Colossus 2
xAI’s second Memphis supercluster, on GB200 and GB300 GPUs. Its first cluster came online in October 2025, per SpaceX’s S-1.
Colossus 2 is an operational GPU cluster run by xAI (SpaceXAI) in Memphis, Tennessee, United States. 440,000 NVIDIA GB200 and GB300 GPUs are planned; no installed count is published. Live power: 946 MW (Epoch AI estimate), 1,500 MW planned (Epoch AI estimate).
Facts
Source per value- Owner / operator Organization that owns or operates the GPUs. Hosting and facility roles are listed under Partners.
- xAI (SpaceXAI) 1
- GPU maker NVIDIA, AMD or Intel, taken from the GPU model.
- NVIDIA 1
- GPU model Accelerator model or family as disclosed. A GH200 is one GPU on a superchip; an MI300A is an APU that contains a GPU.
- NVIDIA GB200 and GB300 1
- GPUs installed Physical GPUs in the cluster, from the operator or a deterministic sum of disclosed quantities (nodes × GPUs per node). Planned counts, H100-equivalents and unaudited press totals are never counted here.
- Not published
- GPUs planned Announced, contracted or target GPU count. Not evidence that the GPUs are installed.
- 440,000 3
- GPU count notes What the count covers (one partition, one site, a design limit), qualifiers and competing figures.
- SpaceX’s S-1 (p. 76, as quoted by Epoch AI): about 110,000 GB200 in the first cluster (Oct 2025), 110,000 GB300 in the second (Apr 2026) and at least 220,000 more GB300 in the next expansion, 440,000 in all, counted here as planned. Press reports 550,000 (Tom’s Hardware, 25 Sep 2026); no filing confirms that total, so the installed column stays empty. 1
- Location City, region and country where the cluster sits. Empty when the operator does not name the site.
- Memphis, Tennessee, United States 1
- Country Country of the cluster, taken from the location.
- United States 1
- Live power MW Power in service, in megawatts of IT load where the source says so. Company figures are marked in the notes; Epoch AI satellite estimates are labelled as estimates.
- 946 3
- Planned power MW Planned or contracted power at full build, in megawatts.
- 1,500 3
- Power notes Basis and date of the power figures, and who published them.
- Epoch AI estimate from satellite imagery, not a company figure: 946 MW IT as of 15 Jun 2026 and about 1.5 GW IT by early 2027. The S-1 gives about 210 MW and 220 MW of GPU server power for the first two clusters, plus over 400 MW for the expansion. 3
- Status operational = running and named as such by the operator or a primary source; partly operational = a phase is live and more is coming; under construction = building underway; announced = disclosed or contracted with no confirmed go-live.
- operational 1
- Operational since Month or year the source gives for the cluster going live (first phase for partly operational clusters). Empty when no date is published.
- 2025-10 1
- Status notes The source’s wording on status, with the date.
- First cluster running since October 2025 per the S-1 timeline. The 440,000 figure is from the filing as quoted by Epoch AI; the 550,000 total is a press figure. 1
Sources
- tomshardware.com accessed 2026-11-11
- SpaceX S-1 (SEC) accessed 2026-11-11
- Epoch AI estimate: Colossus 2 (estimates; quotes SpaceX S-1 p. 76) accessed 2026-11-11
Related clusters
All 87 clustersLists
Need GPU capacity at this scale? Request quotes from GPU cloud providers
Tell us what you need once. We send it to matching GPU cloud providers and they reply to you directly.