Amazon hopes to challenge Nvidia more directly by selling its AI chips

Amazon Web Services is in early-stage talks to sell its custom AI chip, Trainium, directly to other companies for their data centers, a move that would place the cloud provider in more direct competition with Nvidia [1]. The discussions were disclosed by Amazon’s AI chief Peter DeSantis, who told Bloomberg the company is exploring external sales of the chip, though he declined to name potential buyers [1]. The talks follow an April shareholder letter from Amazon CEO Andy Jassy, who noted intense demand for the company’s homegrown processors. “If our chips business was a standalone business, and sold chips produced this year to AWS and other third parties (as other leading chips companies do), our annual run rate would be ~$50 billion,” Jassy wrote [1]. AWS has historically resisted selling its silicon directly, preferring to capture revenue from a cascade of cloud services—storage, security, networking, and monitoring—that customers consume alongside AI compute [1]. A spokesperson, Doron Aronson, confirmed the shift in posture: “While we’ve historically declined requests to sell chips directly, Andy noted it’s quite possible we’ll sell racks of them to third parties in the future” [1]. Any external sales push would face supply constraints. Jassy stated in the same letter that current Trainium capacity sold out almost instantly, and capacity for the next-generation Trainium4 chip, which is more than a year from availability, is already fully reserved [1]. Manufacturing additional volume would require more allocation from partners such as TSMC, where Nvidia has recently become the largest customer, displacing Apple [1]. Nvidia’s revenue run rate stands at $326 billion, dwarfing the $50 billion figure Jassy cited for Amazon’s chip business [1]. Nvidia CEO Jensen Huang has separately identified a $200 billion market for AI CPUs, a segment that would bring Nvidia into the territory of Intel and AMD [1]. AMD, a longtime rival to both Intel and Nvidia, has expanded its data-center and high-performance computing portfolio, including through its 2022 acquisition of Xilinx, which added field-programmable gate arrays to its product line [3]. The potential entry of AWS as a merchant chip supplier adds a new dimension to the AI hardware landscape, which is already shaped by the concentrated spending power of Big Tech firms—Microsoft, Apple, Alphabet, Amazon, and Meta—that collectively account for roughly a quarter of the S&P 500 [5]. Meta, for instance, spent $35.3 billion on research and development in 2022, ranking it third globally, as it builds infrastructure for AI and the metaverse [4]. Amazon’s deliberations signal that the boundaries between cloud providers and chip vendors may continue to blur as demand for AI compute accelerates.