HomeFull speed ahead. Keep track of advancements in every frontier
ColossusxAI—GrokxAI—xAI—GrokxAI—ColossusxAI—GrokxAI—ColossusxAI—GrokxAI—ColossusxAI—
Colossus
Courtesy of xAI · Press kit, editorial use (opens x.ai)

Colossus

by xAIUS

DeployStage 4 of 5

Colossus 1 is running. Colossus 2 is being built as a gigawatt-scale cluster.

Updated 5 Jun 2026Checked 25 Sep0 updates this week

Milestones

Next · Colossus 2 gigawatt-scale build
  1. Colossus online with 100,000 H100s after ~122 daysSep 2024Complete.
  2. Expanded to about 200,000 GPUs2025Complete.
  3. Shelby County air permit for gas turbinesJul 2025Complete.
  4. Colossus 2 gigawatt-scale buildNowCurrent milestone.

Most important updates

  • 5 Jun 2026
  • 6 May 2026
  • 10 Mar 2026
  • 17 Jan 2026
  • 2 Sep 2024

Upcoming

  1. Oct 2026Google's full-rate rental of about 110,000 GPUs on xAI clusters starts ($920M a month) (next)

Current obstacles

  • Power and permitsMemphis power has come from on-site turbines and new substations, drawing regulator attention.

Physics limits

  • Almost all power becomes heatNearly every watt a chip draws ends up as heat. A 72-GPU rack draws about 120 to 130 kW, over 10x an old server rack, so air can't cool it; liquid loops must.
  • A gigawatt is a power plant1 GW running all year is about 8.8 TWh, the output of a large nuclear reactor. Sites need dedicated generation or new high-voltage lines, which take years to build.
  • At this scale something always breaksMeta logged 419 unexpected interruptions in 54 days on a 16,384-GPU Llama 3 run, about one every 3 hours. Bigger clusters fail more often and must checkpoint constantly.

How it works

4 parts
An NVIDIA HGX board with eight GPUs under their heatsinks: Colossus's servers are built from boards of this kind
An NVIDIA HGX board with eight GPUs under their heatsinks: Colossus's servers are built from boards of this kindPhoto: Pokiiri · CC BY-SA 4.0 (opens commons.wikimedia.org)
Speed

Reuse an old factory

Colossus 1 went into a former Electrolux plant in Memphis, from empty building to 100,000 H100 GPUs in about 122 days.

Power

Turbines and batteries

Mobile gas turbines supplied power before the full grid hookup; Tesla Megapacks smooth the sudden load swings when training jobs start and stop.

Network

Ethernet at huge scale

GPUs talk over NVIDIA Spectrum-X Ethernet rather than InfiniBand, which let xAI scale one cluster past 100,000 GPUs.

Colossus 2

The next gigawatt

Colossus 2 in Memphis runs GB200 and GB300 racks, fed by a gas plant built across the state line in Southaven, Mississippi.

Update log

6 updates

Fri 5 Jun

  • Major: PressCommercial

Wed 6 May

  • Major: BlogCommercial

Tue 10 Mar

  • Minor: PressRegulatory

Sat 17 Jan

  • Major: XHardware

Wed 2 Jul 2025

  • Minor: PressRegulatory

Mon 2 Sep 2024

  • Major: XHardware

About xAI

The team behind Colossus

xAI

Grok models, Colossus supercomputer

xAI, founded by Elon Musk in 2023, makes the Grok models used on X, which it absorbed in 2025. It built its Colossus supercomputer in Memphis unusually fast. SpaceX bought xAI in February 2026, valuing it at $250B.

  • Bought by SpaceX on 2 February 2026, in a deal valuing SpaceX at $1T and xAI at $250B.
  • Colossus went from empty building to 100,000 H100 GPUs in about four months in 2024.
  • Grok 4 (July 2025) and Grok 4.1 (November 2025) ranked near the top on reasoning tests.
Founded
20233 yrs
Headquarters
United States
Status
Subsidiaryof SpaceX
Valuation
$250BSpaceX-ownedFeb 2026
Raised
$32B3 rounds
Last round
Series E · $20BJan 2026
Works in
AIFrontier modelsComputingSpaceflight
Coverage
2 programs · 18 updateslatest 19d agochecked 25 Sep
Partners
XTesla
Suppliers
NVIDIA
Acquired by
SpaceX
People
Elon MuskFounder
Primer