To launch the first facility for their $100 billion Stargate infrastructure project, OpenAI and Oracle Corp. intend to start installing tens of thousands of potent AI chips from Nvidia Corp. in a huge new data center in Texas in the upcoming months.

According to a person familiar with the situation, 64,000 of Nvidia's highly sought-after GB200 semiconductors are anticipated to be housed at the location in the tiny city of Abilene by the end of 2026. The person, who spoke on the condition of anonymity to discuss internal preparations, stated that the chips will be deployed to various data centre halls in phases, with an initial rollout of 16,000 scheduled to be finished by this summer.

For just the first stages of a single customer's data centre, the total anticipated shipments amount to an enormous amount of processing power. Additionally, it highlights the potential scope of the Stargate joint venture, which was introduced at a White House event in January by OpenAI, SoftBank Group Corp., and Oracle. In the past, OpenAI stated that Stargate would grow to up to ten locations.

According to a representative for OpenAI, the startup is collaborating with Oracle on the design and construction of the Abilene data center, while Oracle is responsible for acquiring and running the supercomputer that is being constructed there.

Leading tech businesses are competing with Stargate to increase the capacity of Nvidia's latest chips, which are mostly used for generative AI model training and deployment. Recently, Dell Technologies Inc. and Elon Musk's xAI signed a $5 billion agreement for AI servers for a supercomputer in Memphis. By the end of 2024, Meta Platforms Inc. aims to have 600,000 Nvidia H100s, a previous generation of the company's data center chips, in terms of processing power.

CoreWeave Inc., a cloud provider with an AI focus, has more than 250,000 Nvidia graphics processing units spread over 32 data centres.