Oracle and OpenAI Scale Back Mega Stargate US Data Center

Oracle and OpenAI are recalibrating their plans for the much-discussed Stargate AI data center in the United States. Sources indicate the project will be considerably smaller than initially proposed, with the Abilene, Texas site expected to top out at roughly 1.2GW of capacity.

The downscaling appears to be a strategic pivot rather than an outright cancellation. By limiting initial capacity, OpenAI can concentrate on deploying the latest accelerator chips across more modular campuses rather than committing to a single, massive build. That flexibility could speed up rollouts and help the companies adapt to fast-changing AI hardware developments.

For Oracle, the move reduces near-term capital exposure while still keeping the partnership alive. The smaller footprint still delivers a meaningful data-center presence in a region attractive for power and land, but it is less of a headline-grabbing, monolithic facility than some earlier reports suggested.

Industry observers say the shift reflects broader trends in AI infrastructure planning: organizations are prioritizing agility and upgradeability over scale for scale’s sake. Newer chips and more efficient designs can deliver higher performance-per-megawatt, meaning a distributed approach could be more cost-effective and future-proof.

Local officials in Abilene have been engaged in the discussions, with the expected capacity still significant for the area’s economy. While the exact timeline and investment figures remain unclear, the updated plan signals that both Oracle and OpenAI are taking a measured path forward, balancing ambition with operational and technological realities.