What changed
AWS announced the next generation of its Amazon Bedrock AgentCore runtime. AgentCore runtime is the managed compute layer for running AI agents, letting developers deploy agents without building or maintaining infrastructure.
According to the company, sessions now start up to 15x faster, scale to 4x more compute and run up to 68% more efficiently. The core guarantee remains: every session runs in its own hardware-isolated Firecracker micro virtual machine (MicroVM) with its own identity, policy and audit trail.
Bigger workloads, shorter waits
Sessions now scale 4x over the previous version, up to 8 vCPUs and 32 GB, with support for container images up to 10 GB. AWS says sessions start in under 2 seconds regardless of image size and concurrency, so warm pools and pre-provisioned clusters are no longer needed to manage cold starts.
Why it matters
Agents no longer run one-off prompts; they execute multi-step workloads for hours, running code, navigating browsers and accessing sensitive data. Citing industry data, AWS says 40% of large enterprises are scaling AI agents, yet only 15% have scaled orchestrated multi-agent systems. The new version aims to balance security against speed and cost.
