Vercel Edge is what Vercel calls running your code and serving cached content from its global network, right next to the visitor, rather than from one origin region. Simple enough. The catch is that a lot of what got written about it is now out of date. In June 2025 Vercel retired the standalone Edge Functions product, and the Edge runtime moved onto Vercel Functions under the fluid compute model. So here is the 2026 version: what Vercel Edge actually is, what that change did, and when the Edge runtime is still worth picking.
What is Vercel Edge?
Vercel Edge is an umbrella term, not a single product. Four things tend to get bundled under the name:
- The Edge Network: Vercel's CDN, the points of presence that cache your content and send each request to the nearest one.
- The Edge runtime: a light V8-isolate runtime, the same engine that powers Chrome, that runs your function code without spinning up a Node.js container.
- Edge Middleware: code that runs before a request reaches your page. Redirects, auth checks, rewrites, A/B splits.
- Edge Config: a low-latency key-value store you read from the edge, built for feature flags and redirect maps.
When someone says "Vercel Edge", they almost always mean the Edge runtime or Edge Middleware. Worth being precise, because only one item on that list actually changed in 2025.
Is Vercel Edge being discontinued?
No. The Edge runtime is alive and supported. What got deprecated in June 2025 was the standalone Edge Functions product, not the runtime itself. The Edge runtime now rides on Vercel Functions, which pulled all of Vercel's compute onto one pricing model and dropped the old hard wall between edge and serverless. So day to day, you pick a runtime per function, Edge or Node, and the bill behaves the same either way. Those threads asking whether Vercel is quietly killing the edge? That is the rename talking, not a shutdown. Vercel laid the move out in its Edge Middleware and Edge Functions changelog.
Vercel Edge runtime vs Node runtime
Reach for the Edge runtime when latency and startup time beat library support. Reach for Node when you need the full npm ecosystem or longer-running work. The trade-offs, plainly:
- Cold start: the Edge runtime wakes in roughly a millisecond, since a V8 isolate needs no container. Node is slower from cold, though fluid compute has closed a lot of that gap.
- APIs: Edge gives you web-standard APIs (fetch, Request, Response, Web Streams). Node hands you the whole Node surface, file system and native modules included.
- Packages: plenty of npm packages assume Node built-ins and simply will not boot on the Edge runtime. Anything that touches the file system or native bindings needs Node.
- Streaming: both stream fine, but the Edge runtime's low time-to-first-byte makes it the better default for streaming AI tokens.
- Reach: Edge code can run in many regions near your users. Heavy Node work usually stays in your primary region.
Want the same comparison across providers? See our edge compute directory and the Vercel Edge runtime profile.
What is Edge Middleware?
Edge Middleware is code that runs on the Edge Network before a request gets matched to a route. It is the right tool for geolocation redirects, auth gates, bot filtering, locale routing, and A/B tests, because it fires before the page renders and barely adds latency. One rule: keep it small. Middleware runs on every matching request, so anything heavy in there taxes every single page load.
What is Vercel Edge Config?
Edge Config is a globally replicated key-value store you read from the edge with almost no latency. It suits data you read constantly and write rarely. Feature flags, kill switches, redirect maps, a maintenance-mode toggle. It is not a database, though, so for anything you write often or query with filters, reach for a real serverless database instead.
When should you use Vercel Edge in 2026?
Use the Edge runtime when you are already on Vercel and the work fits its limits. A quick decision list:
- Use it for middleware, redirects, auth checks, streaming responses, and globally distributed reads that want low time-to-first-byte.
- Use the Node runtime instead for heavy dependencies, file-system work, long tasks, and anything that leans on the full npm ecosystem.
- Look at Cloudflare Workers if raw cost at scale or a bigger points-of-presence network matters most. Workers runs cheaper per request and in more places.
Choosing a host first and a runtime second? Our cloud hosting comparison and edge computing breakdown cover the wider field.
FAQ
What does "the edge" mean in Vercel?
The edge is Vercel's global network of points of presence, the locations sitting between your visitor and your origin server. Running at the edge means your code or cached content lives in those locations, close to the user, so a request gets answered with less round-trip time.
Is Vercel Edge free?
Yes, on the free Hobby tier for personal projects the Edge runtime is included. Paid usage kicks in on the Pro plan, around 20 US dollars per developer seat each month, plus usage charges. Since the 2025 change, edge and serverless compute sit under one billing model rather than two separate ones.
What is an edge request on Vercel?
An edge request is any request Vercel's Edge Network handles itself: a cache hit, a middleware run, or an Edge runtime function served from a point of presence near the user. Vercel counts these separately from origin function calls, because they happen on the network layer rather than at your origin.
Vercel Edge vs Cloudflare Workers: which should I pick?
Pick Vercel Edge if you already live on Vercel and want tight Next.js integration, one vendor, one bill. Pick Cloudflare Workers if cost at scale or a wider edge network matters more, since Workers runs cheaper per request and in more locations. Both sit on V8 isolates, so the underlying model is much the same.
The short version for 2026? Vercel Edge did not disappear. It got folded into Vercel Functions. The Edge runtime is still the quickest way to run middleware and stream responses on Vercel, the Node runtime handles everything that needs real packages, and the pick is now per function rather than per product.