A secondhand story about Larry Page has become the latest talking point in the growing debate over whether AI compute belongs on Earth at all.
On a recent episode of the Invest Like The Best podcast, investor Patrick O’Shaughnessy recounted hearing that someone had asked the Google co-founder how much of the world’s computing would eventually end up in orbit. Page’s reported answer: five-nines, as in 99.999%. Writer Hosun Chung posted a clip of the exchange on X, summarizing it as “Larry Page reportedly has a number for how much computing will end up in space: five nines.”
Elon Musk, never one to miss a chance to talk up off-planet infrastructure, quote-tweeted the clip with a one-line endorsement: “Larry Page has made many accurate predictions and this is one of them.”
Page himself has not confirmed the figure publicly, and the claim is, by its own telling, hearsay — O’Shaughnessy said he’d “heard this story recently” rather than hearing it from Page directly. But coming from the co-founder of the company that built one of the largest terrestrial data center networks on Earth, and now runs its own orbital compute program, the number is striking enough that it’s being taken seriously by people who build this infrastructure for a living.
Why Musk is nodding along
Musk’s agreement isn’t surprising. He’s spent much of the past year talking up the idea that data centers belong in space, where solar power is constant, cooling is effectively free, and there’s no neighborhood to object to the noise or the water use. SpaceX has filed with the FCC for permission to deploy up to a million satellites as part of what it calls an “Orbital Data Center system,” and the company’s IPO prospectus reportedly points to deploying its first orbital AI compute satellites “as early as 2028.”
That pitch picked up new urgency this week. Just days before Musk’s reply to the Page clip, he posted that “data centers in space” were “looking pretty appealing right now” after Democrats swept the US midterms on a wave of local opposition to datacenter construction. Dozens of cities and counties passed measures restricting or banning new datacenters outright, and two governors, Greg Abbott in Texas and Jon Husted’s would-be successor in Ohio, lost races where datacenter backlash was a central issue. For Musk, orbital compute isn’t just a long-term engineering bet — it’s increasingly framed as an escape hatch from zoning boards, utility fights and angry voters.
He’s not the only one betting on space
Musk and SpaceX are far from alone in moving compute off-planet. Google has its own orbital ambitions: the company’s Project Suncatcher put a prototype satellite carrying TPUs into orbit at the start of October, part of a research effort to test whether solar-powered satellite constellations with high-bandwidth laser links can scale machine learning in space. Google has said it’s planning further test launches in 2027, though executives have been careful to describe Suncatcher as a research moonshot rather than something close to commercial deployment.
Nvidia-backed startup Starcloud has moved faster on paper, sending a data-center class GPU into orbit prior to Google and arguing that space offers energy costs roughly 10 times cheaper than terrestrial datacenters once the vacuum’s natural cooling is factored in. Starcloud’s CEO has gone as far as predicting that “in 10 years, nearly all new data centers will be being built in outer space.” Jeff Bezos has made a similar, if more conservative, bet — he’s said he expects datacenters running in space within 20 years.
Not everyone is convinced the economics or physics work out any time soon. OpenAI’s Sam Altman has publicly dismissed orbital datacenters as unrealistic in the foreseeable future, as has Softbank’s Masa Son, and skeptics point out that SpaceX’s million-satellite filing contains no hardware specifications, no mass budget and no firm deployment timeline — details that matter enormously when each satellite is meant to carry and cool meaningful amounts of compute.
The regulatory squeeze on Earth
What’s changed in the past week is the political backdrop against which these space pitches are being made. Public opposition to datacenters in the US has been building for most of the year, with local officials proposing well over a hundred moratoriums across dozens of states even before election day. In this month’s midterms, datacenter measures appeared on ballots in more than a dozen states, and polling found a clear majority of Americans opposed to new datacenter construction in their communities.
The result was a string of losses for the industry’s preferred candidates and a wave of new local restrictions, from charter amendments banning large datacenters outright to referendum requirements that make future projects far harder to approve. Markets reacted almost immediately: shares in CoreWeave, Oracle, Nebius and SpaceX itself all fell in the days after the results came in, on fears that state and local pushback could slow the AI buildout in some of its biggest markets.
None of this stops terrestrial datacenter construction — the capital already committed by the major AI labs is enormous, and the biggest projects underway aren’t going anywhere. But it does help explain why a secondhand quote attributed to Larry Page, suggesting that almost all compute will eventually leave Earth, found such a receptive audience in Elon Musk. Whether or not 99.999% turns out to be the right number, the argument for getting at least some meaningful share of compute off a planet with increasingly hostile zoning boards is one more of the industry’s biggest names now seem willing to entertain publicly.