loading
loading
You ship functions and the platform runs, scales, and bills them per-request — no servers to manage.
There are still servers; you just don't provision or maintain them. The platform spins instances up on demand, scales to zero when idle, and charges per invocation and compute-millisecond, which is why it's cheap for spiky traffic and wasteful for constant high load. The catch is cold starts, execution time limits, and statelessness — you keep state in a DB or cache, not in memory. Most LLM apps start serverless because traffic is bursty and you'd rather not babysit a GPU box.
Plainly
Think of Serverless as roads and power for the app city. You ship functions and the platform runs, scales, and bills them per-request — no servers to manage.
In practice
Use it when local behavior needs to become a reachable, reliable deployed service. In practice, define the owner, input, output, and failure mode before you rely on it.