Meet the Company That's Quintupled Its Share of AI Chip Shipments in 2 Years

The market for artificial intelligence chips is dominated by Nvidia (NASDAQ: NVDA). Not only does Nvidia have the best-in-class all-purpose chips for AI training and inference, its CUDA software and NVLink networking ecosystem have cemented its position among large data center customers.

But Nvidia graphics processing units (GPUs) aren't the only game in town when it comes to chips for training and inference. A growing amount of compute demand is going toward custom AI accelerators from big tech companies and smaller chip companies. One company has shown tremendous progress in cutting into Nvidia's dominant share over the last two years, and it's well-positioned to keep growing for years to come. In fact, its share of cumulative AI chip shipments has more than quintupled since the start of 2024.

Image source: Getty Images.

The rapidly growing chip business inside a cloud giant

All the major hyperscalers are developing their own chips to support AI training and inference. Alphabet has its Tensor Processing Unit, Microsoft has its Maia line of chips, Meta Platforms (NASDAQ: META) has its MTIA, and Amazon (NASDAQ: AMZN) has Trainium and Inferentia. These chips are designed to be more cost-effective than Nvidia's general-purpose GPUs, enabling significant savings for large-scale customers, including the hyperscalers themselves.

Amazon has seen demand for its chips explode over the last two years as it builds out its AI services on Amazon Web Services (AWS). Its share climbed from 1.4% in the first quarter of 2024 to 7.5% in the fourth quarter of 2025, according to estimates from Epoch AI.

Amazon's management has repeatedly said that its AI services are growing at a triple-digit rate, and more recently, it gave us additional details on demand for its custom chips. CEO Andy Jassy said that the chips business reached an annual run rate exceeding $20 billion in Q1 and is still growing at triple-digit year-over-year rates. If it were a stand-alone business selling chips directly to other cloud providers, Jassy said it would generate $50 billion in revenue.

For reference, Nvidia generated $216 billion in revenue last year, growing 65% year over year. AMD generated $35 billion in revenue, up 34% year over year. That is to say, Amazon's chips business could be worth hundreds of billions of dollars in its own right.

Jassy mentioned that it's possible the company will start selling its chips to third parties. There are only a few major AI labs that might be interested in buying its chips, and it's unlikely to work with its competing cloud companies. The custom AI accelerators offered by Amazon, Google, and Microsoft are quickly becoming a key differentiator among the major public cloud platforms. Still, a deal with OpenAI or Anthropic could dramatically increase production of Trainium and Inferentia chips.

More growth ahead

Amazon's chip business is only getting stronger, and that's great news for Amazon shareholders.

Meta recently signed a deal to deploy tens of millions of Graviton central processing unit (CPU) cores. Meta says the newest generation of Amazon's custom CPUs is purpose-built for the demands of agentic AI. As reasoning models and agentic use cases increase, Amazon could see even more demand for its leading cloud CPU. Jassy said that the custom CPU is now used by 98% of its top 1,000 customers, starting from nothing in 2018.

The demand for Trainium3, which started shipping this year, is strong, with nearly all of its 2026 capacity sold as of the end of March. Jassy says that Trainium4, which comes out in 18 months, is already seeing very strong demand as well.

The success of the chips business translates into very notable effects on Amazon's free cash flow and operating margin. "At scale, we expect Trainium will save us tens of billions of capex dollars per year," Jassy wrote in his letter to shareholders. Since investors balked when Amazon announced plans to spend $200 billion this year on capital expenditures, growing Trainium adoption should come as very welcome news.

Importantly, Amazon should see improved margins on AWS revenue as a result of paying near-cost for its chips, rather than the retail price for Nvidia's. Combined with the big step up in capital expenditures, it should lead to very strong earnings acceleration.

Adam Levy has positions in Alphabet, Amazon, Meta Platforms, and Microsoft. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Amazon, Meta Platforms, Microsoft, and Nvidia. The Motley Fool has a disclosure policy.