Agentic AI puts the CPU back at center stage as Dell and AMD chase every watt
Agentic AI is rewriting the business development playbook for chipmakers and server builders. Graphics processing units still grab the headlines, but as supercomputing and AI converge, the next wave of demand is landing on central processing units that must fit inside the power budgets data centers already have. The shift has been building for years. […] The post Agentic AI puts the CPU back at…
Agentic AI is revitalizing the importance of central processing units (CPUs) in the chipmaker and server builder landscape. While graphics processing units (GPUs) remain the talking point, the convergence of supercomputing and AI is driving a surge of demand for CPUs within data centers' power limitations. These CPUs, particularly in an agentic AI architecture, require high speed and concurrency.
Scott Tease, the vice president of business development for high-performance computing and sovereign AI at AMD, noted that the shifting power dynamics are prompting a reevaluation of CPU-centric design. Similarly, Dell Technologies' David Schmidt highlighted the company's collaboration with AMD in developing rack-scale AI platforms, underscoring the need to optimize power and cooling efficiency.
Dell's PowerEdge R9825 server, built on AMD's EPYC 9006 processors, exemplifies this approach, offering improved power and cooling efficiency. This efficiency is crucial, as it allows for the consolidation of legacy machines and a significant reduction in cooling power requirements. Tease explained that reclaimed power could potentially fuel more agentic workloads, where a single researcher's ideas could be transformed into thousands of compute jobs.
This trend is expected to drive more IT investment and necessitate innovative power solutions to support the growing agentic AI demands. Both Dell and AMD are actively shaping the future of compute by addressing the power constraints that have historically limited CPU performance in data center environments.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.