Amazon is moving to loosen Nvidia's grip on the market for artificial-intelligence chips by leaning on silicon it designs itself.
According to Bloomberg, Amazon is in talks to sell its custom AI chips, a step it frames as a bid to cut into Nvidia's dominance. Rather than only buying processors from the current market leader, Amazon would put its own hardware in front of more customers.
The chips at the center of the effort are Amazon's Trainium line, which Memeburn reports is taking aim squarely at Nvidia. The name signals their purpose: processors built for the heavy computing work behind training and running AI systems, the same territory where Nvidia has set the standard.
For most of the recent AI boom, Nvidia's processors have been the default choice for companies building advanced AI, giving the firm enormous influence over pricing, supply and who gets access to scarce computing power. A move by Amazon to not just use but sell an alternative could chip away at that position.
The sources describe the development at a high level and do not detail pricing, customers, timing or performance comparisons. What is clear from both reports is the direction: Amazon wants to be less dependent on Nvidia, and it is positioning Trainium as a rival rather than a purely in-house tool.
Why it matters: if a company the size of Amazon can offer a credible alternative to Nvidia's chips, it could give AI builders more choice and ease the bottleneck that has shaped the cost and pace of the entire industry.