Elon Musk’s SpaceX is weeks from filing its IPO prospectus, and one of the ideas the company has been pitching is putting AI data centers in space.

A new Wall Street Journal video explored the feasibility of a scheme SpaceX’s own pre-IPO filing reportedly warns “may not achieve commercial viability.”

SpaceX asked the FCC in late January to authorize up to one million solar-powered satellites for what it calls the “Orbital Data Center System,” framing the project as a “first step towards becoming a Kardashev II-level civilization.”

Why The Pitch Matters

NYU Stern professor Aswath Damodaran valued SpaceX at $1.2 trillion in late April, a third below the $1.75 trillion to $2 trillion range the company is reportedly targeting for its IPO.

His base case was built on Starlink, the launch business, and xAI. Orbital data centers, if they work, could push the valuation well past that.

Musk called space-based AI a “no-brainer” at Davos in January and predicted on Dwarkesh Patel’s podcast that orbital compute would beat terrestrial on cost within 30 months.

What It Would Actually Take

A Wall Street Journal video walked through the engineering with a computer architect.

The benefits of going orbital are real: near-constant solar exposure in sun-synchronous orbits, no water or land demands, and solar panels that work five times as efficiently as on Earth, per Musk.

The drawbacks are bigger. Cooling cannot use air or water, so heat has to dissipate through heavy closed-loop radiator panels that bleed infrared into the vacuum.

Radiation is the second problem. A single high-energy particle striking a GPU can flip a bit and turn the right answer into the wrong one, forcing mitigations like error correction, running the same compute across three separate GPUs, or physical shielding.

Latency is the third problem: queries have to round-trip to orbit and back, and bandwidth limits would force satellites to be preloaded with training data rather than streamed from Earth.

Musk says that the project would need 10,000 Starship launches per year. SpaceX flew 165 missions in 2025, almost entirely on Falcon 9, a smaller vehicle, and Starship still has not completed a fully reusable test flight.

What Prediction Markets Say

Polymarket gives SpaceX a 91% chance of going public by August 31, and prices a closing market cap above $2 trillion at 59%.

Morgan Stanley leads underwriter odds at 47%, and the new space ticker is expected to list on NASDAQ.

For investors who can’t wait for the June listing, Tesla Inc (NASDAQ:TSLA) remains the cleanest public proxy on the Musk trade.

Image: Shutterstock