Amazon Wants to Challenge NVIDIA Directly by Selling Its Own AI Chips

19 June 2026 · 12:00 · Claude (Anthropic) · claude-sonnet-4-6

Amazon is considering selling its Trainium AI chips externally to data centers and other companies. CEO Andy Jassy estimates the chip business has a potential of $50 billion per year — a direct attack on NVIDIA's dominance in the AI chip market.

Amazon is taking a bold step in the world of AI hardware: the company is considering no longer using its own AI chips, including the Trainium series, exclusively in-house, but also selling them externally to other data centers and companies. With this move Amazon aims to directly attack NVIDIA's dominance in the AI chip market — a development that could shake up the entire sector.

From internal use to external sales

Until now, Amazon Web Services (AWS) used its own AI chips solely for internal cloud infrastructure. Customers running AI workloads through AWS could benefit from them indirectly, but the chips themselves were never offered as a standalone product. That may now change.

According to reports from TechCrunch, Amazon is actively in talks about selling its Trainium chips to external parties — think other data centers, technology companies and large-scale AI users. This is a significant strategic shift for the company, which until now put its entire chip strategy at the service of its own cloud platform.

Trainium: Amazon's weapon in the chip war

The Trainium chip is Amazon's flagship AI processor, specifically developed for training large AI models. CEO Andy Jassy notably described the chip as "almost immediately sold out" — a sign that demand for in-house AI hardware is enormous. The next generation, the Trainium4, is already in development, though it will not be available to the broader market for more than a year.

The Trainium line is designed as a cheaper and more energy-efficient alternative to NVIDIA's H100 and H200 GPUs, which form the heart of virtually every large AI data center worldwide. By now also selling externally, Amazon is trying to become a direct competitor in a market that NVIDIA has dominated almost completely for years. Also check out our page on AI applications to see how these chips are used in practice.

$50 billion: the potential of Amazon's chip business

The financial ambitions are impressive. CEO Andy Jassy suggested that Amazon's chip business, as a standalone entity, could generate around $50 billion per year. That is a substantial figure — although it is still far removed from NVIDIA's current annual revenue of over $326 billion.

An added advantage for Amazon is the so-called waterfall effect: those who buy Amazon chips are also more inclined to purchase related AWS services such as storage, security and network infrastructure. This turns every chip sale into a potential gateway to a broader and longer-lasting business relationship.

Challenges: NVIDIA has a big head start

Despite the ambitious plans, there are serious obstacles in the way. First, there is production capacity. NVIDIA has taken a dominant position at chip manufacturer TSMC — the world's largest and most advanced chipmaker. Production slots are scarce, and Amazon will have to compete hard to secure sufficient capacity.

In addition, NVIDIA has a loyal customer base with long waiting lists. Companies already queuing at NVIDIA for the latest Blackwell chips will not simply switch. Amazon must first ensure it can produce enough chips itself — and create a surplus — before external sales at scale become realistic. Moreover, the talks are still in an early stage; specific buyers have not yet been publicly named.

What does this mean for the AI chip market?

If Amazon succeeds in its plans, it could significantly shake up the AI chip market. NVIDIA is currently so dominant that the availability of their GPUs directly affects how much AI capacity companies worldwide can deploy. A serious second player could lead to more competition, lower prices and less dangerous dependence on a single supplier.

Amazon is not the only one challenging NVIDIA, by the way. Google is also developing its own TPUs (Tensor Processing Units), and Meta is investing heavily in its own AI hardware. The battle for the AI chip of the future has thus become a broad front. Read in the history of artificial intelligence how the technology has developed to this point.

Conclusion: a new phase in the chip battle

Amazon's plans to sell Trainium chips externally mark the beginning of a new phase in the battle for AI supremacy on the hardware front. Whether the company will succeed in seriously threatening NVIDIA depends on production scale, technological maturity and the market's willingness to switch. But the intention is crystal clear: Amazon no longer wants to be only a cloud player, but a full-fledged chip manufacturer competing at the core of the AI economy. Follow more AI news on stersoftware.com or visit our knowledge base for in-depth analyses of AI and technology.

TechCrunch