Amazon Expands AI Chip Strategy to Compete More Directly With Nvidia

Amazon Web Services is looking to expand its ambitions in the AI hardware market, with plans that could place the cloud giant in more direct competition with Nvidia. If AWS follows through, it would mark one of the strongest challenges yet to Nvidia’s dominance in AI chips.

Speaking to Bloomberg, AWS Senior Vice President of Utility Computing Peter DeSantis revealed that the company is considering selling its Trainium AI chips to external companies for deployment in their own data centres. While DeSantis confirmed that conversations are underway, he declined to identify any potential customers.

According to Amazon, discussions about commercial chip sales are still in their early stages. The idea follows comments by CEO Andy Jassy in his annual shareholder letter, published in early April, in which he suggested the company’s custom-designed AI chips had generated enough demand to become a standalone business.

“If our chips business were a standalone business, and sold chips produced this year to AWS and other third parties (as other leading chips companies do), our annual run rate would be approximately $50 billion,” Jassy wrote. “There’s so much demand for our chips that it’s quite possible we’ll sell racks of them to third parties in the future.”

Even if Amazon succeeds in building a $50 billion chip business, it would still be considerably smaller than Nvidia, which currently operates at an annual revenue run rate of roughly $326 billion. However, such a business would rival companies like Intel in terms of annual revenue, making AWS a significant new competitor in the AI semiconductor market.

So far, AWS has chosen not to sell its AI chips directly because the company earns far more by using them within its own cloud ecosystem. Beyond charging customers for AI computing workloads processed on Trainium chips, AWS also generates revenue from complementary cloud services, including storage, networking, security, monitoring, and other infrastructure products required to build AI applications.

Another major factor is supply. Amazon has repeatedly said demand for its Trainium processors has exceeded available production. In the same shareholder letter, Jassy noted that the capacity for the current generation of Trainium chips sold out almost immediately. He also said the upcoming Trainium4 generation—which is still more than a year away from commercial availability—has already been fully allocated. Those comments came before AWS expanded its cloud offerings to include OpenAI models.

Selling Trainium chips externally would likely require Amazon to divert supply away from existing AWS customers unless it can significantly increase production through manufacturing partners such as Taiwan Semiconductor Manufacturing Company (TSMC). That presents its own challenge, as Nvidia has become TSMC’s largest customer, recently surpassing Apple.

AWS spokesperson Doron Aronson also confirmed that selling AI chips remains under consideration. “While we’ve historically declined requests to sell chips directly, Andy noted we may sell racks of them to third parties in the future,” Aronson said following a recent media tour of AWS’s chip design facility.

The move reflects a broader shift in the AI chip industry. NVIDIA CEO Jensen Huang recently outlined a vision for a new $200 billion opportunity by expanding beyond GPUs into AI-focused CPUs, placing NVIDIA in more direct competition with Intel and AMD. At the same time, Jassy’s vision of building a $50 billion AI chip business signals Amazon’s intention to compete more aggressively in the same market, setting the stage for an increasingly competitive AI hardware landscape.