Orbital Data Centres: The Economics of AI Compute in Space
New! Listen to Concept to Cloud - Real stories from the trenches of software engineering
Orbital Data Centres: SpaceX's Big Bet Still Has an Economics Problem
Cloud Architecture

Orbital Data Centres: SpaceX's Big Bet Still Has an Economics Problem

TB
Tom Barber
July 13, 2026
0 min read

This week's AI Briefing: an Altman, Musk exchange about SpaceX being worth more than the planet turned into the more interesting question underneath it, can you actually run AI inference in orbit, and does the maths close? Not yet, and here's why.

Happy Monday. This one started life as one of my AI Briefings, the short, near-daily look at what’s actually moving in AI, recorded outdoors in a rare bit of Washington DC sun, squirrels fighting behind me and all. The hook was a weekend spat between Sam Altman and Elon Musk, but the interesting part wasn’t the sniping. It was the thing they were sniping about.

The exchange, and the claim underneath it

The two of them went back and forth a few times, and somewhere in the middle Musk floated the idea that SpaceX will be worth more than the entire planet if it hits all its goals. That seems unlikely on its face. But the crux of the argument is worth taking seriously, because it’s the same bet SpaceX is asking investors to make: space data centres.

That’s one of the headline selling points behind SpaceX’s multi-trillion-dollar raise, the same buildout I dug into when the SpaceX, xAI merger landed. The valuation increasingly hinges on SpaceX’s ability to commoditise hardware in space: not just cheap hardware, but cheap hardware in orbit, so it can launch a fleet of orbital data centres that run AI inference on solar power, off-planet, rather than fighting for grid capacity down here.

The pitch is plausible. The economics are the hard part.

From Musk’s side the logic is clean: provide on-orbit compute, powered by abundant solar, away from terrestrial energy constraints. And clearly you can do it, putting compute in space is an engineering problem, not a physics impossibility. The question is whether you can make it cost-effective and sustainable at scale.

To get there, you have to launch rockets, and you have to launch them at a volume large enough to put a meaningful amount of compute into orbit. That’s where it wobbles:

  • Starship isn’t landing and re-flying reliably yet. The latest Starship vehicles haven’t hit the repeat cadence the model depends on.
  • Full reusability may not fully arrive. Even during the raise, SpaceX acknowledged there will likely still be elements, the second stage, for now, that get thrown away rather than recovered. An expendable component on every flight is exactly the kind of line item that breaks the maths.

Put those together and it’s genuinely hard to build an economically viable, repeatable way of sending silicon into orbit until the reusability story is more finished than it is today.

None of that is to say it won’t happen. I know enough about SpaceX to bet that, at some point, a good chunk of this gets solved. The open question is how soon, and that timing is everything.

The clock that actually matters

The reason timing matters is what’s happening on the ground while we wait. Energy prices are rising and there’s a real squeeze on consumer silicon, because AI data centres are already outcompeting everyone else, both on electricity and on access to the actual chips. That’s a slightly different conversation, and one I suspect we’ll keep having: how do you offset the enormous compute and energy loads AI now demands against what we can realistically supply as a planet? Right now those two curves aren’t aligned.

Orbital data centres are one answer to that mismatch. They’re just not a near-term answer yet, and anyone underwriting them at today’s valuations is betting on a reusability and launch-cadence milestone that hasn’t landed. Worth watching closely, worth being honest about the timeline.

That’s enough from me. I’ll be back tomorrow with more AI news and gossip, thanks for reading the briefing.

TB
Written by Tom Barber

Ex-NASA engineer and cloud architect with over a decade of experience building scalable systems for startups and enterprises.

Work with Tom →

Related Articles

Cloud Architecture

SpaceX Acquires xAI: What Orbital Data Centers Mean for Cloud Architecture

Musk's $1.25 trillion merger isn't just corporate consolidation, it's a fundamental bet that terrestrial infrastructure can't scale to meet AI demand. For systems architects, the implications are significant.

Read More →
AI Engineering

How Do I Get Better LLM Uptime? A Reliability Playbook for AI in Production

Most LLM outages are not caused by the model. They are caused by the pipeline around the model. The retries, the routing, the fallbacks, the rate limits, the observability. A concrete playbook for hitting three-nines on an LLM-powered product.

Read More →
Cloud Infrastructure

SpaceX Orbital Data Centers with Starlink V3: 2026 Status & Outlook

SpaceX is turning Starlink V3 satellites into the backbone of orbital data centers, terabit-class capacity per satellite, an FCC filing for one million units, and a $1.25 trillion SpaceX-xAI merger funding the buildout.

Read More →

Ready to Build Your Product?

Let's discuss how we can help you bring your vision to life with expert cloud solutions

Get Started