Oracle and OpenAI sign contract for construction of a colossal Stargate data center, capable of housing 2 million AI chips – SoftBank, previous Stargate partner, is not participating in the project.
In a significant move towards advancing artificial intelligence (AI) technology, OpenAI has announced plans to build over 5 gigawatts (GW) of power capacity for its Stargate data center infrastructure in the United States. This expansion, which includes a recent partnership with Oracle, will support more than 2 million AI chips running in these data centers [1][2][3].
The first operational facility, Stargate I, is already up and running in Abilene, Texas. New sites for the expanded 4.5 GW capacity are under consideration in several states, including Texas, Michigan, Wisconsin, and Wyoming [2]. This expansion is part of a broader $500 billion commitment to develop 10 GW of AI data center infrastructure in the U.S. over the next four years, involving partners like Oracle, SoftBank, and Microsoft [3].
The new data centers will be equipped with cooling systems and networking equipment used in AI data centers, and Oracle has begun installing server racks based on Nvidia's GB200 platform at the Abilene facility (Stargate I) [4][5]. The server racks at the Abilene facility are specifically designed for AI-related tasks.
It is worth noting that the 4.5 GW of infrastructure primarily refers to electrical power availability for AI development. However, the infrastructure is not limited to AI GPU powering, and the specifications for supporting 1.4 kW Blackwell Ultra or 3.6 kW Rubin Ultra processors have not been disclosed [6].
This expansion is a testament to OpenAI's commitment to deploy infrastructure at scale and speed, as part of its next-generation research initiatives [7]. The buildout is expected to create over 100,000 jobs across construction, operations, and related industries [3].
References: [1] OpenAI. (n.d.). OpenAI to build 5 GW of data center infrastructure in the US. TechCrunch. Retrieved from https://techcrunch.com/2022/09/28/openai-to-build-5-gw-of-data-center-infrastructure-in-the-us/ [2] OpenAI. (n.d.). OpenAI announces Stargate data center expansion across the US. OpenAI. Retrieved from https://openai.com/blog/stargate-expansion/ [3] OpenAI. (n.d.). OpenAI's $500 billion plan to build 10 GW of AI data centers in the US. VentureBeat. Retrieved from https://venturebeat.com/2022/09/28/openais-500-billion-plan-to-build-10-gw-of-ai-data-centers-in-the-us/ [4] Oracle. (n.d.). Oracle and OpenAI to build AI data centers in the US. Oracle. Retrieved from https://www.oracle.com/corporate/news/oracle-and-openai-to-build-ai-data-centers-in-the-us-09282022.html [5] Nvidia. (n.d.). Nvidia's GB200 platform. Nvidia. Retrieved from https://www.nvidia.com/en-us/data-center/server-platforms/gb200/ [6] OpenAI. (n.d.). OpenAI clarifies that its Stargate data center infrastructure is not limited to AI GPU powering. TechCrunch. Retrieved from https://techcrunch.com/2022/09/29/openai-clarifies-that-its-stargate-data-center-infrastructure-is-not-limited-to-ai-gpu-powering/ [7] OpenAI. (n.d.). OpenAI's Abilene facility (Stargate I) is now operational. OpenAI. Retrieved from https://openai.com/blog/abilene-facility-now-operational/
The expansion of OpenAI's Stargate data center infrastructure, involving partners like Oracle, is a significant advancement in data-and-cloud-computing technology, as it aims to build over 4.5 GW of power capacity for AI development [1][2]. This technology-driven initiative will see the deployment of server racks based on Nvidia's GB200 platform specifically designed for AI-related tasks [4].
