"Stargate" AI infrastructure project sets sail! Texas' first data center plans to deploy 64,000 NVIDIA Corporation (NVDA.US) GB200 chips.

OpenAI and Oracle Corporation (OCRL.US) plan to deploy tens of thousands of high-performance AI chips from NVIDIA Corporation (NVDA.US) in a new large data center to be built in the state of Texas in the coming months. This is part of their $100 billion "Stargate" infrastructure project. According to sources, the data center in the small town of Abilene is expected to be equipped with 64,000 NVIDIA Corporation's popular GB200 chips by the end of 2026. These chips will be installed in multiple rooms of the data center in phases, with the first batch of 16,000 chips expected to be deployed this summer. For a single client's data center in the initial stages, the planned volume of these chips represents a huge computational capability. This also highlights the potential scale of the Stargate joint venture project, announced by OpenAI, SoftBank (SFTBY.US), and Oracle Corporation at an event in the White House in January. OpenAI previously stated that Stargate will expand to up to ten sites. A spokesperson for OpenAI stated that the company is working with Oracle Corporation to design and deliver the data center in Abilene, with Oracle Corporation responsible for acquiring and operating the supercomputers being built there. Oracle Corporation did not respond to requests for comment, and NVIDIA Corporation declined to comment. The launch of the Stargate project indicates that tech giants are racing to deploy NVIDIA Corporation's latest chips, primarily used for training and deploying generative AI models. Musk's xAI recently signed a $5 billion agreement with Dell Technologies, Inc. Class C Tech (DELL.US) to build an AI supercomputing center in Memphis. Meta (META.US) stated its plan to have computing power equivalent to 600,000 NVIDIA Corporation H100 (the company's previous generation data center semiconductor) chips by the end of 2024. AI-focused cloud service provider CoreWeave stated in its IPO filing earlier this month that it has over 250,000 NVIDIA Corporation GPUs in 32 data centers. It is estimated that the number of GB200 chips required for just the first Stargate facility will cost billions of dollars. While the price of the GB200 chip from NVIDIA Corporation has not been publicly disclosed, the company's CEO, Jensen Huang, stated last year that the slightly less powerful B200 chips sell for between $30,000 and $40,000 each. In addition to the data center in Texas, the team from OpenAI and SoftBank also explored locations in Pennsylvania, Wisconsin, and Oregon as potential sites for future Stargate data center campuses. Salt Lake City is also reportedly a candidate, as Oracle Corporation has already established a presence there in cloud computing.