AMD CEO Lisa Su Anticipates $500 Billion Growth in AI Data Center Accelerator Market by 2028, Fueled by Inferencing Demand

AMD’s Chief Executive Officer has expressed strong optimism regarding the future of the data center sector, particularly emphasizing a burgeoning demand for AI accelerators.

AMD Aims High with Ambitious Plans for the AI Accelerator Market

According to AMD, the current market lacks sufficient computational resources to support the growing applications of artificial intelligence (AI).The company anticipates a robust increase in its AI/data center revenue. In her presentation during the Advancing AI keynote, Lisa Su highlighted that the data center accelerator market is projected to expand at a staggering 60% compound annual growth rate (CAGR).This momentum is expected to continue in the coming years, positioning the AI accelerator sector’s valuation at an impressive $500 billion. This growth trajectory presents myriad opportunities not only for AMD but also for its competitors, including NVIDIA. The AI landscape is ripe for expansion, with various new possibilities surfacing for major technology firms.

The potential for growth in the AI accelerator market is significant, primarily because AI’s role has evolved beyond model training. Today, artificial intelligence is leveraged across a multitude of applications that require substantial computational power, which is facilitated by AI-specific GPUs. According to AMD’s CEO, AI technologies have transcended mere data centers, finding their way into cloud services, edge computing, and client-side applications. Each of these domains necessitates powerful accelerators to generate crucial computing capabilities. As competition intensifies following AMD’s recent announcements, the question of who will best capitalize on this surging demand for accelerators becomes increasingly pertinent.

To seize the opportunities presented by this expanding market, AMD is implementing three distinct strategies aimed at enhancing its AI portfolio. These include developing cutting-edge compute engines, fostering an open ecosystem, and providing comprehensive full-stack solutions to ensure that customers can fully leverage Team Red’s AI capabilities. Recently, AMD unveiled its latest Instinct MI350 AI lineup, featuring a revolutionary CDNA 3 architecture manufactured on TSMC’s 3nm process technology. This new lineup includes powerful models equipped with substantial HBM3E memory stacks and a thermal design power (TDP) of up to 1400W for its flagship model, the MI355X. AMD has asserted that this development allows them to compete head-to-head with NVIDIA’s Blackwell in performance metrics.

On the software side, AMD also introduced the new ROCm 7 software stack, which includes improved frameworks like vLLM v1, llm-d, and SGLang, along with a focus on various optimizations. The enhancements within ROCm 7 include: