OpenAI and Oracle’s Ambitious Stargate Data Center Project

(Bloomberg) — OpenAI and Oracle Corp. are gearing up to fill a massive new data center in Texas with tens of thousands of powerful AI chips from Nvidia Corp. in the coming months. This initiative is part of their ambitious plan to launch the first facility for their $100 billion Stargate infrastructure venture. Located in the small city of Abilene, the site is projected to house 64,000 of Nvidia’s highly sought-after GB200 semiconductors by the end of 2026, according to a source familiar with the plans.

Phased Rollout of Nvidia Chips

The deployment of these chips will occur in phases across several halls of the data center, with an initial rollout of 16,000 chips expected to be completed by this summer. This substantial initial shipment represents a significant amount of computing power for just one data center catering to a single customer. It also highlights the immense potential scale of the Stargate joint venture, which was introduced by OpenAI, SoftBank Group Corp., and Oracle during a White House event in January.

Collaboration and Responsibilities

An OpenAI spokesperson confirmed that the startup is collaborating with Oracle on the design and delivery of the Abilene data center. Oracle is tasked with acquiring and operating the supercomputer being constructed at the site. However, Oracle has not responded to requests for further comments, and Nvidia has also declined to provide additional information.

Stargate’s Role in the AI Chip Race

The Stargate project is entering a competitive landscape where leading tech companies are racing to enhance their capacity for Nvidia’s latest chips, which are primarily utilized for training and deploying generative AI models. For instance, Elon Musk’s xAI recently secured a $5 billion deal with Dell Technologies Inc. for AI servers intended for a supercomputer in Memphis. Similarly, Meta Platforms Inc. has announced plans to achieve computing power equivalent to 600,000 Nvidia H100s—an earlier generation of the company’s data center semiconductors—by the end of 2024.

CoreWeave’s Expansion

Additionally, CoreWeave Inc., an AI-focused cloud provider, has reported having over 250,000 Nvidia graphics processing units across 32 data centers, as detailed in its paperwork for a public offering earlier this month. The anticipated number of GB200s designated for the first Stargate facility is expected to cost billions of dollars. Although Nvidia has not publicly disclosed a price for the GB200, CEO Jensen Huang indicated last year that its less powerful B200 chip is priced between $30,000 and $40,000 each.

Future Stargate Locations

Beyond the Texas site, staff from OpenAI and SoftBank have been exploring potential locations in Pennsylvania, Wisconsin, and Oregon for future Stargate data center campuses. Salt Lake City, where Oracle already has some operational cloud-computing capacity, is also being considered as a viable option, according to the insider.

As the demand for AI capabilities continues to surge, the Stargate project represents a significant investment in the future of artificial intelligence infrastructure. With OpenAI and Oracle at the helm, the initiative is poised to make a substantial impact on the tech landscape.

For more details, visit the original article.