The craziest part of SpaceX's IPO is also the most important

SpaceX's IPO filing this week wasn't about rockets. Or Starlink. Wall Street already understands those businesses. Launches are booming. Satellite internet is scaling profitably. Those stories are mature enough to model in spreadsheets.

The real pitch — the thing meant to justify the next decade of valuation growth — is something far stranger: orbital AI data centers.

In the filing, SpaceX described a plan to deploy "AI compute satellites" into sun-synchronous orbit starting as early as 2028. The idea is simple in theory and insane in practice: move AI infrastructure off Earth and into space, where solar power is effectively unlimited and cooling happens naturally through radiative heat dissipation.

The AI race is increasingly about who controls compute and can generate tokens most efficiently. SpaceX argues the bottleneck is now physical: power generation, data center construction, and chip manufacturing. The company believes terrestrial infrastructure cannot keep up with exploding AI demand, especially as reasoning models and AI agents consume exponentially more compute.

Hence the orbital pivot.

The logic is actually pretty coherent. In orbit, solar arrays receive near-constant sunlight. There's no atmosphere, no weather, and no NIMBYs blocking permits for another gigawatt-scale data center. SpaceX says space-based solar arrays can generate more than five times the energy per unit area of terrestrial systems. Combined with Starship launches and Starlink networking, Elon Musk sees a future where compute itself becomes a new space industry.

And here's the important part: this isn't just Musk freestyling another sci-fi fantasy. Google is pursuing the same idea.

Late last year, Google launched Project Suncatcher and published a paper proposing "space-based ML data centers" powered by solar satellites networked together with optical links. The paper argued that AI's future energy demands may require moving compute infrastructure into orbit.

Google has already tested radiation tolerance for its TPU chips in simulated space environments and is designing AI satellites with Planet Labs, the satellite specialist that went public in 2021. The search giant is also reportedly talking to SpaceX about launching future AI satellites into orbit.

Most importantly, Sundar Pichai endorsed the concept publicly.

"There's no doubt to me that a decade or so away, we'll be viewing it as a more normal way to build data centers," he told Fox News.

That quote matters. Pichai is probably the least theatrical CEO in big tech. If both Musk and Pichai independently conclude that AI compute eventually migrates into orbit, investors can't dismiss the idea as pure science fiction.

Enormous risks remain. SpaceX admits the technology may never become commercially viable. The launch cadence required is staggering. The engineering challenges are brutal.

But IPOs are about belief in the future and the promise of near-monopolies in tech.

And SpaceX just told investors the monopoly it wants to own is AI infrastructure in space.

Sign up for BI's Tech Memo newsletter here. Reach out to me via email at [email protected].