Serverless grows up

Serverless earned a reputation early for cold starts, vendor lock-in, and debugging pain. Much of that critique is now out of date. Runtimes are faster, tooling is mature, and the operational savings are real.

The strongest case is spiky, event-driven work: webhooks, scheduled jobs, image processing, and APIs with unpredictable traffic. Paying only for execution time — and never for idle servers at 3 a.m. — is a genuine advantage.

It is still not a universal answer. Long-running processes, heavy stateful workloads, and latency floors below a few milliseconds are better served elsewhere. Maturity means knowing the boundary, not evangelizing past it.

For most product teams, the right architecture in 2026 is a pragmatic mix: serverless for the spiky edges, long-lived services for the hot core, and a clear reason for each choice.