SpaceX is putting data centers in orbit

SpaceX has launched a satellite carrying 200 Nvidia H200 GPUs — the same kind of chips powering today's big AI models. The trick is that the satellite processes data right there in space rather than sending raw data back to Earth. It sends compressed results down instead. The goal is to cut the cost and latency of running AI workloads, especially for companies that need to process lots of data without bottlenecks.

This is the same move Google, Microsoft, and Amazon are making, just at scale and with SpaceX's launch advantage. The real question is whether satellite-based compute stays locked inside SpaceX's ecosystem or opens up to the broader market. If it opens up, the cost of AI could drop in ways that ripple down to the apps we actually use — the ones that run on the backend of our phones, our payments, our healthcare portals.

The tech press is calling this "orbital data centers" and writing about it like it's a new category. It's not. It's just compute, where the hardware lives is less important than whether it's actually cheaper and faster. The people who'll benefit most are the ones running AI services at scale, not the ones buying the latest GPU.

Why this matters for us: cheap compute in orbit will eventually make the apps we rely on — payments, healthcare, transit — cheaper and more reliable, and the ones who control that infrastructure will set the price.