OpenAI partners with Samsung, SK Hynix for Stargate AI chips
▼ Summary
– OpenAI has signed agreements with Samsung Electronics and SK Hynix to produce DRAM wafers for the Stargate AI project and build data centers in South Korea.
– The companies plan to manufacture up to 900,000 high-bandwidth memory DRAM chips monthly, more than doubling current industry capacity for such chips.
– Stargate is a $500 billion infrastructure project by OpenAI, Oracle, and SoftBank to build AI data centers in the United States.
– OpenAI secured a $100 billion investment from Nvidia for access to over 10 gigawatts of compute capacity and plans five data centers with SoftBank and Oracle to reach 7 gigawatts.
– OpenAI is collaborating with the Korean Ministry of Science and ICT and SK Telecom to build AI data centers outside Seoul and integrating ChatGPT Enterprise into Samsung and SK operations.
In a major push to secure the computational power required for next-generation artificial intelligence, OpenAI has forged strategic partnerships with Samsung Electronics and SK Hynix. These collaborations focus on manufacturing specialized memory chips and constructing advanced data centers in South Korea, all dedicated to powering the ambitious Stargate AI infrastructure initiative. The agreements were formalized in Seoul during a high-level meeting that included OpenAI’s CEO Sam Altman, South Korea’s President Lee Jae-myung, and the top executives from both Samsung and SK Group.
A central component of this deal involves a massive scaling of production for high-bandwidth memory (HBM) DRAM chips. The plan calls for Samsung and SK Hynix to manufacture up to 900,000 of these critical memory wafers each month. According to a statement from SK Group, this output would more than double the current global industry capacity for HBM chips, which are essential for handling the immense data loads of AI model training and inference. These components are destined for use in the Stargate project and other AI-dedicated data centers.
Stargate itself is a colossal undertaking, a joint project by OpenAI, Oracle, and SoftBank with a projected investment of $500 billion. The goal is to establish a network of data centers across the United States specifically designed for AI development. This latest move with South Korean chip giants is part of a broader, aggressive campaign by OpenAI to lock down the hardware and energy resources needed to stay at the forefront of the AI race.
This announcement comes on the heels of several other significant compute-focused investments. Just weeks prior, Nvidia committed up to $100 billion to OpenAI, a deal that grants the AI firm access to over 10 gigawatts of compute capacity through Nvidia’s advanced AI training systems. Shortly after, OpenAI revealed plans with SoftBank and Oracle to construct five new data centers for Stargate, aiming to boost its total compute capacity to 7 gigawatts. In a separate but related agreement earlier in September, Oracle arranged to sell $300 billion worth of compute capacity to OpenAI over a five-year period.
Beyond the chip manufacturing, OpenAI is deepening its roots in South Korea. The company is collaborating with the Korean Ministry of Science and ICT to identify potential sites for AI data centers outside of Seoul. A separate agreement has also been struck with SK Telecom to build a dedicated AI data center. Furthermore, OpenAI has signed additional pacts with various Samsung subsidiaries to explore further avenues for data center construction within the country. As part of the broader partnership, both Samsung and SK Group will integrate ChatGPT Enterprise and OpenAI’s APIs directly into their own corporate operations.
(Source: TechCrunch)
