Lambda, Cloud Run, Vercel Functions. When pay-per-request is cheaper than a server — and when it isn't.
Category · DevOps & Infra
No server, just function.
Serverless doesn't mean no server is running — it means you don't have to look after it. You upload code — AWS Lambda, Google Cloud Run, Vercel Functions — and the provider starts it per request.
You're billed only for actual execution. No request comes in, nothing runs and nothing costs.
When pay-per-request wins.
For sporadic or sharply fluctuating load, serverless is often the cheapest and simplest choice: webhooks, background jobs, event processing, endpoints with irregular traffic.
We use it where a dedicated machine would mostly just burn electricity — scaling from zero to many requests is handled by the platform.
When a server is cheaper.
Under constant high load the maths tips over: a continuously used server is then far cheaper per request than millions of individually billed calls a month.
Add cold starts on the first call and hard runtime limits. Long-running, memory-hungry or latency-critical processes fit the serverless model poorly — here the load profile decides in the end, not the trend.
