Oracle Invests $40 Billion in NVIDIA Chips for OpenAI's Stargate Data Center

Oracle Corporation has announced a substantial investment of approximately 40 billion dollars to purchase high-performance chips from NVIDIA. This investment is aimed at supporting the new data center being constructed by OpenAI in the United States, known as the "Stargate" data center. The data center, located in Abilene, Texas, is a core component of the "Stargate Project," a 500 billion dollar initiative announced by Donald Trump. This project aims to build large-scale AI data centers across the United States, with the first ten centers slated for Texas. Oracle, along with OpenAI, SoftBank, and investment firm MGX, are equity partners in this project, while Oracle, Arm, Microsoft, NVIDIA, and OpenAI are the initial technology partners.

Oracle plans to acquire around 400,000 of NVIDIA's most powerful GB200 chips, which will be leased to OpenAI for a period of 15 years. This strategic move is designed to provide the necessary computational power for OpenAI's data center, which is expected to become one of the largest in the world upon its completion in mid-2026. The data center will have a power capacity of 1.2 gigawatts, with significant financial backing from Morgan Chase, which provided 9.6 billion dollars in loans, and data center owners Crusoe Energy and BlueOwl Capital, which injected approximately 5 billion dollars in cash.

For OpenAI, this data center represents a significant step in reducing its reliance on Microsoft for computational resources. OpenAI's demand for computational power has outstripped Microsoft's supply capabilities, making this new data center a crucial asset. For Oracle, participating in the "Stargate Project" and operating this data center is a key opportunity to enhance its cloud computing competitiveness and catch up with industry giants like Microsoft, Amazon, and Google. Additionally, Oracle, OpenAI, and NVIDIA are collaborating on a similar project in the Middle East, with plans to build a massive AI data center in the United Arab Emirates. This center is expected to house over 100,000 NVIDIA chips, with the first phase scheduled to go online in 2026.

Traditional data centers and power grids struggle to meet the high-intensity computational, data storage, and energy demands of AI. AI data centers, therefore, represent a critical need. These centers require specialized hardware and infrastructure to handle the large-scale parallel computing demands of AI workloads. Unlike traditional data centers, which focus on storage and basic computing, AI facilities need dense configurations of GPUs and AI accelerators, such as NVIDIA's H100 chips, designed to support the complex matrix computations required by AI models. This investment by Oracle underscores the growing importance of high-performance computing in the tech industry and positions Oracle as a leader in providing advanced computational resources for the most demanding applications and workloads.