AWS tells engineers to cut CPU waste as agentic AI strains capacity

Amazon Web Services is pushing internal engineers to reduce low-utilization EC2 instance usage as it struggles to meet CPU capacity demand from external customers. The company points to intensifying agentic AI workloads as a driver.

Why it matters: This is a concrete sign the AI capacity crunch has moved beyond GPUs to ordinary CPU compute, since agent workloads run far more inference and orchestration calls than typical applications. Expect cloud providers to keep tightening internal usage, and possibly pricing, as agentic workloads scale further.