Sam Altman on the AI bubble: what the OpenAI chief admits

Which words did Sam Altman use on the AI bubble?

Altman spoke to reporters over dinner in August 2025, and Fortune reported the remarks on 19 August 2025. He said: "Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes." He also said that when bubbles happen, smart people get overexcited about a kernel of truth.

Three statements stand out: investors are overexcited, bubbles form around a real kernel, and some prices are insane. The words are direct and the source is clear. They are also general, and they name no company, no index, and no date, which sets the limit of what a reader can conclude. Read them as a starting point: a short, sourced admission from a person with first-hand knowledge of AI funding, to be tested against numbers.

What does Sam Altman mean by a kernel of truth in an AI bubble?

A kernel of truth is the real innovation at the centre of a bubble. Altman said that in most bubbles in history, such as the tech bubble, there was a real thing. The railway, the internet, and electricity were real. The excess came from prices, not from the technology. Each of those technologies changed the economy. The pattern is a feature of how markets treat big ideas.

That framing admits and defends at once. It admits that prices can run ahead of reality, and it defends the technology as real. A reader can accept both: AI can be valuable and AI prices can be too high. The two claims address different objects, a product and a price.

Which valuations did Sam Altman call insane?

He pointed to start-ups with a staff of 3 raising hundreds of millions of dollars and called such valuations insane (Fortune). The remark targets the small end of the market, where little revenue supports very high prices. It does not target OpenAI, and the quotes in the Fortune report do not say that OpenAI’s own valuation is too high.

That distinction matters for fair reading. A person who criticises the market below him is making a safer statement than one who criticises his own company. Still, the criticism is real, and it comes from someone who sees AI funding pitches daily and so has a close view of the market.

How does Sam Altman’s stake shape his AI bubble remarks?

OpenAI raises money from investors, so Altman benefits when funding flows and when confidence stays high. A warning about overexcitement can seem against his interest, and that adds credibility to the admission. It can also serve him: a leader who names a risk looks candid, and candour builds trust with investors.

Stake does not make the words false, and it tells you where to look for confirmation. Compare what he says with what funding data show. If round sizes and prices keep rising, the admission describes a market that is still excited, and the warning stays open as of October 2026. Funding data is the best check on any claim in the AI bubble debate.

Which points about the AI bubble did Sam Altman leave out?

In the reported remarks he did not give a date for a fall, a level for prices, or a list of companies at risk. He also did not say that AI demand is weak. The sources checked show an admission about investor mood, with the technology defended. Anyone who summarises the remarks as a prediction of collapse adds more than he said. Silence on those points is a reason to look for other evidence, not a reason to fill the gap.

Careful summaries separate three layers: what was said, what was implied, and what a reader infers. The first layer is the Fortune quotes. The second is a view that prices may be high. The third belongs to the reader and should carry a label, so a reader of your notes can tell fact from reading.

Which numbers can you set next to the Sam Altman remarks?

The EQOS Asia AI study records valuations up ×4 in 18 months, about $200B a year invested in AI, and 95% of enterprise AI projects failing in production (MIT 2025, cited by the study, April 2026). Those figures give the remark a scale: investor excitement runs next to slow business results.

The study also scores the enterprise AI market at 0.337, a fragile zone. A leader who says investors are overexcited and a score that finds fragility point the same way. They use different methods, which strengthens the pairing without making it a proof. Read them as two lenses on one market.

How can you use Sam Altman’s admission about the AI bubble?

Use the admission as a prompt to size your own exposure. If the person closest to AI funding sees overexcitement, a decision maker can ask what happens to a budget, a stake, or a contract if prices reset. Write down the answer for three cases: a soft landing, a correction, and a harder fall.

This page is general information as of October 2026, not investment advice. If an AI exposure sits inside a decision of yours in Singapore, ASEAN, or Hong Kong, the EQOS Asia pricing page lists the diagnostic formats that score the exposure with a method you can recompute.

Frequently asked questions

About the author

Stéphane Jambu

President, EQOS Forecast Asia

Stéphane Jambu leads EQOS Forecast Asia and works as a consultant in e-commerce, search visibility, and AI. LinkedIn profile