🤖 AI Summary
Big AI datacenter projects are flocking to Texas, driven by cheap, colocated energy and massive GPU demand. Meta announced a new El Paso campus scaling to 1 GW of compute by 2028 (initial investment ≈ $1.5B) with power from El Paso Electric’s mix of natural gas, nuclear and solar and a pledge to match consumption with renewables. Nvidia-backed Poolside and CoreWeave plan a 2 GW “bit barn” on West Texas oil fields using onsite natural-gas generators, with CoreWeave as anchor for the first 250 MW slated for 2027; Poolside will meanwhile rent a 40,000‑GPU GB300 cluster from CoreWeave. Other Texas plays include ECL’s 1 GW Houston site (hydrogen fuel cells, tenant Lambda) and OpenAI’s Stargate campus in Abilene, where Crusoe and Oracle have deployed the first 200 MW of a planned 1.2 GW — Oracle says the full site will house ~450,000 GPUs (≈6,250 NVL72 racks).
These builds illustrate how AI’s insatiable compute needs are reshaping infrastructure strategy: developers prioritize proximity to energy sources to cut transmission costs and accelerate phased, modular construction, while hyperscale GPU density drives huge capital outlays (NVL72 racks are estimated ~$3.5M each). The tradeoffs are stark — faster, cheaper deployment using fossil-fuel generation versus claims of renewable matching and experiments with hydrogen — and they foreshadow major grid impacts, supply‑chain pressure on Nvidia hardware, and continued consolidation of compute capacity in energy-rich regions.
Loading comments...
login to comment
loading comments...
no comments yet