AWS orders engineers to cut CPU waste as AI demand strains EC2 capacity

Amazon Web Services has issued internal directives requiring its engineering teams to reduce CPU waste and clean up under‑utilized EC2 instances. The move follows internal meetings in May that highlighted a surge in compute demand driven by AI workloads, which is outpacing available CPU capacity. Engineers now face multi‑day delays when requesting new CPU resources, a shift from the previous hours‑long provisioning.

AWS’s efficiency plan calls for rightsizing instance types, automated shutdowns of idle machines, and broader resource‑management improvements. The pressure on CPU capacity mirrors similar constraints reported by Microsoft Azure and Google Cloud, though AWS’s focus on CPUs is a newer development. The company continues to invest in custom silicon such as Graviton processors and Trainium AI chips, but the long lead times for chip fabrication mean additional capacity will arrive only after several quarters.

For cloud customers, the internal strain could translate into reduced availability of certain EC2 instance types, higher spot‑instance pricing, and longer provisioning times, underscoring the broader impact of AI‑driven compute demand on the cloud industry.