Oracle’s $40 Billion Nvidia Chip Acquisition for Texas Data Center Development
Oracle Corp. is making strides in the field of artificial intelligence by purchasing Nvidia Corp. chips valued at $40 billion to construct an advanced data center aimed at supporting AI applications. This sizable transaction involves the procurement of around 400,000 GB200 Grace Blackwell processors, each of which integrates two of Nvidia’s cutting-edge Blackwell B200 graphics cards coupled with a 72-core central processing unit.
This significant acquisition facilitates the establishment of an expansive data processing hub presently being built in Abilene, Texas. Once completed, the center is anticipated to draw approximately 1.2 gigawatts of power, an amount equating to the energy consumption of roughly one million homes. This installation is a crucial component of Stargate, an initiative spearheaded by OpenAI to develop a comprehensive network of AI-centric data facilities across the United States. The ambitious Stargate project is backed by collaborations with Oracle, Softbank Group Corp., and Abu Dhabi’s MGX investment fund.
The financial commitment to Stargate is substantial, with projections estimating its cost to reach up to $500 billion over the coming four years. Reports suggest both OpenAI and SoftBank will each contribute $18 billion for a majority share in the venture. In contrast, both Oracle and MGX have reportedly pledged $7 billion each to the initiative.
The new Texas data center, designed for housing Oracle’s extensive array of Nvidia chips, is slated for operational launch by mid-2026. This site is a collaborative effort by Crusoe Energy Systems LLC, a Denver-based data center startup, and Blue Owl Capital, an investment firm. The partnership has secured approximately $15 billion in funding, largely in debt form, through institutional investors to support the construction and operation of the facility.
Oracle has secured a 15-year lease on the data center, intending to allocate its computational resources to OpenAI for the advancement of AI-driven projects. This partnership emerged following the termination of OpenAI’s exclusive cloud hosting arrangement with Microsoft Corp., one of its prominent investors in the previous year.
The technological marvel, the GB200 Grace Blackwell chip, promises substantial efficiencies and enhancements. It’s engineered to handle large language models with up to 25 times more energy efficiency compared to its predecessor. Furthermore, the chip incorporates built-in AI models designed to pre-emptively identify and address technical concerns. Additionally, the Decompression Engine component accelerates data retrieval from databases, showcasing its vital role in AI data management.
Furthermore, these chips’ dual GPUs are crafted by Taiwan Semiconductor Manufacturing Co. using an advanced variant of its N4P process. N4P signifies a significant evolution of the company’s five-nanometer node, enabling an increase in GPU manufacturing efficiency. This method decreases the necessity for photomask-dependent chip manufacturing steps, thereby elevating production speeds and improving Nvidia’s capability to satisfy the growing demand for its processors.
