Amazon Web Services (AWS), the global leader in cloud computing services, is currently exploring the possibility of incorporating Advanced Micro Devices’ (AMD) new artificial intelligence chips. While no definitive decision has been made yet, an executive from AWS revealed this information to Reuters.
This announcement was made at an AMD event where the company outlined its strategy for the AI market, an arena currently dominated by Nvidia, AMD’s primary competitor. Despite AMD’s revelation of some technical specifications for an upcoming AI chip that could surpass Nvidia’s current offerings in certain aspects, AMD’s failure to disclose a flagship customer for the chip caused a decline in its stock value.
In discussions with Reuters, Lisa Su, the CEO of AMD, presented a unique approach to securing major cloud computing clients. She emphasized providing a comprehensive selection of components required to build systems capable of powering services like ChatGPT. This approach allows customers to tailor their infrastructure according to their specific needs, utilizing standard industry connections.
Su stated, “We anticipate that many individuals will value the freedom of choice and the ability to customize their data centers based on their requirements.”
Although AWS has not made any public commitments to adopt AMD’s new MI300 chips in its cloud services, Dave Brown, the Vice President of Elastic Compute Cloud at Amazon, confirmed that AWS is actively considering their implementation. Brown expressed the ongoing collaboration between AWS and AMD in determining the best course of action. He acknowledged the benefits derived from AMD’s work on designing chips that seamlessly integrate with existing systems.
While Nvidia does offer its chips individually, the company has also proposed a comprehensive solution called DGX Cloud, which entails cloud providers offering an entire system designed by Nvidia. Oracle is Nvidia’s initial partner for this offering.
In response, Brown revealed that AWS declined to engage in collaboration with Nvidia for the DGX Cloud initiative. After careful examination of the business model, AWS concluded that it did not align with their extensive experience in constructing reliable servers and their existing expertise in the supply chain domain.
Brown further explained that AWS prefers to develop its servers from scratch. In March, AWS began selling Nvidia’s H100 chip but only as part of systems they independently designed and engineered.