By Hamza Faisal

OpenAI chief executive Sam Altman has warned Samsung's top executives that demand for the memory chips and data centres powering the AI boom is being counted twice — and that the market should brace for a bumpy stretch.

The warning came during a closed-door session on October 31 at Samsung's Human Resources Development Center in Yongin, South Korea, where Altman met about 50 presidents of Samsung Group affiliates spanning electronics, finance and biotechnology. South Korea's Financial News obtained the session record and reported it on November 11.

Asked whether demand for AI, memory and data centres could saturate or slow over the next five to 10 years, Altman answered plainly: there are ups and downs in the short term. He said there had been cases of demand being double-counted in the market, and that the industry should expect supply shocks and an unsettled period ahead.

Only two companies are driving real demand

Altman's warning centred on the debate over so-called neoclouds — newer cloud providers that rent out infrastructure specialised for AI computing, such as graphics processing units. He argued that overlapping orders and forecasts across the supply chain are inflating the picture of how much demand actually exists.

"There has been double-counting of demand," Altman told the Samsung executives, according to the Financial News report. He added that only two companies, OpenAI and Anthropic, are actually driving a substantial increase in demand — a concentration that leaves the market exposed if either company's spending plans change.

That concentration matters because the memory industry is already running at full tilt. Samsung's entire 2027 supply of high-bandwidth memory is sold out, according to reports, and rival SK hynix has its production fully committed through the same period. OpenAI signed a letter of intent with both Korean chipmakers in October 2025 to secure memory for its Stargate infrastructure project.

Token use points to an enormous long-term surge

Despite the short-term caution, Altman's long-term message was bullish. He told the Samsung executives that AI demand will grow enormously over time, and that the use of tokens — the units of information processed by AI models — will increase sharply as AI spreads into custom software, robots, spacecraft and scientific research.

He illustrated the trajectory with a striking comparison: in early 2020, OpenAI's heaviest user consumed about 100,000 tokens a month. By early this year, the top user had reached 100 billion — a million-fold jump in six years.

Altman said the biggest constraints on that growth will be chip supply chains and data-centre infrastructure, and he predicted shortages of both as AI use expands. The remarks land as investors question whether the AI infrastructure buildout is running ahead of real demand, with memory prices soaring and suppliers racing to add capacity.

A partnership pitch alongside the warning

The session was not only about demand. Altman also proposed closer cooperation between OpenAI and Samsung, including jointly developing next-generation AI terminal devices — an idea covered separately on November 12.

Neither OpenAI nor Samsung has commented publicly on the reported remarks, and the quotes come from Financial News's Korean rendering of the session record. But the message is consistent with Altman's public stance: the AI buildout will be volatile in the near term, and the companies that control chips and data centres will decide how fast it can grow.

For Samsung, the warning cuts both ways: a short-term correction would hit its memory business just as prices are at historic highs — yet if Altman's long-term forecast proves right, the chip shortages he predicts would keep its order books full for years.