Could SpaceX Put a Data Center in Orbit? One VC Thinks So.

But wait, there's a catch

Honestly, the most interesting thing Schroepfer said wasn't the SpaceX claim. It was this: ocean data centers are 100x cheaper than orbital ones.

Sit with that for a second. If the cost differential is that enormous, why is anyone talking about space at all?

A few possible answers. One is that "cheaper" doesn't always win if the use case is different enough. Orbital data centers could theoretically offer global coverage without the latency problems of routing through terrestrial networks, or serve applications where physical security and jurisdictional ambiguity actually matter. Another answer is that the cost gap narrows as launch costs keep falling, and Schroepfer (and others) are betting on a trajectory, not today's numbers.

A third answer, tbh, is that some of this is hype. It's too early to say which of those is doing the most work.

The SpaceX IPO angle

Bloomberg's second segment on this ties the conversation to the anticipated SpaceX IPO, framing orbital AI infrastructure as "the next frontier" that tech giants and startups are already betting on.

I initially thought this was mostly financial narrative, the kind of story that gets told to justify valuations. But after reading more carefully, it seems like there's something more substantive underneath. The convergence of falling launch costs, AI's voracious appetite for compute, and the physical limits of terrestrial data center buildout (power, water, land, political opposition) creates at least a plausible case that someone will seriously attempt this within the decade.

Whether "seriously attempt" becomes "successfully operate" is a different question entirely.

What this has to do with embodied AI

You might be wondering why I'm covering this on a beat that's supposed to be about humanoids and embodied AI. Fair.

Here's my thinking. The robots we're watching get deployed right now, the warehouse systems, the humanoid prototypes, the autonomous vehicles, they're all increasingly dependent on cloud inference. The physical robot is almost secondary to the compute infrastructure behind it. Where that compute lives, how it's powered, how it's connected, these aren't abstract questions. They shape what embodied AI can actually do in the field.

If orbital data centers become real, even partially, the implications for latency-sensitive robotic applications are significant. Or they could be a disaster, depending on how the connectivity actually works. We don't know yet.

The honest version of this story is that a credible person with relevant experience made a specific, falsifiable claim about a company's capabilities, in a context where there's real money starting to move. That's worth paying attention to, even if most of the details remain unclear.

I should probably know more about the orbital mechanics and thermal management challenges here than I do. But the infrastructure question feels like it's about to become impossible to ignore, whether the answer ends up being ocean, orbit, or something nobody's named yet.

Aisha Patel · 2 days ago · 8 min