Samsung, SK join OpenAI's Stargate to target 900K wafers

The Stargate project just got a serious hardware injection. After a meeting in Seoul between OpenAI’s Sam Altman, Samsung’s Jay Y. Lee, and SK’s Chey Tae-won, the two Korean chip giants signed on as strategic partners for the global AI infrastructure initiative. The headline number is the target: scaling production to 900,000 DRAM wafer starts per month for advanced memory chips. That’s the kind of supply chain muscle you need when you’re serious about training frontier models without bottlenecks.

But this isn’t just a chip deal. The partnerships sprawl across the entire AI stack. OpenAI signed an MoU with South Korea’s Ministry of Science and ICT to scope out next-gen data centers outside the Seoul metro area, explicitly aiming for balanced regional growth. SK Telecom is exploring a separate data center build, while Samsung’s construction, heavy industry, and IT services arms are circling opportunities for additional capacity. It’s a coordinated push to give Korea a physical compute backbone, not just a supplier relationship.

There’s a clear political and economic dimension here. The announcement frames these partnerships as central to Korea’s ambition to become a top-three global AI power. Altman’s quote hit the standard notes about talent and infrastructure, but the real signal is the government’s direct involvement via the MSIT. They’re not just cheering from the sidelines; they’re actively evaluating sites and presumably smoothing the regulatory path. For a country that saw its semiconductor dominance become a geopolitical flashpoint, this is a move to lock in relevance for the next computing era.

SK’s Chey Tae-won called it a “landmark moment” and the “official starting point for comprehensive technological innovation,” which is chairman-speak for betting the farm. He’s promising synergies across memory, data centers, energy, and networks. What’s unsaid is the competitive pressure from TSMC and others. Both Samsung and SK will also deploy ChatGPT Enterprise internally, which is a nice customer win for OpenAI, but feels secondary to the main event: wiring Korea’s industrial base directly into the future of compute supply.