Summary: SpaceX’s confidential pre-IPO presentation of S-1 warns that its plans for an orbital AI data center “involve significant technical complexity and unproven technologies and may not be commercially viable,” contradicting Elon Musk’s claim at Davos in January that space-based AI could be achieved within two to three years. The filing comes as SpaceX targets a $1.75 trillion IPO valuation and applies to the FCC for a million data center satellites, while rivals Starcloud, Google (Project Suncatcher) and Blue Origin pursue their own orbital computing programs.
SpaceX told potential investors it had plans for orbital artificial intelligence data centers in confidential filings filed by S-1 ahead of its IPO.involves 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 extensive and unique space-related risks that could cause them to malfunction or fail.” The disclosure, first reported by Reuters on Monday, is legally standard for a company approaching the largest initial public offering in history. It is also the remarkable bureaucratic sincerity of the same organization described by the chief executive. data centers in orbit like “mindless” three months ago.
At the World Economic Forum in Davos in January, Elon Musk said the cheapest place to deploy artificial intelligence would be in space “within two, maybe three years at the latest.” He called it a space-based sun.It is 10 times cheaper than terrestrial solar energy” because “you don’t need any batteries,” described the cooling problem as being solved simply by pointing a radiator three degrees Kelvin away from the sun, and predicted that within five years, more artificial intelligence capacity would be in orbit than on Earth. In February, SpaceX applied to the Federal Communications Commission to launch and operate up to one million satellites at an altitude of 2,500 km as the “SpaceX Orbital Data Center system.”use direct near-constant solar energy with little operation or maintenance costs.” The S-1, filed confidentially with the Securities and Exchange Commission ahead of a June listing of $1.75 trillion and $75 billion in growth, says otherwise.
The physics of the problem
The contradiction between Musk’s public statements and SpaceX’s legal disclosures is due to a number of engineering limitations that have not changed since Davos. In a vacuum, all heat transfer occurs by radiation. No convection, liquid cooling, no fans. To generate just one megawatt of heat at 20 degrees Celsius, an orbital data center would need about 1,200 square meters of radiator surface, the area of four tennis courts. The entire electrical system of the International Space Station produces only 0.2 megawatts of energy; ground-based hyperscale data centers are racing toward the gigawatt scale. If the radiators you need to use are heavier than the servers they cool, the three degrees background temperature of the space is irrelevant.
💜 of EU technology
The latest rumblings from the EU tech scene, a story from our wise founder Boris and some questionable AI art. Free in your inbox every week. Register now!
Power is equally limited. Solar panels in orbit receive about five times as much energy as Earth without atmosphere, weather or nighttime in certain orbits. But to produce one gigawatt at 30% cell efficiency requires about a square mile of solar array in Earth orbit. The ISS generates 0.2 megawatts of power from arrays the length of a football field. The scale of a single hyperscale data center consuming up to gigawatts on Earth would require deploying and maintaining a solar infrastructure larger than anything humans have ever built in space.
Equipment obsolescence may be the most underestimated constraint. GPUs become obsolete as new architectures emerge every two to three years. Shelves on Earth are constantly being replaced. Each equipment replacement in orbit requires a launch, docking, or robotic servicing mission. Exposure to radiation causes bit flips and permanent circuit damage. Radiation-hardened chips are several generations behind commercial processors. Triple modular redundancy, running three parallel systems and receiving the majority of the volume, will triple the hardware requirements. The AI’s increasing power requirementsThe IEA projects that data center electricity consumption will exceed 1,000 terawatt-hours by the end of 2026. The question is, does solving them in orbit create more problems than it solves?
The competitive landscape in orbit
SpaceX isn’t the only company making orbital calculations, which makes the S-1 waiver more strategically important than the standard risk factor. Starcloud, formerly Lumen Orbit, launched its first high-powered GPU into orbit in November 2025, the Nvidia H100, which represents 100 times the computing power of those running in space. In December, Starcloud became the first company to run a large language model, Google’s Gemma, and launch LLM training in orbit. By March 2026, it had raised $170 million at a valuation of $1.1 billion, the fastest unicorn in Y Combinator history. Its next satellite, with a capacity of 200 kilowatts and a price of about $0.05 per kilowatt-hour, is planned for October.
Google’s Project Suncatcher, in partnership with Planet Labs, plans to launch two test satellites carrying Google TPUs by early 2027 and envisions one-kilometer arrays of computing clusters of 81 satellites in dawn-dusk sun-synchronous orbit. Google’s analysis suggests that launch costs could drop below $200 per kilogram by the mid-2030s, making space data centers comparable to ground-based power costs at that point. Nvidia announced the Vera Rubin Space-1 chip system designed specifically for orbital data centers. Blue Origin has filed its FCC filing for 51,600 data center satellites. a16z-funded startup Orbital is building a satellite constellation with artificial intelligence. The idea is not foreign. It involves serious capital and serious engineering talent. SpaceX’s S-1 is notable because the company, which oversees the constellation of launch vehicles and satellites best positioned to perform orbital computing, tells investors it may not.
Terrestrial alternatives
The S-1 filing comes in a week in which land-based alternatives have absorbed large investments. Massive AI infrastructure deals Like Meta’s $27 billion commitment to Nebius, it shows the scale of spending on terrestrial computing. Nuclear powered AI data centers Valar Atomics is raising $450 million in $2 billion in private funding to build small modular reactors designed for artificial intelligence workloads. The US Department of Energy has identified 16 federal sites for data center construction adjacent to existing nuclear facilities. By 2026, 18 nuclear-powered AI facilities with a total capacity of 31.2 gigawatts are being tracked globally. Microsoft’s Project Natick deployed an undersea data center capsule designed for artificial intelligence workloads in February 2025. The tech industry has spent an estimated $580 billion in 2025 to transform deserts and abandoned factories into GPU-filled facilities.
The pattern is consistent: every approach to the AI power challenge that keeps servers on Earth or, at most, underwater attracts more capital and moves faster than orbital alternatives. Nuclear reactors are a proven technology that can be adapted to a new use case. Orbital data centers are an unproven technology proposed for a use case that doesn’t require them. The S-1 language suggests that SpaceX’s own engineers and lawyers recognize the difference, even if the company’s public messaging doesn’t.
IPO context
The S-1 file serves two masters. SpaceX needs to present its orbital data centers as a credible growth story to justify a $1.75 trillion valuation, the highest for a pre-IPO company. It must also disclose the risks clearly enough to protect itself from securities litigation if the plans do not materialize. The result is a document that simultaneously promotes and rejects the same initiative. This is not unusual in IPO filings. It is unusual for the chief executive to spend the last three months calling this initiative 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 clearly motivated by orbital data centers. Musk said the integration of Starlink’s global satellite network with xAI’s large language models is the main reason. Musk’s AI chip ambitions It includes special processors for orbital deployments through the Terafab project with Intel. One million satellites in FCC filings represents a hundredfold increase in the current population of low Earth orbit. Ars Technica estimates the cost of deploying barebones as “at least $1 trillion.” The vast majority of more than 1,000 public comments submitted to the FCC opposed the plan, citing debris, light pollution and the risk of Kessler syndrome, a chain of collisions that could render entire orbital altitudes unusable.
SpaceX may finally prove that orbital computing works. The company has a track record of achieving what others say is impossible, most notably reusable rockets. But the S-1 filing is not the language of a company that has solved the problem. It is the language of a company seeking credit for trial and error protection. The gap between Davos in January and the SEC in April is the gap between the pitch and the avenue. Both are real. Only one is legally responsible.
