Summary: SpaceX’s confidential pre-IPO S-1 filing warns that its orbital AI data center plans “involve significant technical complexity and unproven technologies, and may not achieve commercial viability,” contradicting Elon Musk’s January statement in Davos that space-based AI was a “no-brainer” that could be achieved in two to three years. The filing comes as SpaceX is targeting a $1.75 trillion IPO valuation and has filed with the FCC for one million data center satellites, while competitors Starcloud, Google (Project Suncatcher) and Blue Origin pursue their own orbital computing programs.
SpaceX told potential investors in its confidential pre-IPO S-1 filing that its plans for orbital AI data centers “They involve significant technical complexity and unproven technologies, and may not achieve commercial viability.The company warned that any future space-based computing infrastructure will operate “in the harsh and unpredictable environment of space, exposing them to a wide and unique range of space-related risks that could cause them to malfunction or fail.” The disclosure, first reported by Reuters on Monday, is a legal standard for a company approaching what could be the largest initial public offering in history. It is also a remarkable display of bureaucratic candor from the same organization whose executive director described data centers in orbit like a “obvious“Three months ago.
At the World Economic Forum in Davos in January, Elon Musk said the lowest-cost place to deploy AI would be space “within two years, maybe three at the latest.” He called space-based solar energy “10 times cheaper than terrestrial solar” because “you don’t need batteries“, described the cooling problem as solved by simply pointing a radiator away from the sun at three degrees Kelvin, and predicted that within five years there would be more AI capability in orbit than on Earth. In February, SpaceX filed with the Federal Communications Commission to launch and operate up to one million satellites as the “SpaceX orbital data center system” at altitudes between 500 and 2,000 kilometers. The presentation described satellites that “Directly harness nearly constant solar energy with little operation or maintenance cost.” The S-1, filed confidentially with the Securities and Exchange Commission ahead of its planned listing in June at a $1.75 trillion valuation and $75 billion raise, says something different.
The physics of the problem.
The contradiction between Musk’s public statements and SpaceX’s legal disclosures relates to a set of engineering limitations that have not changed since Davos. In a vacuum, all heat dissipation occurs through radiation. There is no convection, liquid cooling or fans. To radiate just one megawatt of heat at 20 degrees Celsius, an orbital data center would need approximately 1,200 square meters of radiator surface area, the area of four tennis courts. The entire electrical system of the International Space Station produces only 0.2 megawatts; Hyperscale terrestrial data centers are racing toward gigawatt scale. The three degrees of ambient temperature of the space are irrelevant if the radiators necessary to operate it weigh more than the servers they cool.
The 💜 of EU technology
The latest rumors from the EU tech scene, a story from our wise founder Boris and some questionable AI art. It’s free, every week, in your inbox. Register now!
Power is equally limited. Solar panels in orbit receive about five times more energy than on Earth, with no atmosphere, weather, or night in certain orbits. But it would take about one square mile of solar panels in Earth orbit to produce one gigawatt at 30% cell efficiency. The ISS produces 0.2 megawatts from arrays that stretch the length of a football field. Scaling to the gigawatts consumed by a single hyperscale data center on Earth would require deploying and maintaining solar infrastructure orders of magnitude larger than anything humans have built in space.
Hardware obsolescence may be the most underestimated limitation. GPUs depreciate as new architectures emerge every two to three years. On Earth, racks are continually changed. In orbit, each hardware replacement requires a launch, docking or robotic maintenance mission. Radiation exposure causes bit changes and permanent circuit damage. Radiation-hardened chips are several generations behind commercial processors. Triple modular redundancy, running three parallel systems and getting the majority vote, would triple the hardware requirements. He The increasing energy demands of AIwhich the IEA projects will boost data center electricity consumption to more than 1,000 terawatt-hours by the end of 2026, are real. The question is whether solving them in orbit creates more problems than it solves.
The competitive landscape in orbit
SpaceX is not the only company pursuing orbital computing, making the S-1 disclaimer more strategically significant than a standard risk factor. Starcloud, formerly Lumen Orbit, launched the first high-power GPU into orbit in November 2025, an Nvidia H100 that represented 100 times more computing than had ever operated in space. In December, Starcloud became the first company to run a large language model, Google’s Gemma, and the first to conduct LLM training in orbit. As of March 2026, it had raised $170 million at a valuation of $1.1 billion, the fastest unicorn in Y Combinator’s history. Its next satellite, targeting 200 kilowatts and costing about $0.05 per kilowatt-hour, is scheduled for October.
Google’s Project Suncatcher, a partnership with Planet Labs, plans to launch two test satellites carrying Google TPUs in early 2027 and envisions one-kilometer arrays of 81-satellite computing clusters in sun-synchronous orbit between dawn and dusk. Google analysis suggests that launch costs may fall below $200 per kilogram by the mid-2030s, which would make space data center costs comparable to terrestrial power costs at that time. Nvidia announced Vera Rubin Space-1, a system chip designed specifically for orbital data centers. Blue Origin filed its own application with the FCC for 51,600 data center satellites. Orbital, the startup funded by a16z, is building a constellation of AI satellites. The idea is not marginal. It’s attracting serious capital and serious engineering talent. SpaceX’s S-1 is notable precisely because the company that controls the launch vehicles and satellite internet constellation, the company best positioned to make orbital computing work, is the one telling investors that maybe not.
The terrestrial alternatives
The disclosure of the S-1 comes in a week in which terrestrial alternatives are absorbing huge investments. Massive AI Infrastructure Deals Such as Meta’s $27 billion commitment to Nebius illustrate the scale of spending on terrestrial computing. Nuclear-powered AI data centers are attracting targeted funding, and Valar Atomics has raised $450 million at a $2 billion valuation to build small modular reactors designed specifically for AI workloads. The US Department of Energy has identified 16 federal sites for the construction of data centers adjacent to existing nuclear facilities. By 2026, 18 nuclear-powered AI facilities with a combined capacity of 31.2 gigawatts will be monitored worldwide. Microsoft’s Project Natick deployed an underwater data center pod designed for AI workloads in February 2025. The tech industry spent approximately $580 billion in 2025 converting deserts and abandoned factories into GPU-packed facilities.
The pattern is consistent: Each approach to the AI power problem that keeps servers on Earth, or at most underwater, is attracting more capital and progressing faster than orbital alternatives. Nuclear reactors are a proven technology that is being adapted to a new use case. Orbital data centers are an unproven technology that is proposed for a use case that may not require them. The S-1 language suggests that SpaceX’s own engineers and lawyers recognize the distinction, even if the company’s public messaging has not caught up.
The IPO context
The S-1 filing serves two masters. SpaceX needs to present orbital data centers as a credible growth story to justify a $1.75 trillion valuation, the highest ever for a pre-IPO company. You also need to disclose risks clearly enough to protect yourself from securities litigation if plans don’t materialize. The result is a document that simultaneously promotes and denounces the same initiative. This is not unusual in IPO filings. It is unusual that the chief executive has spent the previous three months describing the initiative as inevitable, obvious and cheaper than the alternatives.
The SpaceX-xAI merger in February, an all-stock transaction that valued the combined entity at $1.25 trillion, was explicitly motivated by orbital data centers. Musk said the integration of Starlink’s global satellite mesh with xAI’s large language models was one of the main reasons. Musk’s ambitions for AI chips through the Terafab project with Intel they include dedicated processors for orbital deployments. The million satellites in the FCC filing would represent a hundredfold increase over the current low-Earth orbit population. Ars Technica estimated the basic implementation cost at “at least 1 billion dollars.The vast majority of more than 1,000 public comments to the FCC opposed the plan, citing debris, light pollution and the risk of Kessler syndrome, a chain of cascading collisions that could render entire orbital altitudes useless.
SpaceX may prove that orbital computing works. The company has a history of achieving what others considered impossible, most notably reusable rockets. But the S-1 filing is not the language of a company that has solved the problem. It’s the language of a company that wants credit for trying and protection if it fails. The gap between Davos in January and the SEC in April is the gap between a proposal and a prospectus. They are both real. Only one carries legal responsibility.
