Amazon CEO takes aim at Nvidia, Intel, Starlink, more in annual shareholder letter

Amazon shareholder Andy Jassy’s annual letter reads something like a Kendrick Lamar diss song, if the rapper was a talking-head CEO and not a Pulitzer Prize-winning lyricist.

Which means, you have to know the history to understand all the opponents Jassy is aiming for, as well as some great personal stories about his unfulfilled dream of becoming a sportscaster and watching hockey games with his father.

Of course, Jassy doesn’t throw down the gauntlet directly. He takes a subtle approach. For example, in his challenge to Nvidia, he writes, “We have a strong partnership with NVIDIA, it will always have customers who choose to use NVIDIA” and will always support these chips in its cloud.

But he also says: “Almost all AI so far has been done on NVIDIA chips, but a new shift has begun.” AWS customers, he says, “want price efficiency” that means Amazon’s home-grown Trainium AI chips.

Jassy says the demand for this chip is so high that the capacity of the new one, Trainium3, is about to run out. Surprisingly, he says this figure is almost sold out with Trainium4, which is still 18 months away from availability.

This means Trainium has reached an estimated revenue of $20 billion. But if Amazon were a manufacturer selling its products to others, it would be at $50 billion in ARR, he says.

Granted, Nvidia made $215.9 billion in actual revenue last year. Nvidia may not be shaking in its boots, just yet. Nevertheless, Jassy presents Trainium as a newcomer and a threat.

Jassy also did not hold back Intel. He points out that AWS’s homegrown Graviton CPU, which competes with the Intel x86 architecture, “is now widely used by 98% of the top 1,000 EC2 customers,” which are the world’s largest companies. Two companies even asked to “buy everything of our Graviton position in 2026,” he writes (emphasis his). “We cannot agree to these requests given the needs of other customers, but it gives you an idea of ​​the need.”

He promised that Amazon’s Starlink competitor, Amazon Leo, which is scheduled to launch in mid-2026 is already successful, too. It has won contracts from Delta Airlines, AT&T, Vodafone, Australia’s National Broadband Network, NASA, among others.

Interestingly, he also said that Amazon could look into selling robots one day. It could turn all the data from its 1 million warehouse robots into “robotic solutions” for industrial and consumer use, he wrote. Is there a humanoid Amazon in the future? We will see. He talked about Amazon’s other businesses, too, like same-day delivery, groceries, and drones.

But mostly, Jassy tried to appear in court about the hundreds of billions of dollars in capital expenditures he made. In February, he announced plans to spend $200 billion by 2026 on capex, mostly to build AWS data centers. That’s more than any other big tech company, which also spends heavily on capex. Jassy’s pitch to shareholders makes sense considering Amazon’s stock has fallen below $200 a share and hasn’t recovered.

“We are not investing nearly $200 billion in capex in 2026,” he wrote, using as an example that his deal with OpenAI includes the model maker promising to spend $100 billion on AWS. Of course, there are some who doubt that OpenAI will meet all of its spending promises.

In a nod to that, Jassy emphasizes that beyond OpenAI, “there are several other customer deals that have been completed (and unannounced), or deep in progress,” lined up to buy AWS capacity.

We will have to wait and see. Those who cause the bubble are never those who recognize (or acknowledge) its existence. “I’ve followed the public debate about whether this technology is widely used, or whether we’re in a ‘bubble.'” But he declares in the book that, at Amazon at least, it isn’t.