Nvidia says old AI chips can still cash checks

Old chips, new swagger

Nvidia CEO Jensen Huang basically took the “out with the old, in with the new” playbook and tossed it in the shredder. On Wednesday, he said the company’s A100 GPUs — first launched in 2020 — can remain “mission-capable” through 2029.

That matters because one of the bear-case arguments around AI hardware has been pretty simple: technology moves so fast that last year’s chips become next year’s paperweights. Huang’s message is the opposite. In Nvidia-land, older chips can still earn their keep, especially when they’re doing inference, smaller-model work, and other tasks that don’t need the latest shiny black box.

The real moat is bigger than silicon

Huang also leaned hard into Nvidia’s software stack, CUDA, as the secret sauce. The pitch is basically: the chip matters, sure, but the platform is what keeps customers coming back like it’s the only coffee shop open on Monday morning.

That’s a big deal for investors because longer useful lives can make GPUs feel less like disposable gadgets and more like financeable assets with recurring rental-like cash flows. Translation: if the hardware keeps generating revenue for years, it becomes easier to fund, deploy, and justify at scale.

CoreWeave and the capital parade

The timing is not random. CoreWeave disclosed a deal to rent Nvidia A100s through 2029, which gives Huang’s thesis some real-world backup. And Nvidia also said it’s working with a squad of heavyweight capital providers — Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR — to help mobilize more than $500 billion over time.

So yes, this is partly a chip story. But it’s also a financing story, a software story, and a “how long can AI infrastructure keep making money before it turns into landfill?” story.

Big picture: if Nvidia can convince the market that its older GPUs are still productive assets, that’s bullish for margins, demand, and the whole AI boom narrative.