Sam Altman and Elon Musk’s Space Data Center Feud Comes Down to Starship’s Timeline

Sam Altman and Elon Musk traded barbed posts online over the weekend, and buried in the insults was a pointed claim about one of the AI industry’s biggest infrastructure bets: orbital data centers. Musk accused Altman of being a scammer. Altman shot back, “homeboy you’re the one sellling [sic] public market investors on short-term space datacenters.” Setting the name-calling aside, Altman’s point lines up with what space and satellite experts have been saying for months, according to TechCrunch’s report on the exchange: the hardware and economics needed to make orbital AI compute real are not close to ready.

What Altman and Musk Actually Said

TechCrunch senior reporter Tim Fernholz, author of a book on the private space industry, described the exchange as drawing “new attention to the gap between vision and reality for the space-compute business.” Responding to Musk’s accusation, Altman argued that Musk is the one selling public market investors on a short timeline for space-based data centers. Musk’s reply, quoted by TechCrunch: “We start flying them next year.” Fernholz’s own assessment follows: SpaceX could plausibly launch a single data-processing satellite next year, but the harder question is when the company can launch and manufacture them at scale, which he calls “likely a question for the 2030s.”

That framing matters because SpaceX is not a bystander in this argument. The company’s own IPO pitch to investors leans heavily on the orbital data center bet, and Musk is the one making it.

The Bet Behind SpaceX’s Valuation

SpaceX took its IPO road show to investors with a targeted valuation of $1.75 trillion, according to Scientific American, and TechCrunch reports that the company’s orbital data center plans are the main driver behind a valuation it now puts at $2 trillion. The pitch is built on three categories: space, connectivity, and artificial intelligence. Two of the underlying bets are unproven: a fully reusable Starship, and a constellation that could eventually run to one million satellites performing AI computation on solar power gathered in orbit. Musk unveiled the first detailed design of the constellation’s prototype, called the AI1 satellite, just days before the IPO.

The experts Scientific American interviewed are not dismissive of the idea, but they are far from convinced it is close. Caleb Henry, director of research at Quilty Space, described the push as SpaceX reinventing itself a second time: from launch provider to satellite internet company and now, potentially, to an AI company running its own space-based infrastructure. He drew a clear line between what SpaceX has already proven (launch and constellation operations) and what it has not: “I can’t tell you if it scales efficiently to an orbital data center,” he said. “But I know who is in the lead.”

Other experts focused on the scale problem. Hugh Lewis, a professor of astronautics at the University of Birmingham, told Scientific American that the proposed AI satellites look larger and more complex than today’s Starlink satellites, which raises collision risk and gives elaborate cooling systems more ways to fail. He pointed to a contradiction: SpaceX has already lowered the orbit of some Starlink satellites to cut collision risk, even as it seeks approval for a constellation far larger, and asked: “If they can’t keep 4,500 safe, how can they expect to keep a million?” Jonathan McDowell, an honorary professor at Durham University’s Space Research Center who maintains a closely watched public catalog of objects in orbit, raised a related concern about what happens as satellites fail or retire at that scale: “Even very small percentages of failures lead to a very large number of space mines.”

Flight 13 Is the Near-Term Reality

The next concrete test of SpaceX’s Starship program is Flight 13, targeting a 90-minute launch window that opens at 5:45 p.m. CT on July 16 from Starbase, Texas, according to Dataconomy. It is the second flight of the Starship V3 configuration, which debuted on Flight 12 in May. The mission will deploy 20 next-generation Starlink V3 satellites, adding 60 terabits per second of capacity to SpaceX’s internet constellation. It will also attempt an in-space relight of a single Raptor engine and a controlled splashdown of the upper stage in the Indian Ocean, while the Super Heavy booster attempts a landing at sea rather than the tower catch SpaceX has used on previous flights to demonstrate rapid reuse.

None of that is an orbital data center payload. Flight 13 is a Starlink launch and an engine test, flying a booster recovery method that is a step down from the tower catch, not a step toward it. That gap, between what SpaceX is actually flying this month and what its AI1 satellite renderings promise, is close to exactly what Altman was pointing at.

Why It Matters

Orbital data centers make for a compelling pitch on paper: no land, water, or power grid constraints holding back AI buildout. But the engineers and analysts closest to the problem keep arriving at the same conclusion Altman did, minus the insult. Routine, cheap, reusable Starship flights have to arrive first. Satellite collision and disposal math at constellation scale has to get solved. Neither looks close, and Scientific American’s sourcing puts a realistic timeline closer to Fernholz’s “2030s” than to Musk’s “next year.” Enterprise infrastructure teams weighing AI capacity plans should treat orbital compute as a bet worth tracking, not a near-term substitute for terrestrial data center capacity.

Sources

• TechCrunch: Sam Altman’s space data center trash talk is what most experts already believe
• Scientific American: SpaceX IPO valuation depends on Starship and orbital AI data centers
• Dataconomy: SpaceX schedules Starship Flight 13 for July 16 launch

Featured image: SpaceX Starship and Super Heavy igniting engines at liftoff during the Flight 5 test (IFT-5). Photo by Steve Jurvetson, licensed CC BY 2.0, via Wikimedia Commons; cropped, resized, and converted to WebP.