# AWS adds faster, larger, cheaper AgentCore runtime for AI agents

> Amazon Bedrock AgentCore now starts sessions up to 15× faster, supports 4× more compute, and reduces memory use by up to 68%, with pay‑as‑you‑go pricing.

Oossa · 2026-11-12 · https://oossa.com/en/aws-adds-faster-larger-cheaper-agentcore-runtime-for-ai-agents

AWS announced today that the next‑generation AgentCore runtime for Amazon Bedrock can launch an isolated AI‑agent session in under two seconds, scale to 8 vCPUs and 32 GB of memory, and reclaim unused memory within minutes. The runtime runs each session in a Firecracker micro‑VM for hardware isolation and now lets agents mount an S3 bucket so data stays inside the same security boundary. Customers say the new version cuts cold‑start latency from a minute to a few seconds and lowers per‑call costs by around 17%.

## The facts

- Sessions start in under 2 seconds, up to 15× faster than the prior version.
- Memory consumption per session drops up to 68% and compute can scale 4× (up to 8 vCPUs, 32 GB).

## Why it matters

Enterprises can move AI agents into production faster and at lower cost while keeping data securely isolated.

## Sources & references

1. [Run agents securely and efficiently with the next generation AgentCore runtime](https://aws.amazon.com/blogs/machine-learning/run-agents-securely-and-efficiently-with-the-next-generation-agentcore-runtime/) – AWS, 2026-11-12

Last updated: 2026-11-12