Recent reports suggest Amazon is considering a significant strategy shift: selling its powerful in-house AI chips to other companies. This move would challenge Nvidia's dominance in the AI hardware market.
This news isn't a complete surprise, as Amazon has been dropping hints for a while. In an April 2026 letter to shareholders, CEO Andy Jassy revealed that Amazon's custom chip business is already a massive success, generating over $20 billion in annual revenue and growing rapidly. He explicitly mentioned the possibility of selling racks of these chips to third parties, estimating it could become a $50 billion business. This clearly signaled Amazon's commercial ambitions beyond its own cloud.
The timing for such a move seems ideal due to several converging factors. First, the demand for AI chips is insatiable. Nvidia reported record-breaking data center revenue of $75.2 billion in one quarter, yet it still struggles to meet all the demand. This supply gap creates a huge opportunity for credible alternatives.
Second, geopolitical tensions, particularly U.S. export controls on advanced chips to China, add another layer of complexity. Reports of Nvidia chips being smuggled have raised concerns about supply chain security and compliance. For companies and governments seeking a reliable and compliant supply, an alternative from a U.S. company like Amazon would be very attractive.
Finally, Amazon's chips have already earned significant trust within the industry. Tech giants like Meta and the leading AI company Anthropic have signed major deals to use Amazon's silicon. Anthropic's commitment alone is valued at over $100 billion and includes access to a massive amount of Trainium capacity. This validation from major players proves that Amazon's chips are a high-performance, viable option, making them an easier sell to external customers.
In essence, Amazon has a proven product, established demand, and a perfect market window. By potentially becoming a merchant silicon provider, Amazon isn't just creating a new revenue stream; it's positioning itself to reshape the entire AI infrastructure landscape.
- Merchant Silicon: Refers to chips designed by a company and sold on the open market to other companies, rather than being used exclusively for its own products. Nvidia is a prime example.
- Rack-level systems: Fully configured and pre-integrated server racks containing processors, memory, networking, and cooling, sold as a complete unit. This simplifies deployment for customers.
- Hyperscaler: A massive-scale cloud service provider that offers computing, storage, and networking services, such as Amazon Web Services (AWS), Google Cloud, and Microsoft Azure.