AMD CEO Lisa Su Says CPUs Will Match GPUs 1-to-1 in AI’s Next Phase

On Bloomberg Businessweek Weekend’s May 8th segment, AMD chief Lisa Su delivered a forecast that should make GPU-only investors uncomfortable. The current ratio of 4 to 5 GPUs per CPU in AI infrastructure is about to compress toward 1-to-1 as agentic AI and inference workloads start hammering the CPU layer.

What Su said on the AMD call

On the Q1 FY2026 earnings call, Su laid out the math behind the thesis. “…if you’re installing a gigawatt of compute, the ratio — there’s a percentage of CPU as part of that gigawatt will increase. Some of the conversation in the industry has been about CPU to GPU ratios… And it’s very hard to call exactly, but we certainly see the movement towards where in the past, the CPU to GPU ratio was primarily just as a host node in like a 1:4 or 1:8 configuration node, now changing and getting closer to a 1:1 configuration or even — you can even imagine if you get lots and lots of agents that you could have more CPUs and GPUs.”

That is a structural claim, and AMD backed it with a TAM revision. Six months ago at the November Analyst Day, AMD pegged the server CPU market at roughly $60 billion growing at 18% annually. The new number is over $120 billion by 2030, growing at greater than 35% annually. Agentic workloads, Su argued, are “largely additive to the TAM” rather than cannibalizing GPU spend.

Data Center revenue hit $5.78 billion, up 57% year-over-year, with server CPU revenue alone growing more than 50%. Q2 guidance calls for server CPUs to grow greater than 70% year-over-year. Total Q1 revenue was $10.25 billion, beating the $9.91 billion estimate, with non-GAAP EPS of $1.37. You can read the 8-K exhibit for the full release. Shares are up 112.55% year-to-date through May 8.

Intel and Flex as the unloved beneficiaries

Intel (NASDAQ:INTC | INTC Price Prediction) CEO Lip-Bu Tan echoed Su nearly verbatim, telling investors “The next wave of AI will bring intelligence closer to the end user, moving from foundational models to inference to agentic. This shift is significantly increasing the need for Intel’s CPUs and wafer and advanced packaging offerings.” Intel’s Data Center and AI segment grew 22% year-over-year to $5.05 billion, and Xeon 6 was selected as host CPU for NVIDIA’s DGX Rubin NVL8 systems. The stock is up 238.54% year-to-date.

Flex (NASDAQ:FLEX) builds power and cooling infrastructure for AI data centers and is up 135.3% year-to-date. Q4 FY2026 revenue hit $7.48 billion, up 16.9%, and the board approved spinning off its Cloud and Power Infrastructure segment, targeting 65%-75% revenue growth in FY2027. Cathie Wood called it “going back to the future” with legacy chipmakers resurging.

If Su is right that gigawatt-scale deployments need CPUs in something approaching parity with accelerators, the AI infrastructure trade widens considerably. Su sees a clear path to more than $20 in EPS over the strategic timeframe. The buildout requires all hands on deck.