# Nvidia's Jensen Huang says two AI labs face too much pressure

Markdown copy of https://realalpha.app/news/nvidias-jensen-huang-says-two-ai-labs-face-too-much-pressure/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

[News](https://realalpha.app/news/) · BG2

By [Rushabh Sheth](https://realalpha.app/news/authors/rushabh-sheth/) Oct. 22, 2026, 11:53 p.m. ET

Jensen Huang, who runs Nvidia, said on the podcast BG2 that the AI building boom does not depend on two big AI labs. The episode was posted last Thursday.

This touches your savings more than your apps. Host Brad Gerstner pointed to a recent pullback in infrastructure companies, on fears that OpenAI could not pay for its computers. Those companies sit in many retirement accounts.

Huang expects spending on data centers to reach $3 to $4 trillion a year by the end of the decade. Data centers are buildings full of computers. He said $1.5 trillion of that is computing power, up from about half a trillion before AI. He said Moore's law is over, so twice the chips now costs twice as much. A gigawatt of older general purpose computing cost $5 billion, he said, against about $50 billion for a Vera Rubin, Nvidia's newest chip system.

On the labs, he said this at 1:12:09. "We're putting way too much pressure on two labs."

[Yahoo Finance](https://finance.yahoo.com/technology/ai/articles/jensen-huang-two-gigawatts-us-163235324.html) reported on Oct. 19 that Huang said two gigawatts is like $80 billion to Nvidia, with 43 new gigawatts coming in 2027. A gigawatt is a measure of the power a data center draws.

Nvidia sells these chips, so higher costs per gigawatt help its own sales. Gerstner said at least a third of next year's build-out is replacing older computers. Huang's forecasts are his own, and he did not say how the labs will pay their bills.

Watch the 5 minutes where he breaks it down. [1:08:17](https://realalpha.app/v/eS7qQ3WJSrw/?t=4097)

## Sources

- The moment: [Jensen Huang | Does AI Revenue Justify Capex, '27 AI Bottlenecks, Singularity, Open vs Closed & more](https://realalpha.app/s/why-the-infrastructure-build-out-doesnt-need-trillion-dollar/?t=4329), BG2, Oct. 15, 2026, from 1:08:17 to 1:12:57
- The story: [Jensen Huang: “Two Gigawatts to Us Is Like $80 Billion.” 43 New Gigawatts Coming in 2027](https://finance.yahoo.com/technology/ai/articles/jensen-huang-two-gigawatts-us-163235324.html), Yahoo Finance, Oct. 19, 2026
- Earlier from Jensen Huang: [Putting CUDA on GeForce nearly sank NVIDIA, but install base won](https://realalpha.app/s/putting-cuda-on-geforce-nearly-sank-nvidia-but-install-base/), Lex Fridman Podcast, Mar 2026
- Earlier from Jensen Huang: [Test every plan against the speed of light, not continuous improvement](https://realalpha.app/s/test-every-plan-against-the-speed-of-light-not-continuous/), Lex Fridman Podcast, Mar 2026
- Earlier from Jensen Huang: [The power grid sits mostly idle, and data centers could use that slack](https://realalpha.app/s/the-power-grid-sits-mostly-idle-and-data-centers-could-use/), Lex Fridman Podcast, Mar 2026

**Quote freely with a link.** Link the moment's page above, which plays the second it was said. [How to cite us](https://realalpha.app/press/). Something wrong? Write to [hello@realalpha.app](mailto:hello@realalpha.app) and we'll correct it ([how we pick and check items](https://realalpha.app/news/about/)).