Amazon is reportedly in talks to sell its custom-designed artificial intelligence chips to external data center operators, marking a significant departure from its long-standing strategy of keeping its proprietary silicon exclusive to its own cloud ecosystem. Peter DeSantis, who leads Amazon’s AI initiatives, confirmed that the company has initiated discussions with third-party providers, though he declined to name specific potential partners.
This move signals a direct challenge to Nvidia’s iron grip on the AI hardware market. By offering its Trainium and Inferentia chips to other data centers, Amazon aims to capitalize on the global shortage of high-performance compute resources while diversifying its revenue streams beyond traditional cloud services.
For years, Amazon Web Services (AWS) has utilized its internal chip-making capabilities to offer lower-cost AI training and inference to its cloud customers. Transitioning into a merchant silicon provider suggests that Amazon believes its hardware is now mature enough to compete on the open market, providing an alternative to the expensive and often supply-constrained GPUs produced by industry leaders.
The decision reflects a broader trend among tech giants to vertically integrate their hardware stacks. As the cost of AI development skyrockets, data center operators are increasingly desperate for more efficient and affordable silicon, making Amazon’s entry into the merchant market a potentially disruptive force in the global semiconductor landscape.