Vercel now deploys a plain Dockerfile as a container, no config needed
Since October 27, 2026, Vercel builds a root Dockerfile on its Container runtime when it finds no other framework. The runtime is in beta on all plans and billed by Active CPU.

Vercel said on October 27, 2026 that it now deploys a project that is just a Dockerfile or
Containerfile at the root. When Vercel detects no other framework, it picks its Container preset,
builds the image and sends all traffic to it. Container Runtime is in beta on all plans, including
the free Hobby plan, and is billed like other Vercel Functions.
What changed
Vercel’s Container Runtime already ran OCI images on Vercel Functions. The change is that a plain
Dockerfile is now enough, with no vercel.json setting. A few rules decide what gets built:
- If Vercel detects another framework, such as Next.js or Django, that framework wins.
- To force the container, run
vercel project update --framework containeror set"framework": "container"invercel.json. - A file named
Dockerfile.vercelorContainerfile.vercelalways selects the container. - Projects with several apps can list
servicesinvercel.json, each with its own Dockerfile.
The container must run an HTTP server. The default port is 80, and you can change it with a PORT
environment variable. Instances with no traffic scale down after 5 minutes in production and 30
seconds in preview, and get a 30-second SIGTERM grace period. Secure Compute and Static IPs don’t
work with custom images yet.
What it costs
Containers use the same limits and Active CPU pricing as Vercel Functions. You pay for CPU only while code runs, and for memory while an instance is handling work. Nothing is charged between requests.
| Item | Hobby | Pro |
|---|---|---|
| Active CPU | 4 hours included | $0.128/hr in Washington D.C., Cleveland and Portland; more elsewhere |
| Provisioned memory | 360 GB-hrs included | $0.0106/GB-hr in the same regions |
| Invocations | 1 million included | $0.60 per million |
Who it affects
Developers who keep a Dockerfile for Railway, Render, Fly.io or a VPS can now try the same image on Vercel without rewriting it. Because idle instances scale to zero, this suits request-driven apps more than always-on workers. For those, Vercel points to a separate product called Daemons.
What to do
Since the runtime is in beta, test with a preview deployment first, and check your usage page after a few days of traffic. To compare always-on servers with per-request billing, see our VPS and app hosting comparison.
Sources
Read on . Prices are in US dollars. How we check facts