According to Economic Daily News, citing The Information, Amazon Web Services (AWS) is adopting a price-cutting strategy to persuade customers to switch to its own Trainium AI chip-powered servers instead of the more expensive NVIDIA solutions.
As noted in the report, a long-time AWS cloud service customer revealed that AWS recently pitched them on renting servers powered by Trainium AI chips, claiming that their computing power is comparable to servers using NVIDIA’s H100 chips but at just a quarter of the cost.
This move comes as tech giants like Google and Microsoft ramp up efforts to develop their own in-house AI chips, aiming to lessen their dependence on NVIDIA’s dominant processors, as highlighted by a report from GuruFocus.
However, as TechStartups notes, pricing alone may not be sufficient to lure customers away from NVIDIA. The bigger challenge for AWS is whether it can provide a smooth transition for developers whose AI workloads are heavily reliant on NVIDIA’s software ecosystem.
AWS executives stated in December 2024 that major tech companies, including Apple, Adobe, and Anthropic—each seeking alternatives to NVIDIA—have been testing the latest Trainium chips and have seen promising results, as noted by Economic Daily News, citing The Information.
According to a Reuters report from December 2024, Apple executive Benoit Dupin has confirmed that Apple is using AWS’s Trainium2 chips.
Meanwhile, the report from Economic Daily News also mentions that although Amazon has listed the hourly rental rates for Trainium-powered and NVIDIA-powered server clusters, the comparison is not made under entirely identical conditions.
In December 2024, Amazon unveiled its most powerful self-developed chip, Trainium3, at the AWS Cloud Conference. According to its press release, Trainium3 will be the first Amazon chip to be fabricated using a 3nm process. The chip will be produced by TSMC and is set for release by the end of 2025, as noted by another report from Economic Daily News.
Read more
(Photo credit: Amazon)