Infrastructure

Cold start

The delay when a serverless function has to boot up from scratch before handling a request.

Because serverless functions shut down when idle, the first request after a quiet stretch has to wait while the platform spins up a fresh environment and loads your code — that's the cold start, often a few hundred milliseconds up to a couple of seconds. Requests that follow hit the now-'warm' instance and are fast. Analogy: turning the key on a cold engine versus one that's already idling. It bites hardest on low-traffic apps and heavy runtimes; lighter or edge runtimes shrink it, and some platforms let you pay to keep functions permanently warm.