OpenAI and Oracle (OCRL) are set to deploy tens of thousands of NVIDIA (NVDA, Financial) high-performance AI chips in a new large-scale data center in Texas. This initiative is part of their $100 billion "Stargate" infrastructure project. The data center, located in Abilene, is expected to house 64,000 of NVIDIA's sought-after GB200 chips by the end of 2026. The chips will be installed in phases, with the first batch of 16,000 expected to be operational by summer. This represents significant computing power for a single customer's data center, highlighting the potential scale of the Stargate joint venture, which was announced by OpenAI, SoftBank (SFTBY), and Oracle at a White House event.
OpenAI is collaborating with Oracle to design and deliver the Abilene data center, with Oracle responsible for acquiring and operating the supercomputers being built there. NVIDIA declined to comment on the project. The launch of the Stargate project underscores the tech giants' race to deploy NVIDIA's latest chips, primarily used for training and deploying generative AI models. Elon Musk's xAI recently signed a $5 billion deal with Dell Technologies (DELL) to create an AI supercomputing center in Memphis. Meta (META) plans to have computing power equivalent to 600,000 NVIDIA H100 chips by the end of 2024. AI-focused cloud service provider CoreWeave reported in its IPO filing that it operates over 250,000 NVIDIA GPUs across 32 data centers.
The GB200 chips required for the first Stargate facility alone are estimated to cost billions of dollars. Although NVIDIA has not disclosed the price of the GB200, CEO Jensen Huang previously stated that the less powerful B200 chips are priced between $30,000 and $40,000 each. Beyond the Texas data center, OpenAI and SoftBank are exploring locations in Pennsylvania, Wisconsin, and Oregon for potential future Stargate data center campuses. Salt Lake City is also a candidate, where Oracle has already established cloud computing operations.
