Elon Musk called Sam Altman a scammer in a post on X over the weekend, and Altman fired back with a jab at Musk's fundraising pitch for orbital data centers. The exchange, reported by TechCrunch, continues a public feud between the two men that has simmered since at least February, when Altman first dismissed Musk's space-based compute ambitions.

Altman's reply read, typo intact: "homeboy you're the one sellling [sic] public market investors on short-term space datacenters." Musk answered with a single line: "we start flying them next year." Neither man's post elaborated further, and neither company has issued a formal statement expanding on the exchange.

The dispute centers on SpaceX's plan to launch a fleet of orbital data centers designed to perform AI inference workloads off-planet. That plan is, according to TechCrunch, the primary driver behind SpaceX's two-trillion-dollar valuation — a figure that puts enormous weight on a hardware concept that has not yet flown. TechCrunch

Starship, the vehicle SpaceX would presumably rely on to loft this infrastructure, was expected to attempt its thirteenth test flight as soon as July 16, 2026. During its IPO road show, SpaceX conceded that Starship may not achieve full reusability in the near term and will need to expend its second stage on every launch — a significant caveat for a company whose valuation rests partly on cheap, repeatable access to orbit. TechCrunch

Altman's skepticism is not new. In February, he called Musk's space data center plans "ridiculous" given current AI computing needs, a comment that predates this weekend's exchange by roughly five months. Fox Business

The underlying engineering argument is straightforward enough that most infrastructure specialists have made some version of it already. Orbital data centers face thermal rejection constraints that terrestrial facilities don't: there is no atmosphere to convect heat away, so radiators must handle 100 percent of dissipation, and radiator mass scales unfavorably against the compute density modern AI accelerators demand. Latency to ground-based users, radiation-hardening requirements for GPUs not designed for space, and the total absence of in-orbit maintenance capability all compound the difficulty. None of this is secret; it is the reason the phrase "most experts already believe" attaches so easily to Altman's position, per TechCrunch's own framing of the story.

None of that settles whether the criticism is fair as applied to SpaceX's specific roadmap, and it is worth separating the general skepticism about orbital compute from the personal animus between two men who have sparred publicly for years, dating back to disputes over OpenAI's founding structure and mission. The scammer accusation and the sellling typo make for good social media theater, but they don't constitute technical rebuttal on either side.

What is genuinely interesting here is the valuation mechanics. A two-trillion-dollar SpaceX valuation resting substantially on a product category that has not demonstrated commercial viability — and that SpaceX's own road show materials flagged as dependent on an expendable second stage, undercutting the reusability thesis that has powered SpaceX's cost advantage for a decade — is the kind of gap between narrative and hardware milestone that public-market investors will eventually have to reconcile against actual flight data. Altman's swipe at Musk "selling public market investors" on the concept reads, stripped of the personal edge, as a pointed reference to exactly that gap.

The July 16 Starship flight, if it proceeds on schedule, will not settle the orbital data center debate on its own. Flight thirteen is a vehicle test, not a payload demonstration, and SpaceX has not indicated that inference hardware will fly aboard it. But any anomaly, successful landing, or stage-loss data point will be read by investors and rivals alike as evidence for or against the broader thesis Musk is asking markets to underwrite.

In this author's view, the more durable story here isn't the insults — public feuds between well-capitalized founders are a recurring feature of this industry, not a bug — but the fact that a company's headline valuation now rests so heavily on an unproven orbital compute product rather than its established launch business. That is a genuinely new dynamic worth watching independent of who calls whom a scammer on a given weekend.