Amazon is exploring a new way to push deeper into the artificial intelligence chip market. The company is in talks to sell its Trainium3 processors to outside customers for use in their own data centers, according to a Bloomberg report. The move would broaden Amazon's chip business beyond its own cloud operations and strengthen its challenge to Nvidia's market lead.

Peter DeSantis, who leads Amazon's AI efforts, said the company has started discussions about offering the custom chips to other businesses. He did not identify any possible buyers. The talks suggest Amazon sees demand for its in-house silicon beyond its own infrastructure, where the chips are already used to support Amazon Web Services customers and internal AI workloads.

Amazon has spent years developing custom chips as part of a strategy to lower dependence on outside suppliers and control more of the hardware stack behind its cloud services. Its Trainium line is designed to provide a lower-cost alternative for training AI models, an area that has been dominated by Nvidia chips. Expanding sales to other companies would mark a notable step from an internal tool to a commercial product aimed at the broader market.

The effort comes as large cloud providers and chipmakers race to secure a larger share of the booming AI infrastructure business. Nvidia remains the leading supplier of accelerators used in AI data centers, but its dominance has encouraged competitors to build their own alternatives. Amazon's move would place it more directly in that contest, potentially giving cloud customers another option as they look for more supply and lower costs.

The company has not disclosed a timeline for a wider rollout, and the discussions are still early. It is also unclear how Amazon would package or price the chips for third-party buyers. For now, the key signal is that Amazon is testing interest in selling the hardware beyond its own cloud ecosystem.

If Amazon succeeds, outside sales of Trainium3 could create a new revenue stream and deepen its role in the AI hardware market. It would also fit with the company's broader push to pair its cloud services with custom technology tailored for AI workloads, an approach that has become increasingly important as demand for computing power continues to rise.

The report adds to a broader industry trend in which major tech companies are building more of the chips that power their own AI products, then weighing whether those designs can attract outside customers as well. For Amazon, Trainium3 may become not only an internal advantage, but also a product it can use to compete more aggressively in one of the most lucrative parts of the AI economy.