the power grid, community backlash, and water shortages caused by resource-intensive AI data centers. With the orbiting data centers, tech companies can overcome Earth's land and electricity constraints altogether, using continuous solar energy and the cold void of space for natural cooling. The company's founder, Euwyn Poon, is an unconventional choice to lead a space race. He previously founded Spin, the dockless electric scooter company that he successfully scaled and sold to Ford. Poon transitioned to space after buying an Nvidia GPU to rent out on Earth, where he hit a wall. Reportedly, he quickly realized that the ultimate challenge for artificial intelligence is electricity. And space holds the potential to solve the power overnight. Each planned Orbital satellite will act as a flying, high-density server rack powered by a massive 100-kilowatt solar array. The fleet would be positioned in sun-synchronous orbits to bask in perpetual, uninterrupted sunlight. Furthermore, cooling becomes an entirely different game. Down here, server farms require millions of gallons of water to keep from melting. In orbit, the hardware will radiate its blistering heat directly into the freezing void of space. "The demand for AI compute is outrunning what we can reasonably build on the ground -- we're short on power, land, and water all at once. Space solves all three. Sunlight is constant, cooling is free, and there's no neighborhood to disrupt. We think the next generation of data centers won't be built in the desert -- they'll be built in orbit," Poon noted. Of course, the plan sounds insane to critics. The vacuum of space lacks air, so heat must be dissipated solely by radiation -- a slow thermodynamic process. Space radiation can also fry sensitive chips, and launching thousands of two-ton objects threatens to choke our orbit with space junk. OpenAI CEO Sam Altman even dismissed the concept of orbital data centers as "ridiculous" earlier this year. Yet, serious capital is moving. Orbital recently closed a $5 million pre-seed round to advance the development. Meanwhile, heavyweights like SpaceX and Blue Origin are quietly sketching out their own orbital compute strategies, anticipating that next-generation heavy-lift rockets like Starship will make mass satellite deployment incredibly cheap. Orbital isn't waiting around. The startup plans to launch a tiny, single-GPU demonstration payload on a SpaceX Falcon 9 rocket next year to test how Nvidia chips hold up against space radiation. If successful, their first full-scale satellite, Orbital-1, will head up in 2028.

Orbital plans space data centres to power AI, seeks FCC clearance for 100,000 satellites

A US startup, Orbital, is looking to set up space-based data centres to meet the surging demand for AI computing power. The company aims to launch its first data centre satellite next year, with plans for 100,000 satellites in low Earth orbit to deliver 10 gigawatts of compute. Orbital plans to scale up deployment towards the end of the decade when the SpaceX-owned Starship will come online and significantly reduce the launch cost, Poon said. It sounds like science fiction, but US-based satellite startup Orbital is looking to set up data centres in space that may power responses given by chatbots as artificial intelligence (AI) drives an unprecedented demand for computing power. Orbital founder and CEO Euwyn Poon told ET it plans its first data centre satellite launch next year on a shared payload. The company has filed an application with the Federal Communications Commission (FCC) seeking clearance to have 100,000 satellites in the low earth orbit (LEO) between 500 to 800 kilometres to deliver 10 gigawatts of compute to help process AI workloads. Orbital plans to scale up deployment towards the end of the decade when the SpaceX-owned Starship will come online and significantly reduce the launch cost, Poon said. "I actually agree that large-scale infrastructure, the kilometres-wide kind, is still science fiction in space today," he told ET. "But what's practical now is what Starlink already demonstrates: you can put up small satellites -- 10,000, 50,000, 100,000 of them -- as Starship brings launch costs down. They're large, yes, but not crazy large. So, the confusion is really about scale."

Elon Musk Says Space Is 'the Only Way to Scale at Scale' for AI Computing Amid Earthbound Bottlenecks - S

Elon Musk on Sunday emphasized the need for orbital infrastructure to prevent artificial intelligence (AI) computing from being held up by Earth's limits and regulatory bottlenecks. Musk Says Space Can Bypass Earth's Bottlenecks Replying to a post about space-based data centers, Musk said, "Space is the only way to scale at scale." Musk was responding to a post by the popular X account X Freeze, which argued that massive AI data centers on Earth face delays from lawsuits, land fights, power-grid constraints and local opposition. The post positioned orbit as a way around those bottlenecks and included a conceptual "AI1 satellite" diagram showing a 150-kilowatt peak compute payload, solar arrays and deployable liquid radiators. Starship And AI1 Anchor Musk's Orbital Vision In early 2026, SpaceX filed plans with the Federal Communications Commission for up to 1 million "orbital data center" satellites. The filing said the constellation would power advanced AI models and operate between 500 kilometers and 2,000 kilometers above Earth. Musk has said Starship would be central to the plan because its heavy-lift capacity could enable the deployment of large numbers of compute satellites. SpaceX also unveiled the first-generation AI1 satellite design last month. Musk said the satellite would be simpler than Starlink hardware because it would need solar cells, radiators and laser links, but not Starlink's more complex communications antennas. Scientists Warn Satellites Could Threaten Astronomy The supply chain remains a major question. Musk has also announced Terafab, a proposed Texas chip project meant to produce 1 terawatt of compute power per year, much of it for space use. SpaceX's manufacturing roadmap is also ambitious. The company plans an 11-million-square-foot Gigasat facility in Bastrop, Texas, targeting AI satellite production by late 2027 and an eventual goal of 100 gigawatts of annual space AI compute by around 2030. Scientists warn the vision could carry steep costs. The European Southern Observatory has previously warned that large satellite constellations can harm astronomy, ecosystems and air quality, while critics say cooling server clusters in orbit and assembling them at scale remain unproven engineering challenges. According to Benzinga Edge Stock Rankings, SpaceX shares continue to trend bearish across the short, medium and long term. Photo courtesy: Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

US Startup Plans Data Centres' Space Odyssey

It sounds like science fiction, but US-based satellite startup Orbital is looking to set up data centres in space that may power responses given by chatbots as artificial intelligence (AI) drives an unprecedented demand for computing power. Orbital founder and CEO Euwyn Poon told ET it plans its first data centre satellite launch next year on a shared payload. The company has filed an application with the Federal Communications Commission (FCC) seeking clearance to have 100,000 satellites in the low earth orbit (LEO) between 500 to 800 kilometres to deliver 10 gigawatts of compute to help process AI workloads. Orbital plans to scale up deployment towards the end of the decade when the SpaceX-owned Starship will come online and significantly reduce the launch cost, Poon said. "I actually agree that large-scale infrastructure, the kilometres-wide kind, is still science fiction in space today," he told ET. "But what's practical now is what Starlink already demonstrates: you can put up small satellites -- 10,000, 50,000, 100,000 of them -- as Starship brings launch costs down. They're large, yes, but not crazy large. So, the confusion is really about scale." Technology giants including Elon Musk's SpaceX, Google and Jeff Beans Blue Origin besides star-tups such as Starcloud (formerly Lumen Orbit) and Kepler Com munications are looking at either building or facilitating satellite hasedcompute that could harness uninterrupted solar energy and eliminate the need for millions of litres of water to cool down pro cessors In fact, Musk has announced plans to launch one million satel lites for Al data contres in space and has filed an application with POC Poon expects orbital compute to be complementary to terrestrial omes for the next 10 years or so "If total data centre capacity to day is on the order of 100 gigawatts on earth, we can offload a ma ningfulchunk. Our 100,000 satelli tes get to 10 gigawatts-and with multiple companies, that alone could roughly double today's ca pacity" he said. Orbital's compute in space will be primarily for indence, where trained Al models make predic tions on new data, Poon said. "We're betting inference will be the bulk of Al workloads going forward, because the models have reached maturity and consumers and applications are now discove ring the use cases," he said The satellites that will be used for Al data centres would be smal ler-roughly the size of a fridge - from where the solar panels and radiators extend out. "It's ally server racks in space-frid go-sized units at scale," said Poon who previously built a micro-mo-bility infrastructure company and sold it to Ford "The only genuinely impractical part today is the launch cost, which we're all betting Starship that it isn't economically prudent to scale a fleet at present. "A satellite costs us roughly $5 million, and launching it on a Fal-comicould cost another $5 million or more. We need launch costs to come down to around $100,000 with Starship-a lot of the econo mics rests there," he said.