When Sam Altman took a swipe at space-based data centers, it read like a bold provocation. But experts in the field largely say he was just stating the obvious.

The OpenAI CEO posted a dismissive comment about companies pitching orbital computing infrastructure to public market investors, writing: "homeboy you're the one sellling [sic] public market investors on short-term space datacenters."

Why Space Data Centers Don't Add Up — Yet

The core problems with space-based data centers are well-documented among infrastructure analysts:

  • Thermal dissipation is dramatically harder in the vacuum of space — heat can only escape via radiation, not convection
  • Launch costs, even post-SpaceX, remain prohibitively high for the hardware density data centers require
  • Latency to ground-based users undermines most real-world compute use cases
  • Maintenance is effectively impossible at scale, meaning hardware failures are permanent

These aren't fringe concerns. They represent the mainstream engineering position on orbital compute.

Hype Outpacing Reality

Despite the skepticism, a handful of startups have attracted significant investor attention by pitching space infrastructure as the next frontier for AI workloads. The argument typically centers on unlimited solar energy and the ability to site infrastructure outside terrestrial regulatory constraints.

But analysts point out that solar collection efficiency and power-to-compute ratios still make Earth-based facilities — especially those near renewable energy sources — far more economical.

"The physics just don't favor it at current technology readiness levels," one infrastructure researcher noted. "You'd need dramatic advances in both launch economics and in-space manufacturing before this pencils out."

Altman's Real Target

What Altman appeared to be calling out wasn't the long-term scientific possibility of space compute, but the short-term investor pitches framing it as an imminent, viable alternative to terrestrial data centers.

That distinction matters. Several companies have raised capital on timelines that most engineers consider wildly optimistic, and Altman's comment — however casually delivered — reflects a broadly held industry view that the hype is running well ahead of the hardware.

For now, the hyperscale data center buildout on Earth continues to accelerate, with AI infrastructure spending projected in the hundreds of billions over the next several years. Space may eventually play a role, but most experts agree that day is measured in decades, not quarters.