Big Tech firms look to make their own AI chips, threatening Nvidia's dominance

It’s Big Tech week on Wall Street. On Wednesday, we got quarterly results from Alphabet, Meta, Microsoft, and Amazon. Shocker: They’re all still spending a lot on AI. One of the biggest expenses is chips, the things that made NVIDIA the most valuable company in the world. Now, Big Tech companies want to get in on that action.

All of these companies are what’s known in the industry as hyperscalers, meaning they operate a ton of data centers.

“Right now, almost the entire ecosystem is built on top of Nvidia's chips,” said Chris Miller, an economic historian at Tufts who wrote the book “Chip War.”

The hyperscalers have been stepping up investments in their own custom silicon, he said, “both because they want alternatives to Nvidia, but also because if you build your own chip, you can specialize it to your particular needs.”

Google and Amazon are also providing AI processors to outside customers, including Meta, which is also building its own chips, and Anthropic.

“There is the potential for this to entirely disrupt Nvidia, so I think it is a pretty significant risk,” said Jay Goldberg, a semiconductor analyst at Seaport Research.

But right now, there’s no such thing as too many chips, said analyst Stacy Rasgon at Bernstein Research. “We just need more,” he said, adding that the rise of AI agents has sent demand for computing power off the charts.

“I don't really care who's winning or losing right now,” he said. “I think it's the wrong question.”

The right one, Rasgon said, is whether these companies can build capacity fast enough.