Andy Jassy used his annual letter to shareholders to tout Amazon's AI chip progress — and lay down the gauntlet to Nvidia.
Amazon is an Nvidia customer, but the e-commerce giant also makes its own AI chips, called Trainium.
Jassy said Amazon's chip business is "on fire" and that this demand is part of a shift in which companies are diversifying where they buy their AI chips from.
"Virtually all AI thus far has been done on NVIDIA chips, but a new shift has started," Jassy wrote. "We have a strong partnership with NVIDIA, will always have customers who choose to run NVIDIA, and we will continue to make AWS the best place to run NVIDIA."
Already, Amazon has chip deals with OpenAI, Anthropic, and Apple by selling access to Trainium chips via the cloud. It could go one step further: Jassy said it's "quite possible we'll sell racks of them to third parties in the future."
Jassy said that customers "want better price-performance."
He compared the shift to Amazon eating into Intel's dominance in the CPU space with its own chip that it launched in 2018, called Graviton.
"The same story arc is unfolding in AI," he said. Amazon's latest chip, Trainium3, is "30-40% more price-performant" than its previous model, Jassy said.
Last year, Business Insider reported that some startups found that Trainium 1 and 2 had been "underperforming" compared to Nvidia chips.
Jassy added that a "significant chunk" of Trainium 4 has already been reserved ahead of its widespread availability in 18 months.
The Amazon CEO's comments underscore the growing number of competitive threats Nvidia faces as it tries to maintain its dominance over the global advanced chip market. Nvidia has reaped the benefits of its advantage, as its record-breaking status in the $1 trillion club can attest.
Increasingly, AI hyperscalers and frontier model makers are looking to supplement their chip suppliers. Arm is partnering with OpenAI and Meta on its chip, the Arm AGI CPU. Meta is also working with AMD. Google is working with Broadcom on its chip.
Jassy said the annual revenue run rate for Amazon's chip business is now over $20 billion, and growing. Amazon uses its AI chips in its own data centers, which Jassy said could help the company reduce its $200 billion capital expenditure bill.
"At scale, we expect Trainium will save us tens of billions of capex dollars per year, and provide several hundred basis points of operating margin advantage versus relying on others' chips for inference," Jassy wrote.
Google has shown just how powerful a "full-stack" AI strategy can be, allowing a company to control everything from AI models to the chips that train them. The search giant's approach — and sheer market reach — is arguably why it was viewed as either catching up to or even surpassing OpenAI in the generative AI race at the end of 2025.
Jassy's comments show that Amazon, too, would like a world where it doesn't have to rely on other companies as much.
Amazon's stock is down 1.2% this year and up more than 25% in the past year.
Nvidia did not immediately respond to a request for comment from Business Insider.
Correction: April 9, 2026 — This story has been updated to correct a misspelling of Trainium.
