Orbital compute is no longer just a speculative idea. It now appears in filings, agency studies, and prototype programs. Some use cases already make sense, especially when data is created in space and bandwidth to Earth is limited. But the broader vision of moving cloud or AI infrastructure into orbit still runs into stubborn constraints in power, heat rejection, launch, replacement, and debris.

Most debates about AI infrastructure start with the wrong metric. Energy and water use describe scale, but they do not tell us what the system produces, what it replaces, or whether the trade is worth it. The harder standard looks at output per unit of energy, how that output is measured, and whether the burden is governed or displaced.

I’ve been joking that I promoted ChatGPT to be my primary care provider. It’s only half a joke. My doctors still have a say in my care, but ChatGPT has become the place where I lay the record out, pressure-test interpretations, and figure out what I actually need to ask next.