Altman Courts Samsung in Seoul as Memory Shortage Bites AI Data Centers

OpenAI CEO Sam Altman sat down with roughly 50 senior Samsung executives in South Korea, as the ChatGPT maker scrambles to lock up memory chips for the AI data centers it plans to build. Bloomberg first reported the gathering, which Samsung confirmed.

The meeting took place at Samsung's Resources Development Center in Yongin on Saturday, October 30. Chiefs from the conglomerate's financial, electronics, and biotechnology businesses all showed up. That roster signals Samsung treated the visit as a group-wide strategic event, not a routine component negotiation.

Samsung's own account of the talks leaned general. "The Samsung executives and CEO Sam Altman discussed the direction of AI technology development and its potential applications, as well as new markets and business opportunities that will emerge from it," the company stated. The discussion reportedly centered on strengthening ties established by a letter of intent signed in October to cooperate on global AI infrastructure — including supplying memory semiconductors for OpenAI's Stargate project.

No new supply agreement emerged. Samsung emphasized AI transformation and future business opportunities rather than announcing fresh memory commitments. Altman did, however, hand Samsung's AI Center head a Codex Micro Keyboard, a device built for use with OpenAI's coding agent.

Why is OpenAI chasing Samsung now?

The backdrop is a memory market stretched to its limit. AMD forecasts DRAM revenue will jump 177% in 2026 to roughly $416 billion. Trendforce estimates DRAM supply will fall 1% to 2% short of demand this year, and expects the imbalance to worsen in 2027. New fab capacity takes years to build, which is why Micron recently warned the shortage could persist into 2028.

Data centers aren't the only buyers fighting for those wafers. PCs, smartphones, and other electronics are competing for the same limited DRAM supply — and demand may be nowhere close to peaking.

Who controls the supply?

Samsung and SK hynix together controlled about 63% of global DRAM revenue in the second quarter, according to Counterpoint Research. In high-bandwidth memory — the specialized DRAM that AI accelerators depend on — the two Korean firms held roughly 83% of global revenue. That concentration gives South Korea a central role in supplying the hardware behind the AI buildout, and it explains why technology executives keep flying to Seoul.

AMD CEO Lisa Su visited earlier this month to discuss securing additional HBM supplies from Samsung. "It is important to plan three to five years ahead with our partners," Su said.

What did Altman say about memory?

Before the trip, Altman stirred controversy with comments on the Chick Pink podcast about how the world's limited memory supply should be allocated. He argued that DRAM sitting inside personal computers is often underused, calling the situation "quite terrible," and suggested data centers could use memory more intensively by sharing compute resources across AI workloads. The argument drew widespread criticism, given that OpenAI's own services, including ChatGPT, depend heavily on centralized computing infrastructure.

How are manufacturers responding?

Samsung and SK hynix have been upgrading their factories in China to meet surging demand for AI-related memory and storage. Those particular upgrades target NAND flash production rather than DRAM — but they show how capacity plans are shifting toward data center hardware as demand moves.

With HBM sold out years ahead and new DRAM capacity years away, expect more pilgrimages to Yongin — and harder bargaining over every wafer — as AI builders race to secure the memory their GPUs can't run without.