Amazon.com is reportedly in discussions to offer its internally developed artificial intelligence processors for deployment in third-party data centers, a significant strategic move aimed at eroding Nvidia's market stronghold.

Peter DeSantis, who leads AI initiatives at Amazon, confirmed the cloud computing giant has initiated talks with potential clients, though he refrained from disclosing their identities.

He remarked in a Paris interview that the company perceives AI infrastructure as a field undergoing rapid transformation and is continuously exploring avenues to reach a broader customer base.

The announcement propelled Amazon's stock, which climbed as much as 1.8% during Thursday's trading session to hit an intraday peak of $241.82.

Amazon's AI accelerator chip, named Trainium and launched in 2020, has already secured several high-profile clients, including OpenAI, Anthropic, and Uber Technologies, who utilize the hardware through Amazon Web Services. The company stated in April that this chip family has garnered revenue commitments exceeding $225 billion.

In his shareholder letter that same month, CEO Andy Jassy indicated it was "quite possible" the company would sell complete racks of its chips to external parties. This forms part of a wider strategy to reorient the conglomerate around artificial intelligence, an area where it is perceived to be trailing competitors.

Amazon, along with other major cloud providers, has been creating its own substitutes for Nvidia's widely used graphics processing units, intensifying these development efforts following the emergence of ChatGPT.

While the AI surge has led to a sharp increase in cloud service revenues, it has also given rise to new specialized AI cloud firms and spurred demand in Europe and elsewhere for "sovereign" services that comply with local regulations and typically keep data processing within national borders.

In a parallel move, Alphabet CEO Sundar Pichai announced in April that Google would start supplying its tensor processing units—a competitor to Nvidia GPUs—to a chosen set of customers for use in their own data centers.

According to DeSantis, Amazon is pursuing a similar path with Trainium partly due to rising international demand, particularly outside the US, for computing resources that are managed locally.

This trend, especially pronounced in Europe, has included calls for reduced dependence on American technology. Speaking at the VivaTech conference in France, DeSantis noted that Amazon Web Services has not experienced any negative impact from this sentiment.

He added that the third iteration of the Trainium chip, which started shipping earlier this year, is now "largely sold out," and that there is already significant interest in a fourth version anticipated to launch next year.

DeSantis rejected the notion that selling Trainium chips outside of AWS would cannibalize the company's core cloud business, stating that the current level of AI resource consumption leaves ample room for growth and that he is not concerned about such competition.

He also highlighted the expansion of Amazon's Graviton chips, a general-purpose processor recently made available to Meta Platforms. Over the past three years, DeSantis said, Amazon has integrated more Graviton chips into its computing infrastructure than any other chip type.