Sam Altman’s reported message to Samsung’s leaders was two-sided: AI infrastructure demand may be exaggerated or uneven in the near term, but he expects AI use to grow substantially over time. The distinction matters for chipmakers and data-center investors weighing long-range expansion against less certain short-term forecasts.
1
2
What Altman reportedly told Samsung
Financial News said it obtained a question-and-answer record from Altman’s October 31 meeting in Yongin with roughly 50 presidents of Samsung affiliates.
1
5 In that discussion, he acknowledged cases where demand for memory semiconductors and data centers had been counted more than once, and said there could be short-term ups and downs.
1
The reported remarks came amid debate over a possible “neocloud” bubble. Altman reportedly identified OpenAI and Anthropic as the main companies driving the sharp increase in demand, and cautioned that supply shocks and a turbulent period could follow.
2 These comments describe a risk in how demand is forecast and timed; they do not establish how much demand was double-counted.
Why he was still optimistic about long-term demand
Altman reportedly contrasted near-term volatility with a much larger long-term opportunity. When asked whether demand might slow or saturate over the next five to ten years, he acknowledged short-term fluctuations but predicted strong growth in AI usage.
1
4
One illustration in the reported Q&A was monthly token consumption by the heaviest user: about 100,000 tokens in early 2020, rising to 100 billion earlier in 2026.
4 That comparison was presented as evidence of how quickly usage could expand. It is a usage example, however—not proof that every projected data-center or memory requirement will materialize.
A secondary account of the meeting described the business logic as providing AI at high volume and thin margins, supported by vertically integrated infrastructure intended to lower costs.
9 In that framing, investment in computing capacity is a bet on serving much greater future use efficiently, even if near-term demand estimates remain uncertain.
The meeting was about cooperation, not a confirmed new supply deal
The visit took place as OpenAI sought memory chips and the infrastructure to support its expansion, with Bloomberg describing the meeting as an effort to deepen cooperation with Samsung on memory and AI infrastructure.
17
26 Financial News separately reported Altman’s interest in working closely with Samsung on a joint device.
8
Those discussions should not be confused with confirmed agreements. Reports said neither company announced a new memory supply deal following the meeting.
20
39 The available reporting supports a distinction between exploring cooperation and securing a newly announced supply commitment.
The takeaway for AI infrastructure watchers
Altman’s comments were not simply a warning that AI demand is a bubble, nor a guarantee that all planned capacity will be needed. They point to a tension: near-term demand forecasts may be volatile or double-counted, while rising AI usage underpins his longer-term optimism.
1
2
4 For readers assessing the AI buildout, the key question is whether projected demand represents distinct, durable usage—and when it will translate into actual demand for chips and data centers.