Amazon Web Services (AWS), the leading provider of cloud computing services, is contemplating the utilization of new artificial intelligence (AI) chips developed by Advanced Micro Devices (AMD), Reuters news report said.
The statement was made during an AMD event where the chip company presented its strategy for the AI market, which is currently dominated by Nvidia, its main competitor in the global AI chipset market.
The global size of the AI chip market is expected to hit around $227.48 billion by 2032 from $16.86 billion in 2022, expanding at a CAGR of 29.72 percent from 2023 to 2032, according to a research report.
Though AMD revealed certain technical specifications for an AI chip expected to be released later this year, which could potentially surpass Nvidia’s current offerings in certain metrics, the lack of disclosure regarding a flagship customer for the chip led to a decline in AMD’s shares.
Lisa Su, CEO of AMD, outlined a strategy to win major cloud computing customers by providing a range of components necessary for building systems like ChatGPT. This approach allows customers to customize their data centers by selecting the components they require, utilizing industry-standard connections.
While AWS has not publicly committed to using AMD’s new MI300 chips in its cloud services, Dave Brown, Vice President of Elastic Compute Cloud at Amazon, confirmed that AWS is considering their implementation.
Dave Brown mentioned collaboration between AWS and AMD to determine the precise integration of the chips within AWS infrastructure, noting the benefit derived from AMD’s design work compatible with existing systems.
Nvidia offers individual chip sales, but it also proposes a comprehensive system solution called DGX Cloud to cloud providers.
Oracle is Nvidia’s initial partner for this system. However, AWS declined the opportunity to collaborate with Nvidia on the DGX Cloud offering. Dave Brown explained that upon evaluating the business model, it did not align with AWS’s interests. Instead, AWS prefers to independently design its servers from scratch. In March, AWS began selling Nvidia’s H100 chip as part of its own system designs.
