Sam Altman appears to be more concerned about AI capex waste than the market does.
The Nasdaq is at fresh records today and up 0.7% but he spoke with CNBC and touched on a point that many critics have highlighted: wasteful spending. AI capex right now is at astronomical levels and the spending among the US hyperscalers this year will be roughly equivalent to total German government spending this year. About $1 of every $20 generated in the US economy this year is going to AI capex.
Speaking alongside the Stargate Michigan announcement — a one-gigawatt campus carrying a price tag north of $45 billion that OpenAI says it’s “very confident” will pay off on demand signals — the man who anchors the entire buildout volunteered the case against it:
“I know some great stuff is happening but there is also a tonne of waste. How long do I have to wait for it to show up in revenue? How long do I have to wait to really get costs under control? I assume
the industry will figure that out pretty quickly but that is a fair issue… I would bet that by a year or two from now
there is a much better rationalization of company spend relative to outcomes.”
When AI capex starts to get cut, I strongly suspect that many of the pick-and-shovel names in AI will see a sharp correction in stock prices.
At the same time, there was plenty of cheerleading. The bullish framing was everything you’d expect from someone whose product is compute:
“We understand what demand looks like and how much people want to use these models.”
“The more compute we can provide, the more we can offer.”
“It seems clear that AI is going to massively increase prosperity.”
He singled out coding as the standout use case — “coding is the magic right now” — and argued the market underrates how much AI each person will eventually consume once it’s “running in the background” for each of us.
The other thing worth flagging was the messaging pivot on jobs. By a wide margin the topic Altman kept circling back to was unwinding the idea that AI is coming for your job:
“I think it’s a terrible message that AI is going to take jobs.”
That’s a notable shift in tone. As recently as February, at the India AI Summit, he was telling the same network that people are right to feel anxious, leaning on the argument that firms adopting AI fastest are also the ones hiring most. Today he reframed the whole jobs narrative as a bad message rather than an open question — convenient for a CEO who needs public buy-in for $450 billion of data-center construction.
What’s also notable is that he spent a longer time talking about that topic than anything else, and it came at a moment where other AI executives are saying the same thing
