Decomposed dinosaurs make Texas a top destination for AI bit barns

Steven

Decomposed dinosaurs make Texas a top destination for AI bit barns

Summary

Texas is rapidly becoming a major hub for hyperscale AI datacentres as companies chase cheap, abundant power — often from natural gas extracted from local oilfields. Meta has announced a phased El Paso site targeting a gigawatt of compute by 2028. Nvidia-backed Poolside and CoreWeave plan a two-gigawatt West Texas facility powered by onsite natural-gas generators, with CoreWeave as an anchor customer for the first 250MW. Poolside will also rent a 40,000-GPU cluster while the site is being built; the full build is estimated at around $16bn (excluding hardware). Other projects include ECL’s planned one-gigawatt hydrogen-powered datacentre near Houston and OpenAI’s Stargate campus in Abilene (200MW live, expected to reach 1.2GW next year) with Oracle and Crusoe partners. Massive GPU counts and high costs — e.g., NVL72 racks at an estimated $3.5m each — underline the scale and expense of these builds.

Key Points

  • Meta has announced a new El Paso datacentre planned to scale to 1GW of compute capacity by 2028.
  • Poolside and CoreWeave plan a two-gigawatt GPU ‘bit barn’ in West Texas, to be powered by onsite natural-gas generators drawn from local oilfields.
  • Poolside expects to invest roughly $16bn in the project (excluding compute hardware) and will rent 40,000 Nvidia GB300 GPUs from CoreWeave while building.
  • ECL claims a 1GW hydrogen-fuel-cell-powered datacentre planned for Houston, with Lambda as a tenant — but earlier phases have seen delays.
  • OpenAI’s Stargate campus in Abilene has brought 200MW online and aims for 1.2GW; Oracle/partners say the full site could house around 450,000 Nvidia GPUs.
  • The scale drives huge capital and hardware costs (NVL72 racks are estimated at ~$3.5m each), but operators expect volume discounts and modular construction to manage expenses.
  • Energy sourcing varies: some projects rely on fossil fuel generators for cost and proximity, others tout renewable matching or hydrogen as an alternative.

Context and relevance

This wave of projects shows where the AI infrastructure market is heading: extreme scale, proximity to cheap fuel sources, and a mix of energy strategies (fossil gas, hydrogen, renewable matching). For cloud and infrastructure teams, investors, and policymakers, these builds matter because they reshape regional energy demand, carbon profiles, supply-chain dynamics for GPUs, and the geography of compute capacity. The announcements also highlight trade-offs — lower build and operating costs vs environmental and regulatory scrutiny — that will shape planning and procurement decisions across the industry.

Why should I read this?

Because it’s where the money, GPUs and power are heading — and that affects costs, supply and the planet. If you care about who gets access to vast AI compute, how it’s being powered, or what it means for GPU availability and energy grids, this quick read tells you which players are building what in Texas and why it matters.

Author’s take

Punchy and to the point: these are not small experiments — they’re multi‑billion‑dollar bets that will determine where AI runs and who controls the iron. Worth paying attention to if you work in AI infrastructure, procurement, energy policy or enterprise strategy.

Source

Source: https://go.theregister.com/feed/www.theregister.com/2025/10/16/texas_ai_bitbarns/