Elon Musk called Sam Altman a scammer on social media over the weekend, and Altman fired back by questioning Musk's plan to investors. The dispute centers on SpaceX's idea to launch artificial intelligence computers into orbit—machines that would do the computational work for AI systems from space instead of from Earth.

Altman's response was direct: "homeboy you're the one sellling [sic] public market investors on short-term space datacenters." Musk answered simply: "we start flying them next year."

The disagreement is about whether putting AI computers in space actually makes sense. SpaceX wants to launch data centers—facilities full of computing machines—into orbit, where they would process AI inference work. Inference is the moment when a trained AI model actually answers a question or performs a task. Currently, all of this happens in buildings on Earth.

SpaceX is valued at two trillion dollars in the stock market. According to reporting by TechCrunch, much of that value depends on this space data center concept working. The trouble is, no one has ever built this kind of infrastructure before. SpaceX would use its Starship rocket to launch it, with a test flight scheduled around July 16, 2026. During meetings with investors, SpaceX admitted that Starship will not be fully reusable anytime soon—the company will need to use up the rocket on each launch rather than catching and reusing it. That is important because SpaceX's entire business case has rested on being able to reuse rockets cheaply.

Altman's concerns are not new. In February, months before this weekend's exchange, he called Musk's space data center plan "ridiculous" given how much AI computing is actually needed.

There are real engineering problems with putting data centers in space. Data centers get very hot because the computers inside them use enormous amounts of electricity. On Earth, that heat is carried away by air and cooling systems. In space, there is no air. All the heat has to be shed through special radiators—large metal panels that beam heat away into the vacuum. The radiators themselves become very heavy, and heavy things are expensive and difficult to launch into orbit. Modern AI chips generate so much heat that the radiators needed to cool them would be impractically large and heavy.

Other problems exist too: sending data up to space and back down takes time, which slows things down. Computers not built for the radiation environment of space can fail. And if something breaks on an orbital data center, there is no practical way to fix it.

These are not invented complaints. Infrastructure experts have raised them repeatedly. Altman is essentially saying what many people in the industry already think.

It is worth stepping back from the personal quarrel between Musk and Altman—they have disagreed publicly for years—and looking at what Altman is actually pointing at. When he suggests Musk is "selling" this vision to stock market investors, he is raising a different kind of question: whether SpaceX's enormous valuation can be justified by a product that has never been built or tested.

SpaceX is worth two trillion dollars partly because the company has a proven track record launching rockets reliably. That is real. But the orbital data center idea is unproven. And SpaceX's own materials to investors admitted that Starship won't work the way the long-term cost case assumes—meaning the financial story underlying the entire valuation has a gap in it.

The July 16 Starship launch, if it happens, will be a vehicle test, not a test of space data center hardware. So it will not settle the argument by itself. But investors watching that launch—and the ones that follow—will be trying to figure out whether the orbital data center vision is realistic or hype.

The larger question here is simple: when a company is valued at two trillion dollars because of a product idea that has never flown, how much of that value is based on real capability and how much is based on a story. That is the argument underneath the insults.