Altman’s Memory Critique Meets Samsung’s Executive Team

🔍 Read the full analysis: Altman’s Memory Critique Meets Samsung’s Executive Team on ThorstenMeyerAI.com

OpenAI CEO Sam Altman met about 50 senior Samsung executives at the company’s training center in Yongin, South Korea, on Oct. 31, Samsung said, as memory-chip supply tightens and demand from AI data centers grows. Samsung described the session as leadership training about AI; no new chip-supply deal, prices or delivery volumes were announced.

Samsung said the attendees came from its electronics, finance and biotechnology affiliates. The meeting’s theme was translated by the Korea Herald as “The Next 10 Years Shaped by AI: The Enterprise’s Choice.” Samsung’s account focused on where AI is heading, its effects on business decisions and management, and the markets it could open. It did not describe the session as a contract negotiation.

Bloomberg reported that the visit was intended to deepen cooperation on memory chips and AI infrastructure as shortages intensify. The Seoul Economic Daily reported that Richard Ho, the OpenAI vice president who leads its in-house chip work, accompanied Altman. That outlet also reported, citing industry sources, that OpenAI’s accelerator may use Samsung’s next-generation HBM4 memory. Samsung and OpenAI have not confirmed that reported design choice.

Samsung, as relayed by Yonhap, said Altman told the executives that an advantage comes not simply from adopting AI, but from using it to move faster, make better products and compete. Separately, Yonhap reported that SK Group chairman Chey Tae-won met Altman in Seoul on the same day with SK hynix executives. No details of that discussion were disclosed.

AI Demand Meets Memory Supply

The meeting draws attention to a competition over how scarce memory capacity is allocated. AI data centers require large quantities of memory, including high-bandwidth memory used alongside advanced processors. OpenAI’s expanding infrastructure needs put it among the companies seeking access to that supply, while Samsung and SK hynix are major producers. Any supply agreement could matter to the pace and cost of AI infrastructure, but this meeting produced no announced commitment.

Memory availability also affects businesses and consumers beyond the data-center sector. TrendForce expects conventional DRAM contract prices to rise 10% to 15% in the fourth quarter of 2026. It has warned that PC memory supply could decline in 2027 as more capacity shifts toward servers. These are forecasts, not confirmed price changes for every buyer or device. If tight supply persists, manufacturers could face higher component costs and choices about device prices or specifications.

The competition is not simply between AI companies and PC owners. Memory capacity, product types, manufacturing schedules and long-term contracts all shape what can be supplied. Altman’s remarks express one view of how scarce resources might be used; they do not establish that personal computers are the cause of the shortage or that their memory can readily be redirected to data centers.

Popular right now

Dyson V8 Cyclone Cordless Vacuum

AmazonView Latest Price

Ninja NC501 CREAMi XL Deluxe 11-in-1 Ice Cream and Frozen Treat Maker with Two XL Containers

AmazonView Latest Price

Ninja Foodi Smart 6-in-1 Indoor Grill and Air Fryer, 4 Qt, FG551

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

Altman’s PC Memory Argument

On Oct. 29, two days before the Samsung meeting, Altman discussed memory supply on Cheeky Pint, a podcast hosted by Stripe co-founder John Collison. In response to a question about why high prices had not yet brought more supply, Altman said “the invisible hand is failing right now.” He argued that, from a global allocation perspective, it was “quite terrible” for limited memory to sit in people’s computers while being used “5% of the time.”

That 5% figure was Altman’s estimate, not a measurement supplied by the podcast or an independent study. His point was that shared data centers could make more continuous use of memory than individual computers, not that personal computers are defective. The argument also comes from the CEO of a company that sells access to cloud-based AI services, a relevant perspective when considering his proposed allocation.

OpenAI and four Samsung companies signed a letter of intent in October 2025. Samsung said at the time that OpenAI’s projected demand could reach up to 900,000 DRAM wafers per month. That was a projection in a letter of intent, not a confirmed delivery schedule or a new contract announced at the October 2026 meeting. OpenAI later said Samsung Electronics would offer ChatGPT Enterprise and Codex to employees in Korea and its global device division. Samsung executive chairman Lee Jae-yong visited OpenAI in San Francisco in July 2026, according to the Korea Herald.

The supply issue has become more commercially important to Samsung. In the third quarter of 2026, Samsung’s chip division recorded 106.5 trillion won of the company’s 107.4 trillion won in operating profit, according to the supplied report. Its phone and network business recorded a 0.6 trillion won operating loss, which the report said was partly due to cost pressures. Those figures describe that quarter; they do not by themselves establish how the meeting will affect Samsung’s later results.

„“From a global allocation perspective, it’s quite terrible” that limited memory sits in personal computers “being used 5% of the time.”“

— Sam Altman, speaking on the Cheeky Pint podcast on Oct. 29

No Supply Terms Disclosed

Samsung has not released a full attendee list or described any specific supply negotiation. Neither company announced a new contract, chip volumes, pricing or delivery dates. The reported possibility that OpenAI’s accelerator could use Samsung HBM4 comes from the Seoul Economic Daily’s sources and remains unconfirmed by the companies.

It is also unclear whether the Oct. 31 meeting led to any commitments beyond the existing relationship, or how the earlier letter of intent relates to actual purchases. The supplied reporting does not establish how much memory OpenAI currently receives from Samsung, how much it may seek from other suppliers, or what effect the meeting could have on availability for other buyers.

Forecasts of higher DRAM prices and reduced PC supply in 2027 are not guarantees. Actual prices and availability will depend on production, demand, product mix and supplier decisions. Altman’s estimate about how often PC memory is used does not quantify spare capacity that could be transferred to data centers.

Watch for Supply Agreements

The next concrete indicators would be a public supply agreement or further company statements specifying memory products, volumes and timing. Any confirmation of Samsung HBM4 in an OpenAI accelerator would also clarify the reported connection between OpenAI’s chip-development work and Samsung’s memory business.

For buyers, forthcoming supplier guidance and industry price forecasts will show whether the projected pressure on conventional DRAM and PC memory is developing as expected. As of Nov. 1, 2026, the meeting is confirmed, but its commercial outcome remains undisclosed.

Key Questions

What was confirmed about Altman’s meeting with Samsung?

Samsung said about 50 senior executives from its electronics, finance and biotechnology affiliates met Altman on Oct. 31 at its training center in Yongin. Samsung described the event as an AI leadership program.

Did OpenAI and Samsung announce a new memory-chip deal?

No. The supplied reporting says no new supply contract, volumes, prices or delivery dates were announced. The companies’ October 2025 letter of intent is separate from a confirmed purchase agreement.

What did Altman say about memory in personal computers?

On the Cheeky Pint podcast, Altman argued that memory in personal computers is idle much of the time and could be used more fully in shared data centers. His reference to use “5% of the time” was his estimate, not an independent measurement.

Is Samsung HBM4 confirmed for an OpenAI accelerator?

No. The Seoul Economic Daily reported, citing industry sources, that OpenAI’s accelerator is expected to use Samsung’s next-generation HBM4. Neither company has confirmed that report.

Could the memory shortage affect PC prices or specifications?

It could put pressure on costs and supply, but the exact effect is uncertain. TrendForce forecast a 10% to 15% rise in conventional DRAM contract prices in the fourth quarter of 2026 and warned PC memory supply could decline in 2027 as capacity shifts toward servers. Those are forecasts, not guarantees for retail prices or individual products.

Source: ThorstenMeyerAI.com