FILE PHOTO: A view shows the logo of Amazon at the Amazon Fulfillment Center during a media tour ahead of the holiday season in Tepotzotlan, Mexico, December 13, 2023. REUTERS/Gustavo Graf/File Photo
(Reuters) - Amazon.com's cloud computing unit on Tuesday said it will offer free computing power to researchers who want to use its custom artificial intelligence chips, aiming to challenge Nvidia's popularity among those researchers.
Amazon Web Services (AWS) said it will offer credits to use its cloud data centers that it values at $110 million to researchers who want to tap Trainium, its chip for developing artificial intelligence models that competes with chips from Nvidia, as well as Advanced Micro Devices and Alphabet's cloud division.
AWS said researchers from Carnegie Mellon University and the University of California, Berkeley, are taking part in the program. The company plans to make 40,000 of the first-generation Trainium chips available for the program.
The move comes as AWS, still the largest cloud computing company by sales, has seen a sharp challenge from Microsoft as software developers look to harness new types of chips for AI work. AWS is hoping to gain attention for its own AI chips by taking a different strategy than Nvidia, said Gadi Hutt, who leads business development for the AI chips at AWS.
To program Nvidia's chips, most AI developers use what is called Cuda, Nvidia's flagship software, rather than programming the chip directly. AWS instead plans to publish documentation about the most fundamental part of its chip - what is called the instruction set architecture - and let customers program the chip directly.
Hutt said the approach is aimed at luring large customers who might want to make small tweaks that could add up to big gains when using tens of thousands of chips at a time.
"Think about folks that are using infrastructure and putting hundreds of millions of dollars, if not more" toward rented computing power, Hutt said. "They would take any opportunity possible to increase performance and reduce the cost."
(Reporting by Stephen Nellis in San Francisco; Editing by Leslie Adler)
