What is Vercel in October 2026?
Vercel is a managed hosting platform for frontend and full-stack web apps, built by the company that maintains Next.js. You connect a Git repository, it builds every push, serves the output from its CDN and runs server code as Vercel Functions. We use it as the default home for Next.js, and as a good one for Astro and headless CMS frontends that need server rendering or fast previews.
Functions run on several runtimes: Node.js, Bun, Python, Go, Ruby, Rust, Wasm, Edge and OCI containers. By default a function runs in one region (iad1), and Pro teams can pick more. The “edge” label matters less than this: a function far from its database is slow no matter how close the CDN is.
What do we deploy there?
- Next.js apps with the App Router, ISR, Server Components and middleware. Here Vercel saves the most setup time. Our Next.js team starts here unless a constraint says otherwise.
- Astro and other content sites that rebuild when an editor publishes. A webhook from the CMS triggers a build or an on-demand revalidation, so editors see changes without a developer. For static-heavy builds see our Astro services.
- Headless CMS frontends for Payload, Strapi, Contentful and Sanity, with preview mode wired to draft content.
- Marketing sites with auth or personalisation where middleware decides what a visitor sees.
How do builds and previews work?
Each pull request gets its own deployment and URL. Reviewers, designers and QA open the exact build that will ship, against the same environment variables as production unless you scope them differently. Environment variables are set per environment: production, preview and development. Deployment protection can put previews behind Vercel login, a password or SSO, which we turn on for any client work that is not public yet.
Promotion to production is a branch merge, and rollback is a one-click switch back to an earlier deployment, because old builds stay available. We keep the pipeline simple: Vercel builds, GitHub Actions runs tests and checks before merge, and a failed check blocks the merge instead of the deploy.
On a media platform we built a Vercel and AWS pipeline with per-branch previews and automated rollback gates, and the case file records a deploy time cut of about 70% and rollback under two minutes. The details are in the media platform pipeline case study.
Edge and serverless functions: what should run where?
Edge middleware is right for small decisions made before the page: redirects, rewrites, geo routing, cookie-based A/B splits and bot rules. It starts fast because it runs in isolates, but it has tight limits on what code and libraries it can use.
Serverless functions, now running with Fluid compute, are right for anything that talks to a database, calls a payment API or renders a page on demand. Put them in the region next to your database. Vercel’s docs list a read-only file system with 500 MB of writable /tmp, and a maximum duration per function. Long jobs belong in a queue or a container, not in a request.
What does Vercel cost, and how does the bill grow?
As of October 2026, the Vercel pricing page shows Hobby at $0 and Pro at $20 per month, with Enterprise on custom terms. Hobby is capped and cannot buy extra usage. Pro includes 1 TB of bandwidth and a $20 usage credit, and beyond that every meter is pay-as-you-go and uncapped: bandwidth from $0.15 per GB, build minutes from $0.007 per minute on the basic machine, plus active CPU time, memory and invocations for functions. CDN requests are a separate meter. New teams get a default spend budget of $200 that can pause projects when reached.
The bill rarely grows on seats. It grows on four things:
- Traffic that hits functions instead of the cache (server-rendered pages with no caching).
- Image optimisation and bandwidth on media-heavy pages.
- Build frequency, especially monorepos that rebuild every app on every push.
- Bot and crawler traffic on dynamic routes.
We cut cost by caching more routes with ISR, setting cache headers, ignoring builds for unchanged apps and setting spend alerts. If spend still outgrows the value, that is a signal to look at another host.
When do we pick another host?
- Mostly static site, many frameworks, form handling: Netlify is simpler and often cheaper. Our Vercel vs Netlify comparison covers the trade-offs.
- Very high traffic, mostly cacheable: Cloudflare Pages or Workers have no bandwidth charge on the paid Workers plan.
- Long-running jobs, WebSockets at scale, strict data residency: containers on AWS or a private cluster, which our DevOps support team runs.
How do we migrate to or from Vercel?
Moving in from Netlify or Cloudflare means translating redirects and headers into vercel.json or framework config, mapping functions to the right runtime, and copying environment variables per environment. We deploy the new project on a preview domain, compare responses route by route, then lower the DNS TTL and cut over.
Moving out is easier when the app avoids platform-only features. Next.js can run in a container or on another host through its standalone output, but ISR, image optimisation and middleware each need an equivalent. We list them before any decision. For ongoing platform work, you can hire a DevOps engineer from us, or start from our Cloud and DevOps services.
