AMD reveals next-generation AI chips with OpenAI CEO Sam Altman

Advanced Micro Devices (AMD) has made headlines with the announcement of its next-generation AI chips, the Instinct MI400 series, set to ship in the upcoming year. These innovative chips can be integrated into a comprehensive server rack known as Helios, allowing for thousands of units to work in unison as a single 'rack-scale' system. At a launch event in San Jose, California, AMD CEO Lisa Su emphasized the groundbreaking design, stating, "For the first time, we architected every part of the rack as a unified system." OpenAI CEO Sam Altman joined Su on stage to share that OpenAI plans to utilize AMD's new chips. He expressed his initial skepticism about the specifications, saying, "When you first started telling me about the specs, I was like, there's no way, that just sounds totally crazy. It's gonna be an amazing thing." This rack-scale configuration aims to present a seamless system to users, which is vital for AI customers such as cloud service providers and developers of large language models. These customers are in search of extensive 'hyperscale' clusters capable of powering entire data centers with significant energy resources. "Think of Helios as really a rack that functions like a single, massive compute engine," Su elaborated, drawing a comparison to Nvidia's Vera Rubin racks expected to debut next year. AMD's rack-scale technology positions its latest chips to compete directly with Nvidia's Blackwell chips, which currently feature configurations with up to 72 graphics-processing units. As the primary competitor in the realm of data center GPUs for AI applications, AMD is stepping up its game. OpenAI has been providing valuable feedback on AMD's MI400 development, indicating a collaborative effort in advancing AI technology. With the MI400 series and the already released MI355X chips, AMD is targeting a competitive pricing strategy against Nvidia. A company executive noted that AMD's chips are designed to operate at lower costs due to reduced power consumption, stating, "Across the board, there is a meaningful cost of acquisition delta that we then layer on our performance competitive advantage on top of, so significant double-digit percentage savings." The market for AI chips is projected to surpass $500 billion by 2028, and AMD anticipates significant growth in this sector, despite Nvidia currently holding over 90% market share. Both companies are committed to releasing new AI chips annually, underscoring the escalating competition and the critical importance of state-of-the-art chip technology for major corporations like Microsoft and Amazon. In the past year, AMD has made significant investments in the AI sector, including the acquisition of ZT Systems, enhancing its capabilities to build advanced rack-sized systems. Su remarked, "These AI systems are getting super complicated, and full-stack solutions are really critical." Currently, AMD's leading AI chip, the Instinct MI355X, started shipping last month and will be available for rental through cloud providers in the upcoming third quarter. As companies seek alternatives to Nvidia to manage costs and meet the increasing demand for AI deployment, AMD's latest chips are believed to have superior capabilities for inference tasks due to their high-speed memory and significant computational power. AMD has reported that its chips have been adopted by several of the largest AI customers, including OpenAI and Tesla. Oracle plans to utilize clusters featuring over 131,000 MI355X chips for its clientele, while Meta is deploying AMD technology for its Llama model. Although AMD's AI chip business is still developing, analysts predict substantial growth, with expectations for a 60% increase in the category this year.