The Physical Stack of AI · Datacenters and gigawatt campuses
You can name the four flagship non-Stargate AI campuses of 2026 — Meta Hyperion, xAI Colossus, AWS Project Rainier, and Stargate UAE — and place each one on the map by operator, location, and approximate GW scale.
Stargate is the largest, but it isn't alone. The 2026 gigawatt landscape includes four other campuses worth knowing by name, each shaped by a different competitor's strategy:
- Meta Hyperion — Richland Parish, Louisiana. Meta's investor release describes an approximately $27B JV with funds managed by Blue Owl; Meta's data-center page says the site will deliver over 2 GW of compute capacity for future open-source LLM training; Entergy's agreement names the supporting generation and 500 kV transmission buildout. - xAI Colossus 2 — Memphis, Tennessee. Reported at 2 GW, 555K NVIDIA GPUs, and roughly $18B of capex by early 2026; also the site of a live air-permit appeal and local public hearing process. It is the speed-prioritized campus in the comparison set. - AWS Project Rainier — Indiana, dedicated to Anthropic. Anthropic says it uses more than 1M Trainium2 chips through Rainier, with nearly 1 GW of Trainium2/Trainium3 capacity coming online by end-2026 and up to 5 GW under the expanded April 2026 deal. - Stargate UAE — Abu Dhabi. 1 GW total, with the first 200 MW expected to go live in 2026. The first international Stargate deployment and sovereign-partnership AI campus, built by G42 and operated by OpenAI and Oracle with NVIDIA, Cisco, and SoftBank in the partner stack.
Chapter contains 4 lessons.