Amazon Eyes AI Chip Sales to Challenge Nvidia

According to media reports, Amazon is in discussions with potential customers about supplying its in-house Trainium AI chips for deployment in third-party data centers, marking a significant step in the cloud giant’s efforts to expand beyond cloud-based AI services and challenge Nvidia’s dominance in the AI accelerator market.

Peter DeSantis, Amazon’s Senior Vice President responsible for AWS utility computing and AI initiatives, confirmed that the company has begun conversations with prospective customers. However, he did not disclose specific names. An Amazon spokesperson described the initiative as being in the early exploratory stage, while noting that customer interest has already emerged.

Introduced in 2020, Trainium is Amazon’s proprietary AI accelerator designed for machine learning training workloads. To date, the chip has primarily been available through Amazon Web Services (AWS), allowing enterprises to access AI computing resources via the cloud. Major organizations, including OpenAI, Anthropic, and Uber Technologies, have utilized Trainium-powered services. In April, Amazon revealed that Trainium-related offerings had accumulated more than $225 billion in procurement commitments.

Amazon’s custom silicon strategy dates back to its 2015 acquisition of Israeli chip design company Annapurna Labs. Since then, AWS has steadily expanded its portfolio of internally developed processors and AI accelerators.

As demand for generative AI infrastructure continues to surge, leading cloud providers such as Amazon and Google are accelerating investments in proprietary AI chips. These custom processors are increasingly viewed as strategic alternatives to Nvidia’s GPUs, helping cloud operators reduce costs, optimize workloads, and strengthen control over their AI technology stacks.

The potential expansion of Trainium beyond AWS services could significantly broaden Amazon’s presence in the AI hardware market, positioning the company as a more direct competitor to established AI chip vendors while offering enterprises additional options for high-performance AI computing infrastructure.