Compute that takes any shape
Vercel's Fluid compute now powers 15M daily builds, 25M weekly sandboxes, and 1 trillion monthly requests — and it's purpose-built for agents.
Why it matters
Vercel has shipped a unified compute layer (Fluid) that dynamically assembles machines for different workload shapes — builds, functions, sandboxes, and now agents. For practitioners deploying agentic systems, this eliminates the infrastructure provisioning friction that slowed earlier models; for enthusiasts, it's a signal that platform infrastructure is adapting faster to agent demand than traditional cloud VMs.
The key facts
7 to knowFluid now runs 15M builds/day, 25M sandboxes/week, 1 trillion requests/month
Hive provisions full VM in milliseconds with filesystem state attached
Fluid images (VHS format) resume rather than boot; custom machine ready in milliseconds
Vercel Drives decouple storage from compute; files persist across machine teardown/rebuild
Active CPU pricing: pay only while code executes, not during I/O or model waits
Single control-plane API across builds, functions, sandboxes, and agents
Agent-specific: untrusted code isolation + environment + persistent state, assembled on-demand
Go to the source
Vercel Blogvercel.com
Publisher excerpt: Workloads differ in how long they run, how much memory they take, and how much of the environment they control. That previously meant a different compute primitive for each job, so product iterations required provisioning infrastructure, not just working on a feature. We built Vercel so that none…