The agreement, one of the largest of its kind, will see Oracle build and expand hyperscale facilities, including a 1.2 GW “Supercluster” campus in Abilene, Texas, equipped with tens of thousands of Nvidia’s latest GB200 AI chips.
The move, confirmed by multiple reports including the Financial Times, aims to underpin OpenAI’s continued rapid model development while diversifying its cloud infrastructure beyond Microsoft Azure. OpenAI is also working with CoreWeave and Google Cloud as part of this multicloud strategy.
Oracle’s data centre footprint is set to expand dramatically. In total, it plans to develop up to 4.5 GW of AI compute infrastructure across sites in Texas, Michigan, Ohio, Wisconsin, Pennsylvania, New Mexico, Georgia, and Wyoming.
The scale is unprecedented: 4.5 GW is enough electricity to power millions of homes, reflecting the huge energy demands of next-generation AI models.
Market reaction was also swift. Oracle’s share price surged to record highs, with analysts predicting that the deal could more than double the company’s data centre revenues and reshape its position in the AI cloud market.
“This is Oracle’s breakout moment,” said one analyst. “It’s gone from enterprise laggard to an AI infrastructure powerhouse almost overnight.”
The Abilene campus alone is expected to eventually house 400,000 Nvidia GPUs, representing a hardware investment worth nearly $40 billion.
The deal cements Oracle’s position as a key infrastructure provider for OpenAI and reinforces the growing arms race in AI compute.
While Microsoft remains OpenAI’s closest strategic partner, Oracle’s infrastructure will be crucial for the scale and speed of future model training.
RELATED STORIES
AI, APIs and industrial IoT: Oracle’s Andrew De La Torre on the future of telco transformation
Meta’s AI ambitions heat up with OpenAI talent hunt and $15bn scale AI investment
