xAI Colossus
100,000 GPUs in 122 days — the build that reset the industry's clock.
You're looking at a data-center record: a real facility — who runs it, what hardware fills it, and where its power comes from.
xAI converted a former Electrolux appliance factory in Memphis into the world's most famous AI supercomputer: roughly 100,000 H100-class GPUs brought online in 122 days in 2024, doubled to ~200,000 within months, running on Spectrum-X Ethernet with Supermicro and Dell systems and Tesla Megapacks buffering power. Dozens of on-site gas turbines bridged the grid gap — and drew air-permit challenges from local groups. Grok 3 and 4 were trained here.
Specifications & key facts
Each figure is labeled by basis: vendor spec, vendor claim, reported, estimate, or official. "Not publicly disclosed" means exactly that — we never invent values.
Overview
Scale
Power
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Cooling
Hardware deployed
Operated by
Built by
Related
Sources
NVIDIA on Colossus/Spectrum-X ↗
Record reviewed as of 2026-01-15 — see Methodology.