Serverless at the Edge: When It Makes Sense (and When It Doesn't)

The Latency Map Matters More Than the Framework

Edge compute wins where the work is close to the user and cheap to repeat. It loses where the request has to reach a central database anyway — in that case you have only added a hop.

Where It Earns Its Place

The pattern across every successful deployment we have measured is the same: small, stateless, and on the critical first paint.

• Auth checks and redirects that would otherwise block the document
• A/B variants and personalisation assembled before HTML lands
• Aggregating two or three APIs into one response near the caller
• Rewriting and caching third-party content you cannot control
• Webhooks that acknowledge fast and queue the real work

Where It Costs You

The trade-offs are real and mostly show up months later, when the runtime limits become architectural ones.

• Cold-start behaviour differs per runtime, so benchmarks do not transfer
• Local parity is weaker than a container you can run on your laptop
• CPU-time and payload caps push logic back into the origin
• Observability across regions is harder than in one cluster
• Vendor lock-in creeps in through platform-specific APIs