“SpaceX stood up a 100,000 GPU supercomputer cluster in 19 days, a process that typically takes a year after equipment delivery.”