OpenAI, led by CEO Sam Altman, is spearheading a transformative $500 billion project named Stargate that aims to solidify the United States’ position as a global AI powerhouse.
The initiative involves strategic partnerships with major tech players like Oracle and Microsoft and is expected to generate over 100,000 new jobs in the US, according to a report by Forbes.
At the heart of Stargate is the creation of advanced AI infrastructure, including massive data centers to support the training and deployment of OpenAI’s next-generation models. One of the key sites will be in Abilene, Texas, where the company plans to install 100,000 Nvidia GPUs by the end of 2025. The facility will largely rely on West Texas' wind energy, which already produces more electricity than local needs on most days.
To power this massive AI effort sustainably, OpenAI has joined hands with Crusoe Energy - a company known for turning waste gas from oil wells into usable electricity. Instead of flaring natural gas, Crusoe captures and converts it using specialised engines, significantly reducing pollution. The company’s co-founder, Chase Lochmiller, describes the project as a “factory of factories,” where AI systems will eventually be able to design and build even more sophisticated AI technologies.
Previously, Crusoe had used this clean energy approach to mine Bitcoin. Now, their focus has shifted toward supporting the immense energy needs of AI computing.
The Stargate project is being developed with backing from a consortium that includes SoftBank, Oracle, OpenAI, and MGX. Technical collaborators also include Nvidia, Microsoft, Arm, and others. Though the full $500 billion has not yet been deployed, $100 billion has already been allocated to constructing two state-of-the-art data centers.
Stargate’s urgency is amplified by rising competition from China, particularly with the rapid growth of AI firm DeepSeek, which has started offering high-performance AI models at lower costs and with fewer complications. In response, OpenAI has ramped up innovation, launching GPT-4o - featuring image generation - and making new models like GPT-4.1, GPT-mini, and GPT-nano available via API.
Altman highlighted that these new models excel in long-context tasks (up to 1 million tokens), instruction following, and coding.