OpenAI enlists Samsung and SK for Stargate

OpenAI isn't just securing GPUs anymore—it is now locking down the entire world's memory supply.

On Wednesday, OpenAI announced major strategic partnerships with South Korean tech giants Samsung and SK Group as part of its massive Stargate infrastructure initiative. The announcement followed a high-profile meeting at the Presidential Office in Seoul featuring OpenAI CEO Sam Altman, South Korean President Lee Jae-myung, Samsung Electronics Executive Chairman Jay Y. Lee, and SK Chairman Chey Tae-won.

The agreements aim to secure essential hardware for next-gen AI and expand data center capacity across South Korea, backing the country's push to become a top-three global AI powerhouse.

Securing 900,000 wafers a month

The most eye-popping part of the deal sits right in the chip fabs.

Samsung Electronics and SK hynix plan to aggressively ramp up advanced memory production, targeting up to 900,000 DRAM wafer starts per month to power OpenAI's upcoming models. In a separate statement, SK Group pointed out that this volume would more than double the current industry capacity for high-bandwidth memory.

With memory bottlenecks remaining one of the nastiest constraints in AI scaling, locking down the two biggest players on the planet is a massive logistical flex. In return, both conglomerates will deploy ChatGPT Enterprise and OpenAI APIs across their own internal operations.

Expanding data centers beyond Seoul

The partnerships go well beyond buying silicon—OpenAI is also helping reshape South Korea's computing footprint.

OpenAI signed a Memorandum of Understanding with the Korean Ministry of Science and ICT to explore building AI data centers outside the Seoul metropolitan area, aiming to spur balanced regional growth. On the corporate side, OpenAI inked a deal with SK Telecom to evaluate local AI data centers, alongside agreements with Samsung subsidiaries—including Samsung C&T, Samsung Heavy Industries, and Samsung SDS—to explore additional capacity.

This move caps off an unreal month of capital expenditure for OpenAI. The company recently saw Nvidia commit up to $100 billion for over 10 gigawatts of compute, Oracle sign a $300 billion compute sales deal over five years, and SoftBank join Stargate's wider $500 billion infrastructure blueprint. OpenAI is systematically securing every piece of the AI physical stack.

Why it matters

For developers and everyday users, memory shortages have quietly been the hidden tax on AI progress, causing rate limits, latency spikes, and high token costs. By locking down massive dedicated HBM pipelines, OpenAI is ensuring its next frontier reasoning and multimodal architectures won't stall on the assembly line.

From an industry standpoint, we think OpenAI is playing a game of capital encirclement that few rivals can match. Instead of simply buying off-the-shelf servers, Altman is negotiating directly with foreign heads of state and dominant industrial chaebols. Aligning Stargate with South Korea's national ambition to become a top-three AI power gives OpenAI unprecedented leverage over global supply.

The strategy is brilliant for locking out competitors from critical chip allocations, but the sheer scale carries undeniable financial risk. Tying up hundreds of billions of dollars across data centers, memory fabs, and international infrastructure means OpenAI must deliver revenue growth that matches this unprecedented industrial buildout.