OpenAI, the creator of ChatGPT, is accelerating its efforts to dominate the artificial intelligence landscape with a monumental expansion of its infrastructure. In partnership with cloud computing company Oracle and Japanese investment firm SoftBank, OpenAI has announced plans to stand up five new AI data center sites. This move is part of the ambitious 'Stargate' project, an initiative aimed at scaling up the infrastructure needed to train and operate the next generations of AI models.
The new facilities will be located across the United States. Specifically, OpenAI and Oracle will develop sites in Shackelford County, Texas; Doña Ana County, New Mexico; and a yet-to-be-disclosed Midwest location. Concurrently, OpenAI and SoftBank will collaborate on building centers in Lordstown, Ohio, and Milam County, Texas. These new locations are in addition to an existing site in Abilene, Texas, and other projects already under development.
The scale of the Stargate project is immense. The entire project is expected to include 10 gigawatts of computing power and cost $500 billion. With the recent announcements, the project brings its capacity to nearly 7 gigawatts and involves more than $400 billion in investment over the next three years. This puts OpenAI on a path to secure the full $500 billion, 10-gigawatt commitment ahead of schedule, by the end of 2025. OpenAI CEO Sam Altman emphasized the importance of this infrastructure, stating, "AI can only fulfill its promise if we build the compute to power it."
This massive investment highlights the tech industry's incredible thirst for computing power, driven by the advancement of generative AI. Data centers require large expanses of land, vast quantities of water for cooling, and astronomical amounts of electricity. The expansion also offers OpenAI a way to break out from its longtime partnership with Microsoft, which until recently was its exclusive computing partner. Altman mentioned his company has been "severely limited for the value we can offer to people," noting that ChatGPT is slow and not as smart as they would like.
The Stargate project is not just about scaling capacity but also about pushing technological boundaries. In a partnership with Nvidia, OpenAI plans to deploy up to 10 gigawatts of AI systems, which will require between 4 million and 5 million GPUs. This collaboration includes a potential investment of up to $100 billion from Nvidia into OpenAI. Nvidia CEO Jensen Huang described the project as "giant," noting the scale is equivalent to his company's entire annual GPU shipment.
The ultimate goal, as envisioned by Altman, is to unlock future breakthroughs that benefit humanity. "Maybe with 10 gigawatts of compute, AI can figure out how to cure cancer," he said. While the realization of such a massive undertaking faces challenges like financing and permitting, the commitment from OpenAI and its partners signals a new, aggressive phase in the race for artificial intelligence supremacy.