Financing challenges and shifting AI compute demands have derailed Oracle and OpenAI's planned expansion of their flagship Stargate data center in Abilene, Texas, creating an opening for Meta to step in with Nvidia's help. The companies scrapped plans to lease additional capacity at the Crusoe-developed campus after negotiations dragged on for months.
While the 1,000-acre site continues construction with several operational sections already running, Oracle and OpenAI elected not to proceed with what would have been a major expansion.
Meta Platforms is now considering leasing the planned expansion space from developer Crusoe, according to people familiar with the discussions. Nvidia has taken an active role in brokering the potential deal, paying a $150 million deposit to Crusoe to secure future capacity before approaching Meta about occupying the space.
The move reflects Nvidia's strategic interest in ensuring its AI semiconductors fill new data center capacity rather than those of rival chipmakers like Advanced Micro Devices. The GPU maker became involved when it became clear Oracle and OpenAI wouldn't expand their Texas footprint.
"Our flagship Stargate site is one of the largest AI data center campuses in the United States," OpenAI compute scaling executive Sachin Katti said in a statement. "Today we have more than half a dozen sites under development across multiple states."
The original Stargate project was announced last year as part of a $500 billion joint venture between OpenAI and Oracle. Despite canceling the Abilene expansion, that broader agreement remains intact, Oracle still plans to deploy 4.5 gigawatts of compute capacity for OpenAI across multiple locations.
OpenAI has around $600 billion in projected compute spend by 2030, according to reports, though this is an apparent reduction from earlier projections of $1.4 trillion by 2033. The company reportedly wants to wait for power infrastructure at alternative sites that will support Nvidia's upcoming Vera Rubin chips rather than deploying current Blackwell GPUs at Abilene.
Meta has been aggressively expanding its AI infrastructure spending, projecting capital expenditures of up to $135 billion this year alone. The social media giant announced during its January earnings call that it would invest heavily in GPU compute capacity.
Oracle faces its own financing challenges despite securing what it values as a $300 billion contract with OpenAI over its lifetime. The cloud provider recently announced plans to raise an additional $50 billion in debt and equity to fund its data center buildout while reportedly planning layoffs to control costs.
