AMD CEO Lisa Su backs CPUs in Agentic AI debate as we shift closer to a 1:1 ratio between CPUs and GPUs

Backed by the AI boom, AMD recently disclosed some impressive earnings figures for Q1 2026. The company brought in $5.8 billion from its data center division, driven by record deployment of AMD EPYC processors and Instinct GPUs. AMD CEO Dr. Lisa Su had some interesting remarks regarding the ongoing debate about the importance of CPUs in today's agentic AI applications.

As we have covered before, CPUs are becoming increasingly important in Agentic AI deployments. While GPUs remain the primary driver, CPUs are no longer a back-of-the-mind consideration in these applications, and thus the demand for CPUs is going through the roof. This has led to multiple CPU price hikes from both Intel and AMD, as they reallocate resources toward more profitable AI and data center ventures.

Responding to an analyst, Dr. Lisa Su explained that the industry is moving swiftly toward a 1:1 ratio of CPUs to GPUs in Agentic AI applications. While this ratio has traditionally been 1:4 or even 1:8 in favor of GPUs, Dr. Su says we might eventually see CPUs overtake GPUs and become even more important. That does sound like a stretch, but such a prediction from an industry leader like Dr. Su cannot be simply ignored.

"And it's very hard to call exactly, but we certainly see the movement towards, you know, where in the past the CPU to GPU ratio was primarily just as a host node in like a one to four or one to eight configuration, now changing and getting closer to a one-to-one configuration or, even, you know, you can even imagine if you get lots and lots of agents that you could have more CPUs than GPUs."

- AMD CEO Dr. Lisa Su

2

Furthermore, when asked whether the rising demand for CPUs would cannibalize the current GPU market, Dr. Su was adamant that it would instead complement GPUs, saying it was "largely additive." For context, Agentic AI involves several autonomous agents within an LLM that complete set tasks without any need for human involvement. For this purpose, CPUs and GPUs need to be deployed in a "host node" for coordination and execution.

We are already seeing movements from major tech giants that back up Dr. Su's prediction. Arm has recently launched its Arm AGI CPU for Agentic AI in collaboration with Meta. Intel and Google have entered a long-term deal to deploy Xeon processors. Meta and Amazon even jumped in on the action with AWS Graviton chips. While currently the ratio still favors GPUs, it wouldn't be a stretch to see CPUs become equally important in the near future.