other
GPU reported as the first AI accelerator flown and workload-tested in orbit, by Starcloud, in November of the prior year.
No evidence-backed relationships are recorded.
Moonshot published 2.8 trillion open weights. At four bits per parameter that is about 1.4TB resident before any cache, which rules out the eight-way H100 node most teams assume.
Reality
Two contract labs made and tested Claude's designs, which is genuinely new. The field baseline those results are scored against comes from the company that ran the experiment.
Reality
GPU rental prices fell more than 60% in 2025. SemiAnalysis data says H100s have since climbed about 40% to $2.35 per GPU-hour, with 12 to 18 month waits for clusters.
Reality
The binding constraint on AI capacity is power, water and permits, not silicon. Orbital compute still has to survive the arithmetic: 150kW per satellite against gigawatt-scale ground sites.
Reality
SpaceX, Blue Origin, Google and Starcloud have filings out for well over a million compute satellites. The deadline is a sales pitch; the paperwork is a read on ground power and permitting.
Reality