Amazon's New AI Engine Cuts the Cost of Always-On Assistants
Cheaper AI that never sleeps — and you only pay for what it actually does.
Deep Dive
The new Amazon Bedrock AgentCore runtime is here — a managed compute layer for agents that process claims, write and review code, coordinate across systems, and run for hours without supervision. It reclaims memory as a session releases it instead of holding it at the peak, and releases memory back the instant a session ends, so a long-running or bursty agent isn't paying for its high point the whole time it runs. Cold start times stay consistent regardless of container size
Key Points
- Amazon's new AgentCore runtime runs AI agents — software that does tasks by itself — and frees up memory the moment a session ends instead of charging for peak usage.
- New sessions start in under 100 milliseconds, the same speed whether the agent is small or handling many tasks at once.
- You pay only for actual use and idle agents cost nothing, making always-on assistants affordable for companies to deploy.
Why It Matters
Cheaper, faster AI agents let businesses automate more routine work — which can cut costs and change who does what.