Altman meets about 50 Samsung executives on memory chips

Felix Laurent 31 Oct 2026 3 min read

Big tech

Sam Altman met about 50 Samsung executives in South Korea on Saturday, Oct 31. Bloomberg reported that the aim was to deepen cooperation on memory chips and AI infrastructure as global shortages intensify. Samsung's own account of the day talks about AI strategy, not chips.

What happened in Yongin

Samsung said it hosted Altman at its Human Resources Development Center in Yongin, south of Seoul, the Korea Herald reported. About 50 presidents and senior executives came from Samsung affiliates in electronics, finance and biotechnology. The Korea Herald names none of them.

The session was titled "The Next 10 Years Shaped by AI: The Enterprise's Choice". According to Samsung, it covered how AI can reshape industries, corporate decision-making and management, and what leaders will need. Before meeting Altman, the executives took part in hands-on training with generative AI. The meeting builds on a deal that is a year old.

The earlier deal

In October 2025, Samsung Electronics signed a letter of intent with OpenAI to cooperate on global AI infrastructure, including advanced memory chips for OpenAI's Stargate project. Samsung said then that OpenAI's memory demand was projected to reach up to 900,000 DRAM wafers per month. SK Hynix signed a similar letter at the same time.

The 900,000 figure is OpenAI's projected demand, as Samsung described it. It is not a supply target set by Samsung. Under the same deal, Samsung SDS agreed to look at AI data centers and enterprise AI services, and Samsung C&T and Samsung Heavy Industries agreed to work on floating data centers.

Ties have grown since. In June 2026, OpenAI said Samsung would roll out ChatGPT Enterprise and Codex to employees in Korea and in its global Device eXperience division. In July, Samsung chairman Lee Jae-yong visited OpenAI's headquarters in San Francisco. The Yongin meeting is the second round of top-level AI training at Samsung this year, after one in June. About 2,300 executives have been trained since then, and Samsung aims to train all employees by February 2027. Bloomberg tied the visit to memory for a reason.

The memory constraint

Samsung's preliminary results put its third-quarter operating profit at 107.4 trillion won, about $81.1B, the Korea Herald reported. That would make it the first tech company to earn over 100 trillion won in a quarter. Samsung began mass shipments of HBM4 in February 2026.

Its rival SK Hynix reported third-quarter operating profit of 83.33 trillion won, up 632% from a year earlier, with a record operating margin of 79.7%. It says its DRAM and HBM are effectively sold out through 2029. "Customers are requesting more than twice the volume we can supply," its head of HBM sales, Kim Ki-tae, said. The company expects memory shortages to last through 2030.

Altman is not the only visitor. Over the past year Lee Jae-yong has met Nvidia's Jensen Huang, Tesla's Elon Musk, AMD's Lisa Su and Altman, the Korea Herald notes. A post by analyst Rohan Paul adds that Huang and Su have each visited South Korea twice since last year. We could not check that against a primary page. What the meeting produced is another matter.

What has not been announced

Tech-insider, retelling the coverage, reports that no purchase agreement, chip allocation, delivery schedule or dollar figure came out of the day. No participant was quoted saying the shortage caused the meeting. Samsung's account, as reported by the Korea Herald, does not mention memory at all.

So was Altman in Yongin to secure memory chips? Possibly, but nothing on the record says so. Samsung describes a session on AI strategy for its top executives, part of a company-wide training push. Bloomberg ties it to the memory crunch, and the context supports that reading: OpenAI's projected demand is huge, and Korea's two memory makers say they are sold out for years. But no deal, volume or price came out of the meeting. Until one does, the chip angle is the outlets' framing, not an announced result.