AWS makes its faster AgentCore runtime the default for new AI agents
AWS's next-generation AgentCore Runtime is now the default for new agents in 19 regions: sessions start in under 2 seconds, scale to 32 GB, and bill for less idle memory. Existing agents must be moved by hand.
- Author
Umair Mateen- Published
- Reading time
- 4 min
- Category
- AI
Ideas in this post
Tap one for a 5-second answer. Underlined words in the article work the same way.
0 of 4 understood

Amazon Web Services has made its next-generation AgentCore Runtime the default for new runtimes in 19 regions, announced on 12 November 2026. AgentCore Runtime is the part of Amazon Bedrock AgentCore that hosts and runs your AI agents; the new version starts sessions much faster, gives each one more computing power and, according to AWS, bills for less memory.
What actually changed
AgentCore Runtime runs each agent session (one user's conversation or task) in its own small, hardware-isolated virtual machine, so one session cannot see another's data. According to the AWS announcement and the AWS Machine Learning blog post published the same day, the new version brings:
- Faster starts. AWS says sessions on a 2 GB container image now start in under 2 seconds at the 75th percentile (three in four starts are at least that fast), down from 29.7 seconds on the previous version. Its blog headline puts this at up to 15 times faster.
- Bigger sessions. Each session can scale to 8 vCPUs and 32 GB of memory, with container images up to 10 GB. AWS describes this as four times the previous version.
- Lower memory bills. Memory is handed back within minutes once an agent stops using it, and idle agents scale to zero. AWS says multi-turn agents can see up to 68% less billed memory per session.
- A predictable pricing option. The blog describes "committed baseline" pricing: you pay for a minimum amount of memory per session, can burst to four times that baseline when needed, and steady workloads are charged at AWS Lambda MicroVM rates. AWS does not list the actual rates in either post; they are on the AgentCore pricing page.
- Per-session storage. Agents can mount and save to a file system such as an Amazon S3 bucket. AWS also says that by its re:Invent conference, each session will be tied to its own storage folder; that part has not shipped yet.
Two customers are quoted in the blog with their own early numbers: Mission reports 35% faster cold starts for inference requests and 17% lower measured runtime cost per call, and Zuora reports cutting memory per invocation by around 60% in many cases. These are the customers' own figures, not independent tests.
Quick check · Myth or fact
AI agents: myth or fact?
Decide for yourself first, then tap each statement to check it.
0 of 3 checked
From the explainer What is an AI agent? Explained with an assistant you send on errands
Who should care
Teams already running agents on Amazon Bedrock AgentCore, and teams choosing where to host customer-facing agents. A slow first response is one of the most visible problems in a chat or : a user waiting half a minute for an agent to wake up often leaves. Faster starts and per-session isolation also matter to companies in regulated sectors that need to show each customer's session is kept apart.
Finance leads should note the billing shift. Paying only for memory an agent is actually using can cut costs for agents that sit idle between turns, while the committed baseline gives a steadier monthly bill for busy, always-on agents.
Limits and caveats
- Existing runtimes do not move on their own. AWS says "existing runtimes stay on the current version until you move them." To upgrade, you set the platform version to V2 when you create or update a runtime. To keep a new runtime on the old version, set it to V1.
- Region list. The new version is the default in 19 regions across North America, South America, Europe and Asia Pacific, including Frankfurt, Ireland, London, Paris and Stockholm in Europe. In other regions you can opt in with V2, according to the announcement.
- No end date for the old version. Neither post gives a retirement date for the previous runtime.
- Vendor figures. The speed and memory numbers are AWS's own measurements and depend on your container image and workload. Test with your own agent before relying on them.
What to do this week
- If you run agents on AgentCore, list your runtimes and which version each uses. New ones in the 19 default regions are already on V2; older ones are not.
- Pick one non-critical agent, update it to V2 in a test environment, and measure first-response time and memory use against the current version.
- Check the pricing page and compare pay-per-use against committed baseline for your busiest agent.
- If you handle customer data, confirm your needs are met in the region you use, and watch for the per-session storage folders AWS has promised.
The DevDrift helps teams put AI agents into real workflows with approvals and , including on AWS.
Sources
You’ve read the news. Now understand it.
- Learn · 5 min · Myth or factWhat is an AI agent? Explained with an assistant you send on errandsAn AI agent takes a goal, plans the steps and carries them out with software tools, checking each result as it goes. Explained with an assistant you send on errands.Start in 5 seconds →
- Learn · 5 min · Side by sideWhat is a voice agent? Explained with a receptionist who never puts you on holdA voice agent is an AI you talk to out loud that listens, checks your systems and answers in a spoken voice. Explained with a front-desk receptionist who answers every call straight away.Start in 5 seconds →
- Learn · 5 min · Myth or factWhat is data residency? Explained with picking a warehouse's countryData residency means choosing which country or region your data is stored and worked on in. Explained with a storage company that has warehouses in many countries, and lets you pick one.Start in 5 seconds →
- Learn · 5 min · Myth or factWhat is an audit trail? Explained with a visitor logbookAn audit trail is a time-ordered record of who did what in a system, when and with what result. Explained with the visitor logbook at an office reception desk.Start in 5 seconds →


