Stargate Project Under Threat: Why OpenAI Is Not Interested in Expanding the Texas Data Center
13:26, 10.03.2026
OpenAI, along with Oracle Corporation, is stepping back from scaling one of the world's most ambitious data centers. The project in Abilene, Texas, which was intended to be a part of the massive Stargate initiative, has faced financial and logistical difficulties.
Why the 2 GW Negotiations Fell Through
The original plan implied increasing the facility's capacity from 1.2 GW to 2 GW. However, agreements between OpenAI, Oracle, and the contractor Crusoe fell through due to complex financing conditions and unstable growth forecasts from OpenAI itself.
The situation was further complicated by opposition from local authorities and technical incidents: earlier this year, winter weather disabled the cooling system, paralyzing the functioning of several buildings for several days.
A capacity of 1 GW is equivalent to the output of a modern nuclear plant. While ideas for building nuclear power plants for data center needs are being discussed in the industry, Texas residents are not fond of such plans. People fear that the region's environment may suffer to support AI infrastructure.
NVIDIA’s Intervention
The conflict between Oracle and Crusoe became so serious that NVIDIA Corporation intervened. The chipmaker sought to ensure the steady use of its own hardware and prevent its replacement by products from competitor AMD.
To stabilize the situation, NVIDIA provided Crusoe with a $150 million deposit and attempted to bring in Meta as a potential tenant for the additional capacity that OpenAI declined.
Despite these changes, the facility that occupies over 370,000 square meters continues to develop, and Oracle is rapidly deploying NVIDIA-based servers for current AI needs.
Why Scaling Data Centers Can Be Risky
Scaling AI data centers brings energy and resource challenges. According to forecasts, by 2030, U.S. data centers will consume as much water as all of New York City. By 2026, AI infrastructure will claim up to 70% of the world's memory resources, which could worsen component shortages for household appliances and consumer devices.