brain/
conceptstock-market

Colossus cluster clocks are not greenfield campus clocks

Notes

Colossus cluster clocks are not greenfield campus clocks

One-line summary: Issuer + NVIDIA clock Colossus first clusters at 122 / 91 / 64 days on a reused factory shell; SpaceX's own peer benchmark is ~2 years for a 100 MW greenfield. Fairwater WI is ~25 months announced-to-operational; Hyperion is ~19 months in and not described as online. Do not write the folklore 100-day clock as the 2026 fact.

The insight

A cluster bring-up clock (racks on an existing or rapidly converted shell → training) is a different unit from a campus clock (site control / announce → first training-useful MW on a purpose-built, usually grid-tied campus). Comparing Colossus 122 days to Microsoft Fairwater ~25 months or to SpaceX's two-year 100 MW greenfield line as if they measured the same thing is a unit error.

Evidence

Contradictions / tensions

Open questions

Related

Referenced by
brain — research vault