On October 1 (local time), manufacturer ChatGPT said it had signed a memorandum of understanding with Samsung and SK Hynix on the production of DRAM wafer panels for the Stargate AI infrastructure project, and is studying the possibility of building more data centers in Korea.
The deal follows a meeting in Seoul between OpenAI CEO Sam altman, South Korean President Lee Jae-myung, Samsung Electronics Chairman Jay Y. Lee and SK Group Chairman Chey Tae-won.
According to the commitment, Samsung and SK Hynix will expand their scale to produce up to 900,000 high-bandwidth DRAM chips (HBM) per month, providing for the Stargate data center system. This figure, according to SK Group, is double the current high-bandwidth memory chip manufacturing capacity of the entire industry.
Stargate is a $500 billion super-scale infrastructure project jointly implemented by OpenAI, SoftBank and Oracle to build data centers to serve AI training and development in the US.
Over the past month, OpenAI has steadily increased its investment in computing capabilities. Nvidia has also previously said it will pour up to $100 billion into OpenAI, giving the company access to more than 10 gigawatt of computing power from advanced GPUs.
In parallel, OpenAI, softBank and Oracle will build five new data centers for Stargate, bringing the total expected capacity to 7 gigawatt.
In addition, Oracle has also agreed to sell $300 billion in computing capacity to OpenAI within 5 years. In Korea, OpenAI is working with the Ministry of Science and Technology to find a location for AI data centers near Seoul, and has signed a separate agreement with SK Telecom and Samsung's subsidiaries to expand infrastructure.
Samsung and SK Group will also integrate ChatGPT Enterprise and OpenAI's APIs into internal operations, marking an important step forward in large-scale AI application in Asia.
