SpaceX Reveals Orbital AI Data Center Specs During IPO Week Roadshow

SpaceX has released its most detailed look yet at a planned orbital AI data center satellite, unveiling a 70-meter-wingspan structure during the same week the company is pitching investors on a $1.75 trillion public listing.

By Olivia Grant Edited by Maria Konash Published:
SpaceX Reveals Orbital AI Data Center Specs During IPO Week Roadshow
Musk unveiled the AI1 satellite on the eve of SpaceX's IPO roadshow. Image: Daniel Olah / Unsplash

Elon Musk shared the most detailed specifications to date for SpaceX’s planned orbital AI data center in an interview posted to X on Monday, timed to coincide with the opening of the company’s IPO roadshow. The satellite – which Musk described as a first-generation prototype – stands 20 meters tall with a wingspan of 70 meters, making it the largest spacecraft SpaceX has prepared for launch. The structure consists of a rack of AI processors supported by large solar arrays and liquid radiators for heat management.

Musk framed the design as a straightforward extension of technology SpaceX already uses in its Starlink satellite internet constellation, which the company has been operating since 2019. The orbital platform draws on the same manufacturing knowledge and component supply chains as Starlink, and Musk said SpaceX does not consider the engineering challenge particularly difficult relative to what the company already builds. SpaceX CFO Bret Johnson confirmed separately that the initial satellites will run Nvidia GPUs.

The solar panels for the platform will be produced at a new facility in Bastrop, Texas, where SpaceX already manufactures Starlink hardware. Musk referred to the new plant as the “Gigasat” factory and put its footprint at more than 11 million square feet. Construction is ongoing, with Musk saying the facility will be running at meaningful production volume before the end of 2027.

SpaceX has indicated it intends to eventually place up to one million orbital data center units into service. That figure forms a central pillar of the company’s investor materials, which attribute a total addressable market of $28.5 trillion to SpaceX’s combined businesses, with $26.5 trillion of that assigned to AI. The company is targeting a public listing valuation of $1.75 trillion this week.

The Strategic Logic

The push toward orbital computing reflects two pressures that have complicated AI infrastructure buildout on the ground. First, land-based data centers require large quantities of power, water, and physical space – resources that have increasingly drawn regulatory friction and community opposition in multiple U.S. states. Space-based platforms avoid those constraints by drawing on direct solar energy and radiating heat into the vacuum of space, where thermal management is a matter of engineering rather than permitting.

Second, compute capacity itself has become a competitive bottleneck. AI model training requires enormous quantities of processing power over extended periods, and the companies with the largest compute reserves have a structural advantage in developing more capable models. Placing that compute in orbit – at a scale of one million units – would represent a resource pool that no ground-based competitor has matched.

The Hardware Supply Chain

Beyond Nvidia GPUs for the initial deployment, SpaceX is building a longer-term chip supply through Terafab, a semiconductor fabrication facility it is developing in partnership with Tesla and Intel. Musk put the planned footprint of the Terafab facility at approximately 10 million square feet – roughly ten times the floor area of Tesla’s largest existing factory in Austin, Texas. Terafab will produce radiation-hardened chips designed specifically for the space environment, where standard commercial semiconductors degrade more quickly due to cosmic ray exposure. No production timeline for Terafab was disclosed.

The IPO Backdrop

The Monday reveal was not incidental to the timing. SpaceX kicked off its investor roadshow the same day, and the orbital data center program represents the primary justification for the company’s AI revenue projections – which Goldman Sachs, the lead underwriter, has forecast growing from $3.2 billion in 2025 to $322 billion by 2030.

Giving prospective institutional investors a concrete look at the physical hardware behind those projections, at the moment they are being asked to commit capital, is a deliberate sequencing of information. Whether the satellite’s specifications – a 70-meter prototype that Musk himself characterized as a draft of version one – are sufficient to support a $26.5 trillion market assumption is the question investors on the roadshow are now working through.

AI & Machine Learning, Cloud & Infrastructure, News, Space & Future Tech