Scale to one: How Fluid solves cold starts
By Steven Van ·
Vercel says Fluid compute now prevents cold starts on 99.37% of requests and softens the ones that still occur.
Vercel says Vercel's Fluid compute now avoids cold starts on 99.37% of all requests, and that the rare ones which still happen are shorter and faster than on a typical serverless platform. The original Fluid compute announcement lays out how this works, from four angles:
- Prevention: a scale-to-one model keeps an instance warm before it's needed, instead of shutting it down after a few idle minutes.
- Frequency reduction: Fluid reuses existing instances to serve multiple requests rather than spinning up a new one each time.
- Anticipation: predictive scaling warms instances ahead of expected demand.
- Impact reduction: when a cold start does happen, a Rust-based runtime with Node.js and Python support speeds up initialization, and bytecode caching pre-compiles function code to cut startup overhead.
Vercel
The platform for frontend developers — deploy, preview, and scale web apps and AI agents with zero config.
View Vercel →