Nvidia’s two biggest customers may be turning into its biggest threats.
For three years, Nvidia’s stock price has defied gravity on the assumption that the AI industry needs its chips. When Google and Amazon reported first-quarter earnings on Wednesday, both companies signaled their ambition to sell their custom AI chips directly to customers.
So far, Google’s TPUs and Amazon’s Trainium chips have only been available through Google and Amazon’s cloud services. Customers can pay to use them, but they don’t own them.
Nvidia is currently the undisputed leader in AI chips. There’s no sign of this momentum stopping anytime soon, but recent statements from Google and Amazon suggest they’re ready to challenge Nvidia’s throne.
Amazon CEO Andy Jassy gave Nvidia a tough fight in his annual letter to shareholders earlier this month.
“Nearly all AI to date has been done on Nvidia chips, but a new transition has begun,” Jassy wrote, adding that it is “very possible” that Amazon could start selling its chips directly to customers.
He laid out the timeline for the plan during Wednesday’s earnings call, saying there is “a good chance” Amazon will start offering full racks of Trainium chips beyond its own cloud “within the next few years.”
Google showed even stronger commitment and a closer timeline.
On Wednesday, Google CEO Sundar Pichai said publicly for the first time that the company plans to deliver TPU chips to a “select group of customers” in its data centers this year, but said the “vast majority” of revenue from those sales won’t materialize until 2027.
Pichai said there are big opportunities in Google’s semiconductor business, which could lead to funding for next-generation chips. That flywheel could generate huge business for Google. Morgan Stanley said in a December research note that selling 500,000 TPU chips could add about $13 billion in revenue to Google’s balance sheet in 2027.
The company said in its 10th quarter filing that it has so far signed a “limited number of contracts” to supply TPUs to customers.
That’s also where things can get tricky. Google and Amazon are also big customers of Nvidia. They buy Nvidia’s chips and lease them to customers in their own data centers. Amazon and Google both announced that they will continue to work with Nvidia.
Nvidia stock fell more than 4% on Thursday. The company did not immediately respond to Business Insider’s request for comment.
“I’m worried, but I’m not worried.”
Breaking Nvidia’s chip dominance won’t be easy, analysts told Business Insider.
“NVIDIA should be concerned, but not worried,” said Alvin Nguyen, senior analyst at Forrester. Nvidia has built a strong ecosystem of hardware, software and support, making it easy for customers to choose between them, he said.
“Selling a product and having access to a product are two very different things,” he said, adding that Amazon and Google need to provide services such as education and support to companies looking to buy their chips.
Google and Amazon’s chip racks are also “very bespoke and proprietary,” customized for their respective data centers, said Patrick Moorhead, CEO and chief analyst at Moor Insights & Strategy. That poses a challenge in achieving mass adoption, he said.
Moreover, the chip market is not a zero-sum game. AI companies are becoming increasingly diverse, using chips from multiple suppliers simultaneously. For example, OpenAI has chip deals with Nvidia, AMD, Broadcom, and others.
But Bernstein analyst Mark Schmulik wrote in a note Thursday that custom silicon is becoming an “increasingly important part of the AI story” for Google and Amazon. Both companies tout their chips as more cost-effective than Nvidia’s.
As the needs of the AI industry change, so do the opportunities. Earlier this month, Google announced a new TPU chip focused on inference (the process of running a model after training). This reasoning becomes increasingly important as more companies bring AI agents online.
Béatrice Vallée, senior analyst for enterprise technology and services at GlobalData, said Google and Amazon’s decision is an “unusual move” to diversify the chip space and reduce the cloud provider’s dependence on Nvidia chips.
“This process will take years, but there is no going back,” she said.