AMD CEO Lisa Su revealed that server CPU demand has 'far exceeded' expectations, driven by the rapid growth of agentic AI applications. Speaking at the Morgan Stanley Technology Conference, Su explained that the CPU-to-GPU ratio in AI workloads has shifted dramatically, with top customers reporting that CPU demand alongside AI was 'under-forecasted.'

Supply constraints are now emerging as a result, with lead times stretching to eight to ten weeks for some orders. The sudden spike in commitments left AMD's supply chain with little time to adjust, though the company is working with partners to address bottlenecks. Su expressed confidence that AMD's product lineup can handle training, inference, and agentic workloads This demand surge mirrors broader industry trends, with Intel also reallocating wafer capacity from PC production to Xeon processors after hyperscalers ordered larger quantities than expected. Both Intel and AMD warned customers in China about tightening supplies in February, with Intel Xeon delivery times stretching to six months and prices rising over 10%. The situation has prompted AMD to reportedly prioritize Zen 6 chiplets for EPYC server processors, potentially delaying consumer Ryzen 'Olympic Ridge' desktop chips to 2027.