List of AI Datacenters in USA
Here is a compiled list of some of the most prominent AI-focused data centers (or major clusters/supercomputers) operational or coming online around 2025–2026, based on recent reports as of mid-January 2026. Many details are approximate, planned, or phased, as these massive projects evolve rapidly with expansions, power constraints, and supply chain factors.I've focused on well-documented examples from key players like xAI, Meta, Microsoft/OpenAI (Stargate), and Oracle (Stargate-related). Numbers for GPUs and power are often current or near-term (early 2026) estimates, with totals sometimes higher in future phases.Owner
Start Date (Initial/Operational)
GPU Total (approx.)
Electricity Demand (approx.)
GPU Make & Model
City
State
xAI (Colossus complex)
September 2024 (Phase 1); expansions through 2025–2026
555,000+ (target ~1M across site)
Up to ~2 GW (total site capacity)
Primarily NVIDIA GB200, with H100/H200 mixes
Memphis (multi-building site)
Tennessee
Meta (Prometheus)
Expected 2026
Hundreds of thousands (part of >1.3M fleet)
~1 GW
NVIDIA (various high-end models)
New Albany
Ohio
Meta (Hyperion)
Phased, initial by late 2020s
Massive scale (part of multi-GW plans)
Up to 5 GW (long-term)
NVIDIA (various)
Richland Parish
Louisiana
Microsoft (Fairwater, part of OpenAI support)
Early 2026
Hundreds of thousands
Not fully specified (multi-GW multi-site)
NVIDIA GB200/GB300
Mount Pleasant
Wisconsin
OpenAI/Oracle (Stargate Abilene campus)
September 2025 (initial buildings); scaling 2026
Up to 450,000+
~1.2 GW
NVIDIA GB200
Abilene
Texas
Oracle (Stargate-related)
September 2025 (initial); expansions mid-2026
>450,000 (planned)
1.2 GW+
NVIDIA GB200
Abilene
Texas
Key Notes:xAI's Colossus stands out as one of the fastest-built and largest current clusters, starting small in 2024 but expanding aggressively with on-site gas turbines to bypass grid delays. Meta is pursuing gigawatt-scale "titan" clusters (Prometheus, Hyperion) to support its LLaMA models and broader AI ambitions, aiming for >10 GW total capacity by end-2026 across projects. Stargate (OpenAI/Microsoft/Oracle/SoftBank partnership) is a multi-site, multi-GW initiative targeting up to 10 GW overall, with Abilene as a flagship (often cited for its rapid scale-up to support millions of GPUs in the long run). Electricity demand is a major bottleneck—many sites use a mix of grid, on-site natural gas, renewables, and batteries. GPU counts and power are dynamic; expansions (e.g., to 1M+ GPUs) are common goals for 2026+. Other notable mentions (less detailed public data): Anthropic/Amazon clusters, Google TPU-based facilities, and emerging neoclouds like CoreWeave.
These facilities represent the cutting edge of the AI infrastructure boom, but they also highlight challenges like power availability, environmental impact, and regulatory hurdles in 2026.
References
This article "List of AI Datacenters in USA" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:List of AI Datacenters in USA. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
