Nvidia said its tech is still a full generation ahead of Google’s AI chips

Nvidia pushed back hard on Tuesday after fresh doubts from Wall Street about whether Google’s AI chips could weaken its grip on the industry.

The company said its tech is still a full generation ahead, and it made sure to say it with a straight face.

Nvidia also reminded everyone that it continues supplying Google, even as talk grew that Meta might test Google’s tensor processing units for its data centers.

Nvidia’s stock fell 3% earlier in the day after that Meta report came out, but the company used its X post to calm markets and smack back at the idea that its lead was slipping.

Nvidia insisted its platform “runs every AI model and does it everywhere computing is done,” and said its Blackwell chips stay more flexible and powerful than Google’s ASIC-based TPUs, which only work for specific tasks inside one company.

Nvidia hits back with Blackwell advantages

Nvidia said its chips offer “greater performance, versatility, and fungibility than ASICs,” and it pointed out that its graphics processors now control more than 90% of the AI chip market, based on analyst data.

Google’s in-house TPUs have gained more attention lately, mostly because Blackwell chips are pricey, but Nvidia stressed that these custom chips are limited to Google’s own systems.

Google does not sell TPUs to outside buyers. It only uses them internally or rents them to customers on Google Cloud.

Google has tried to show the strength of its line by releasing Gemini 3 earlier this month. The model was trained fully on TPUs, not Nvidia GPUs.

A Google spokesperson said demand is rising for both its custom TPUs and for Nvidia’s GPUs, adding, “We are committed to supporting both, as we have for years.”

Nvidia CEO Jensen Huang also stepped in during an earnings call this month. He said Google continues to buy Nvidia’s GPUs and confirmed that Gemini can also run on Nvidia hardware.

Jensen said he has been in touch with Demis Hassabis, the DeepMind chief, who texted him that the industry’s “scaling laws” remain “intact.”

Scaling laws are the idea that larger data loads and more chips produce stronger AI models. Nvidia believes these laws will push demand even higher for its own chips and full systems across the industry.

If you're reading this, you’re already ahead. Stay there with our newsletter.