Vercel introduces Fluid compute to solve serverless cold starts through a multi-layered approach including scale-to-one (keeping at least one instance warm), instance reuse for concurrent requests, predictive scaling, bytecode caching, and rolling releases. This combination achieves zero cold starts for 99.37% of requests,
Table of contents
What a cold start is and why it mattersHow Vercel prevents cold starts: The complete solution stackScale to one: Eliminating first visitor cold startsFluid compute: Reuse what's warmPredictive scaling: Warming ahead of demandBytecode caching: Start faster when you mustRolling releases: Preventing deployment-induced cold startsCold starts solved at every scaleSort: