Amazon May Start Selling Its Custom AI Chips to Outside Companies. Should Nvidia Investors Be Worried?

For years, Amazon (NASDAQ: AMZN) has designed its own artificial intelligence (AI) chips and kept them to itself, renting out the computing power through Amazon Web Services (AWS) rather than selling the chips themselves. A new report, however, suggests that this could be changing. Bloomberg reported that Amazon is in talks to sell its custom Trainium accelerators for use inside other companies' data centers -- a step that would put its chips in direct competition with the graphics processing units (GPUs) from Nvidia (NASDAQ: NVDA) that power most of today's AI.

For Nvidia shareholders, that raises a fair question. Does a new seller of AI chips -- one with Amazon's scale and years of custom-silicon work behind it -- threaten the company that has come to define the market?

Image source: Getty Images.

What Amazon is exploring

While Amazon is clearly building out its chip business, there's still work to be done before it scales up chip sales in data centers beyond its own. Amazon AI chief Peter DeSantis told Bloomberg that the company has only begun discussions with potential buyers and declined to name any specifically. And there's no public timeline for when chips might ship outside AWS.

CEO Andy Jassy had already flagged the direction during its first-quarter earnings call in April, saying there was "a good chance" Amazon would start offering full racks of Trainium chips beyond AWS "in the next couple of years."

"We view AI infrastructure as rapidly evolving," DeSantis told Bloomberg. "And we're constantly looking at ways to get to more customers."

What gives the talks weight is how far along Amazon's chip effort already is. The company's broader custom-silicon business -- Trainium plus its Graviton and Nitro chips -- crossed a $20 billion annual revenue run rate in the first quarter of 2026 and is growing at a triple-digit pace. And big AI customers are already committed. OpenAI has agreed to about two gigawatts of Trainium capacity through AWS, and Anthropic has signed on for up to five gigawatts of current and future Trainium chips. Uber is a customer too. Demand has been strong enough that Amazon's latest chip, Trainium3, is nearly sold out.

Marvell Technology sits behind much of this, with a five-year, multi-generational AWS agreement that includes the supply of custom AI products and other data center semiconductors, so a wider market for the chip would reach well beyond Amazon itself.

The pitch to outside buyers is cost. Amazon has argued for a while that Trainium does the same work more cheaply than comparable GPUs, and DeSantis suggested that selling chips externally wouldn't eat into AWS, since demand for AI computing still outstrips supply.

Should Nvidia investors worry?

Some caution is fair. Amazon's customers, and Nvidia's, have every reason to want a second source of chips, and a credible lower-cost option puts pressure on the pricing power that has made Nvidia extraordinarily profitable.

This is also part of a broader pattern. Alphabet has its own tensor processing units and has recently begun offering them to a select group of external customers.

But the threat is easy to overstate, at least for now. Nvidia's business isn't shrinking. In fact, it's still accelerating. In its fiscal first quarter of 2027 (the period ended April 26, 2026), Nvidia's data center revenue rose 92% year over year to a record $75.2 billion.

Even Amazon underscores the point. Alongside its Trainium commitments, the company said it would deploy more than one million Nvidia GPUs starting in 2026. In fact, the most aggressive builder of custom chips is also one of Nvidia's largest buyers -- a sign the AI market is widening fast enough to support both, rather than one simply replacing the other.

The market seems to agree that the market is big enough for Amazon to take some share without hurting Nvidia, as Nvidia stock didn't sell off on the news. Both it and Amazon rose about 3% the day the report appeared, helped along by a broad rally in technology stocks.

My take? This is more of something to keep an eye on than fear. Sure, Amazon selling Trainium to outsiders could result in real long-term pressure on Nvidia's pricing power. But Nvidia still holds the bulk of the AI accelerator market.

Daniel Sparks and his clients have no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Amazon, Marvell Technology, Nvidia, and Uber Technologies. The Motley Fool has a disclosure policy.