# How Fluid compute works on Vercel

By Steven Van · 2025-03-03

Vercel's post covers how its function router sends requests to warm instances and runs concurrent invocations to cut cold starts and idle compute.

Vercel has published an explanation of how [Vercel](<https://creatorstoolbox.com/tools/vercel>) Fluid compute works, with a focus on the Vercel Functions router. [The post](<https://vercel.com/blog/how-fluid-compute-works-on-vercel>) describes Fluid as a compute model aimed at requests that spend much of their time waiting on external models or APIs, where traditional serverless wastes compute.

The router sits at the centre of Fluid. It assigns requests to instances based on real-time metrics such as load, availability and type, and sends invocations to pre-warmed or active instances to cut cold starts. Fluid instances handle multiple concurrent requests, and capacity is added only when needed. Existing resources are reused before new ones are provisioned.

Instances do not accept direct traffic. They open persistent TCP tunnels to the router, receiving invocation data and streaming responses back over the same connection. Vercel says this design improves security and gives the router control over workload distribution and scaling.

[![](<https://assets.vercel.com/image/upload/q_auto/front/vercel/twitter-card.png>)](<https://vercel.com/fluid>)

![Vercel](<https://www.google.com/s2/favicons?domain=vercel.com&sz=128>)

Vercel

The platform for frontend developers — deploy, preview, and scale web apps and AI agents with zero config.

[View Vercel →](<https://creatorstoolbox.com/tools/vercel>)

[Original source](<https://vercel.com/fluid>)

[Read How Fluid compute works on Vercel on Creators Toolbox](<https://creatorstoolbox.com/blog/vercel-how-fluid-compute-works-on-vercel>)

---
Canonical source: https://creatorstoolbox.com/blog/vercel-how-fluid-compute-works-on-vercel