Huang Rejects AI Doom Talk, Says US Needs More Fossil Plants for AI
Nvidia CEO Jensen Huang says the United States has not built out enough fossil fuel plants to meet power demand from AI. He laid out that position while discussing the future of energy and AI's impact on the planet in the latest episode of The Ezra Klein Show. The Verge
The episode, titled 'Jensen Huang Thinks A.I. Alarmism Has Gone Too Far,' was published on September 23, 2026. The New York Times The New York Times also published an edited transcript of the conversation. The New York Times
Huang argued that fossil fuels must be used over the next several years because there is not enough sustainable energy to make a difference. His formulation was sequential. Use fossil generation now, with hope to transition afterward. The Verge
He framed the tradeoff in stark medical terms. AI, he said, is like surgery. In order to save people you have to hurt them first. In his words, AI must inflict "an enormous amount of pain and suffering" to save them. The Verge
That language fits the episode's larger premise. Unlike many leaders of frontier AI labs, Huang does not think AI could wipe out humanity. Apple Podcasts
A separate exchange earlier in September underlined how tightly compute policy and energy policy are now linked. Donald Trump called Huang while Huang was onstage in Los Angeles. During that call, Trump criticized calls to slow down and regulate A.I. The New York Times
The broader context here is a collision between deployment timelines. Frontier systems are being scaled on commercial schedules measured in quarters. Power plants, transmission upgrades and permitting run on schedules measured in years. For practitioners managing capacity planning, that mismatch is familiar. It forces near-term choices about existing generation while longer-term supply is still being built.
In this author's view, Huang's comments are best read as an infrastructure argument rather than a research argument. He is not debating model architectures or alignment methods. He is saying the constraint on AI progress, at least in the United States, is electrons. That is a pragmatic claim, and worth flagging because it shifts responsibility away from labs alone and toward utilities, regulators and builders.
Looking at what this means for technologists, the near-term path Huang describes is uncomfortable but clarifying. If sustainable supply cannot yet carry incremental AI load at the scale being deployed, operators will face explicit tradeoffs between throttling deployment and accepting higher-carbon power in the interim. Over the long arc, technology has repeatedly found ways to do more work per watt once incentives and engineering attention align. The open question is whether that efficiency gain, plus new clean supply, arrives fast enough to shorten the fossil bridge Huang is describing.
