Amazon prioritises Trainium chips over Nvidia for AWS
Updated
Updated · The Motley Fool · May 9
Amazon prioritises Trainium chips over Nvidia for AWS
12 articles · Updated · The Motley Fool · May 9
Chief executive Andy Jassy said Trainium compute has a $225bn backlog within AWS's $364bn total, even as Amazon expects to receive one million Nvidia GPUs by end-2027.
Amazon plans about $200bn in data-centre and chip spending this year, up from $131.8bn, and says using Trainium can save tens of billions in capital expenditure.
The shift could weaken Nvidia's grip on AI accelerators as cloud rivals including Alphabet and Microsoft also push in-house chips to cut costs and differentiate their platforms.
Why is Amazon spending billions on Nvidia GPUs while betting its future on its own custom Trainium chips?
With tech giants building their own AI chips, is Nvidia's reign as the undisputed hardware king about to crumble?
As custom AI chips proliferate, are we heading towards a fragmented hardware landscape that stifles innovation?
Amazon’s Trainium Surge: Massive Backlog, $50B Revenue Ambition, and the Battle for AI Chip Dominance
Overview
Amazon is making a major shift by investing heavily in large-scale AI infrastructure, focusing on custom silicon like Trainium chips. This move allows Amazon to enhance its cloud offerings, lower per-inference costs, and help companies capture more margin as they expand their AI capabilities. By developing proprietary AI hardware, Amazon gives practitioners greater public-cloud capacity for high-throughput model training and inference, more ways to access Trainium-style resources, and faster iteration cycles for deploying large models. This aggressive strategy positions Amazon as a leader in the evolving AI chip market and supports rapid growth in advanced AI services.