Amazon Doubles Down on Chip Security – Strengthening Homegrown Tech and Expanding Nvidia Partnership
In order to secure an adequate supply of essential data-center processors, Amazon.com Inc.’s cloud-computing division has unveiled upgraded versions of its own computer chips and strengthened its partnership with Nvidia Corp.
The performance of the new domestic Graviton4 chips is up to 30% better than its predecessor, Amazon Web Services said at its annual re:Invent conference in Las Vegas. Computers using processors will start coming online in the coming months.
The company also announced Trainium2, an updated version of the processor designed for artificial intelligence systems. It will begin offering new services starting next year, Amazon said. This chip offers an alternative to the so-called AI accelerators sold by Nvidia — processors that have been vital to building artificial intelligence services.
But Amazon also touted an “expanding partnership” with Nvidia, whose CEO Jensen Huang was joined on stage by his AWS counterpart, Adam Selipsky. AWS is the first major user of the updated version of that company’s Grace Hopper Superchip, and is one of the data center companies hosting Nvidia’s DGX Cloud service.
The new internal chips are part of AWS’s effort to maintain its lead over Microsoft Corp.’s Azure and Alphabet Inc.’s Google Cloud Platform. Earlier this month, Microsoft announced its own processors and AI accelerators. Like Amazon, it has emphasized the efficiency they bring compared to buying parts from traditional vendors like Intel Corp.
AWS said it has built more than 2 million Graviton processors since the project began about five years ago. All of the top 100 users of its EC2 — AWS’s family of processing power for rent — have opted for Graviton-based computing, the company said.
“By focusing our chip design on the real customers who matter, we’re able to deliver the most advanced cloud infrastructure to them,” said Dave Brown, senior vice president of AWS. “Graviton4 is the fourth generation that we have delivered in just five years. And as interest in generative artificial intelligence grows, Trainium2 helps customers train their machine learning models faster, cheaper and with better energy efficiency.
Amazon faces a delicate balancing act. While it wants to tout the technical prowess of its new chips, the company also needs to maintain its relationship with Nvidia. Procuring enough Nvidia products has become a status symbol in the tech industry, with figures like Larry Ellison and Elon Musk boasting about their ability to acquire chips.
So far, Nvidia’s products are considered the best in the industry. Until Trainium or other alternatives can match its capabilities, companies like Amazon cannot afford to alienate Nvidia. And catching up won’t be easy. Nvidia’s acclaimed H100 chip will get a more advanced update called the H200 in the first half of 2024, followed by an all-new design later this year.
Nvidia also continues to diversify. With the DGX Cloud service, it offers software and services aimed at spreading the use of artificial intelligence more widely. AWS also uses a supercomputer based on Nvidia hardware. When not used for internal research and development, the “supercluster” is made available to other users as a service.
