On September 25, 2025, according to a report from The Wall Street Journal, leading artificial intelligence company OpenAI officially announced a large-scale global computing infrastructure expansion plan with a total investment expected to reach $1 trillion. This move marks a significant shift in the global AI industry, which is transitioning from the stage of model innovation competition to a new phase of “arms race” in infrastructure construction.
OpenAI has clearly shifted its strategy from the previous “model-driven” approach to “computing power-driven”. The core goal of this globally disclosed computing infrastructure expansion strategy is to meet the massive computing power demands of next-generation artificial intelligence models during training and inference. The plan not only involves the deployment of ultra-large-scale computing center clusters in multiple locations across the United States but also promotes the construction of overseas computing power infrastructure, setting a new record for investment scale in AI industry infrastructure worldwide.
Strategic Partnerships: Securing Hardware and Capital Support
Notably, one day before the announcement of this $1 trillion investment plan, OpenAI had just reached a deep strategic partnership worth $100 billion with chip giant NVIDIA. Industry analysts believe that this collaboration will provide OpenAI with crucial high-performance computing hardware support, effectively resolving its concerns regarding the supply of computing hardware. At the same time, it eliminates market worries about OpenAI’s financial sustainability, laying a solid foundation for the subsequent large-scale infrastructure construction.
In addition, on Tuesday of the same week, OpenAI announced another important cooperation plan—joining hands with Oracle and SoftBank to build five new artificial intelligence data centers in the United States and upgrade existing facilities. According to the cooperation plan, OpenAI will jointly construct 3 data centers with Oracle and 2 with SoftBank, while expanding an existing Oracle data center in Abilene, Texas. This multi-party cooperation not only accelerates the construction progress of computing infrastructure but also enriches OpenAI’s resource channels in terms of technology and operation.
Abilene Computing Center: The Core of Global Layout
As the core project of this expansion plan, the first computing center in Abilene, Texas, USA, has attracted much attention. Located in the grasslands approximately 180 miles west of Dallas, the center covers an area equivalent to that of New York’s Central Park. Initial construction work has been completed, and subsequent projects are progressing as scheduled. OpenAI revealed that the final computing capacity of this completed center will be more than 13 times the scale of the company’s originally deployed computing power.
During a recent media visit event, executives from Oracle and OpenAI jointly led journalists on a tour of this 1,100-acre computing power base and officially positioned it as the “world’s largest artificial intelligence supercomputing center”. On-site observations showed that despite the local construction environment facing high temperatures of 100 degrees Fahrenheit (approximately 37 degrees Celsius), workers still wore dust masks and sunglasses, advancing various construction tasks in an orderly manner, demonstrating the efficient progress of the project construction.
The Wall Street Journal commented on this that OpenAI’s large-scale computing power infrastructure construction initiative aims to build an insurmountable computing power barrier, providing strong underlying support for the research and development of its next-generation models such as the GPT series. This layout not only consolidates OpenAI’s leading position in the global AI field but also is poised to shape the development direction of the entire AI industry’s infrastructure, ultimately emerging as a key driver behind the latest AI trend centered on robust, large-scale computing power foundations.
