Why does OpenAI need six giant data centers?

In a significant development for AI infrastructure, OpenAI, in collaboration with Oracle and SoftBank, has unveiled plans to establish five new data centers across the United States. This ambitious initiative is part of their joint project, Stargate, which aims to enhance AI capabilities and support the growing demand from ChatGPT's impressive user base of 700 million each week. The expansion will contribute to a staggering 7 gigawatts of planned capacity and is projected to involve over $400 billion in investments over the next three years. Despite the scale of this undertaking, some critics have raised concerns about the sustainability of the investment model. The newly announced sites will include three locations resulting from the partnership between OpenAI and Oracle: Shackelford County in Texas, Doña Ana County in New Mexico, and an undisclosed location in the Midwest. These facilities, along with a 600-megawatt expansion near the existing Stargate hub in Abilene, Texas, are set to provide more than 5.5 gigawatts of capacity, enabling the data centers to consume up to 5.5 billion watts of electricity at peak operation. Moreover, this expansion is expected to generate over 25,000 jobs on-site, bolstering local economies. SoftBank is also contributing to the initiative with two additional sites. One, located in Lordstown, Ohio, is already under construction and is anticipated to begin operations next year. The second site, in Milam County, Texas, will be developed in partnership with SB Energy, a subsidiary of SoftBank Group, and is projected to reach a capacity of 1.5 gigawatts within the next 18 months. These new data centers will complement the flagship Stargate campus in Abilene, where Oracle has begun supplying Nvidia hardware. OpenAI has already started utilizing this facility for training new models and running inference operations for ChatGPT, marking a pivotal moment in the evolution of AI infrastructure.