Oracle, OpenAI Scrap Data Center Plan
Oracle and OpenAI have scrapped plans for a massive data center expansion in Texas. The abrupt cancellation is putting pressure on AI-related stocks, raising questions about the sustainability of the sector's rapid infrastructure build-out and capital expenditure.
The now-canceled expansion was slated for a flagship AI data center site in Abilene, Texas, operated by developer Crusoe. This facility is a key part of the ambitious "Stargate" initiative, a multi-hundred-billion-dollar project involving OpenAI, Oracle, and SoftBank to significantly scale up AI infrastructure in the U.S. Talks to add approximately 600 megawatts of capacity stalled due to prolonged negotiations over financing and OpenAI's evolving infrastructure requirements. OpenAI's engineers have reportedly determined that 1-2 gigawatt sites represent a performance "sweet spot" for their training clusters, shifting focus away from massive single-site expansions. Chipmaker Nvidia played a pivotal role in the fallout, paying a $150 million deposit to Crusoe to facilitate discussions for Meta Platforms to potentially lease the expansion space. This move is seen as an effort by Nvidia to ensure its own chips, rather than those of rivals like AMD, would be utilized in the facility. Despite the Abilene cancellation, the broader agreement for Oracle to develop 4.5 gigawatts of data center capacity for OpenAI remains in effect. OpenAI has indicated that the computing capacity from the scrapped expansion will be sourced from other data center campuses currently under development. The project has faced operational challenges, including disruptions to liquid cooling systems during winter weather earlier this year, which caused some data center buildings to go offline for days. This cancellation occurs amidst a massive surge in AI infrastructure spending, with U.S. tech giants expected to spend around $650 billion in 2026. However, a recent survey indicated that only about one-third of AI projects currently deliver a positive return on investment, though a majority of companies plan to increase their AI infrastructure budgets.
