AMD is being yanked towards server CPUs by customers who suddenly realised AI boxes need more than GPUs.
AMD chief executive Lisa Su is seeing ‘unexpected’ volume for EPYC-class server CPUs, with customer commitments ramping fast and the supply chain struggling to keep pace.
Speaking at the gathered throngs at the Morgan Stanley Conference, she said the CPU side has surprised even her, despite going in optimistic.
Su said: “We’re seeing, actually, as much as, you know, I’m very, very excited about the GPU portion of the business, I mean, the CPU portion of the business has actually far exceeded my expectations in terms of demand. I was pretty bullish to begin with, right?”
She says big customers are flagging that CPU compute sitting next to AI workloads was under-forecast, forcing AMD to catch up.
“If you talk to our top customers, they’re like, “Wow, you know, Lisa, the, like, the demand for CPU compute sitting along AI was perhaps something that was under-forecasted. We are in the process of catching up.”
Su admits there is “supply tightness”, pinning it on demand spiking across the past few quarters and leaving little time for manufacturing and packaging to adjust. AMD says it is working with partners to ease bottlenecks and expects capacity to expand in the coming year, with its line-up aimed at training, inference and agentic workloads.
Intel has been singing a similar tune, admitting it failed to meet hyperscaler commitments because it lacked sufficient production capacity. Nvidia turning up with CPU-only commitments, including Vera, is another sign that AI infrastructure is diversifying and the humble server CPU is back in fashion.
