OpenAI, Oracle abandon 2 GW AI data center expansion in Texas

OpenAI and Oracle have stepped back from plans to expand a major artificial intelligence data center in Texas, according to a report by Bloomberg.

The decision follows extended negotiations over financing and shifting infrastructure needs for AI development.

The project involves the Abilene, Texas site, a key part of the massive Stargate initiative.

The joint venture between OpenAI, Oracle, and SoftBank aims to build large-scale computing infrastructure for advanced artificial intelligence systems.

Originally, OpenAI and Oracle planned to expand the Abilene facility’s capacity from 1.2 gigawatts to 2 gigawatts.

Instead, Meta Platforms is now evaluating whether it should lease the additional planned capacity.

Financing and demand shifts

People familiar with the matter told Bloomberg that financing challenges and changing demand forecasts slowed the expansion plans.

As negotiations dragged on, OpenAI reassessed its near-term infrastructure needs. The companies eventually decided not to move forward with the larger expansion.

Construction continues at the existing site, which covers roughly 1,000 acres in Abilene. Oracle and OpenAI will continue operating within the Stargate facility already under development.

The Stargate initiative itself remains one of the most ambitious AI infrastructure projects announced in the United States. U.S. President Donald Trump introduced the plan in January, stating that the companies involved could invest as much as $500 billion to build AI infrastructure across the country.

In September, OpenAI, Oracle, and SoftBank also revealed plans for five U.S. data centers tied to the Stargate effort.

Three sites would involve Oracle, two would link to SoftBank operations, and one would expand the Abilene complex.

While OpenAI and Oracle pulled back from the expansion, Meta Platforms may take advantage of the available capacity.

According to the report, Meta is exploring a deal to lease the additional infrastructure planned for the site. The expansion is owned and developed by the energy-focused cloud company Crusoe.

NVIDIA also played a role in steering Meta toward the opportunity, Bloomberg reported.

The chip designer supplies many of the advanced processors used to train and run AI models in large data centers.

OpenAI and Oracle already rely on NVIDIA hardware at the Abilene site.

NVIDIA reportedly intervened earlier to ensure that its AI chips power the expanded facility rather than processors from rival Advanced Micro Devices.

Meanwhile, Meta continues to expand its own AI computing footprint aggressively.

Last month, the company signed a multi-year agreement with AMD to deploy as much as 6 gigawatts of AMD Instinct AI chips.

The deal aims to support Meta’s next generation of AI infrastructure and model development.

Stargate project moves forward

Despite the canceled expansion plan, companies involved say progress at the Abilene facility continues.

“We are very proud of our relationship and our progress in bringing capacity online,” Oracle said in a statement, according to the report.

The companies building the site also emphasized the scale and speed of the project.

“Together, Crusoe and Oracle are operating in lockstep to deliver one of the world’s largest AI factories in Abilene,” Crusoe added.

“Our collaboration can deliver massive-scale infrastructure faster than anyone else in the industry.”

OpenAI and Oracle also continue to pursue additional data center capacity elsewhere.

In July, the companies announced plans to build another 4.5 gigawatts of AI data center infrastructure, and that project remains on track.

The shifting plans highlight the rapidly evolving demand for computing power as major technology firms race to build the infrastructure needed for advanced artificial intelligence.