Sam Altman and experts pump brakes on Musk's space data centers

Sam Altman and Elon Musk just got into another public spat over the weekend, and this time, the target was space data centers.

After Musk took a shot at Altman by calling him a scammer, the OpenAI CEO fired back with some pointed trash talk: "homeboy you're the one sellling public market investors on short-term space datacenters."

Altman's comment hit a raw nerve because it says what most aerospace and AI engineers have been quietly calculating for months: orbiting AI clusters are still a distant dream.

The two-trillion-dollar orbit

SpaceX's orbital data center ambitions have become a massive pillar behind the company's staggering 2-trillion-dollar valuation. The pitch to investors is irresistible: launch fleets of high-powered compute satellites into orbit to run AI inference tasks, fueling Musk's SpaceXAI models and serving as an unprecedented orbital cloud that neatly sidesteps Earth's power constraints.

When Musk dismissed critics by declaring, "We start flying them next year," he was technically right and practically misleading. SpaceX can certainly shoot a satellite packed with AI chips into low Earth orbit next year. But running an occasional experiment is light-years away from building an economically viable fleet.

Engineers at rival space-compute startups, Google's orbital compute division, and independent rocket scientists keep reaching the exact same barrier. Until launch costs drop into the basement and high-powered, heat-dissipating satellites can be mass-produced dirt-cheap, the business case simply will not close.

Starship's dirty little secret

Musk's entire timeline leans on Starship, which is slated for its 13th test flight as soon as July 16. Even if that launch manages the dream scenario of successfully recovering both stages, commercial operations remain years away. Worse yet, orbital data centers are stuck at the back of the line behind lucrative NASA exploration milestones and SpaceX's own Starlink deployment.

More tellingly, SpaceX already let the cat out of the bag during its recent IPO road show. Presenters admitted to prospective investors that Starship might not achieve full reusability in the near term, meaning flights will likely have to expend their second stages.

Throwing away a Starship upper stage every single flight completely demolishes the economic rationale for space computing. Launching massive server clusters cost-effectively isn't a problem for next year; realistically, it's a puzzle for the 2030s.

Why it matters

Altman's jab landed because it called out a glaring mismatch between tech marketing and launchpad reality. For developers, orbital compute is nowhere near becoming a usable alternative to terrestrial GPUs, and latency, bandwidth limits, and replacement costs mean nobody should be waiting on space clouds anytime soon.

Inside the industry, this squabble reveals the sheer pressure behind mega-cap valuations. To justify a 2-trillion-dollar price tag, SpaceX needs investors to view Starship as more than a rocket: it must be a floating supercomputer platform.

We think Musk's strategy of promising flights next year is smart marketing to keep public excitement peaked, but promising near-term commercial viability while secretly expending rocket stages looks untenable. Escaping Earth's power grid sounds poetic, but gravity and economics always collect their due.