Johnny UnarNot interchangeable
Serverless functions (AWS Lambda, Google Cloud Functions) and edge functions (Cloudflare Workers, Vercel Edge Functions) both let you run code without managing servers. But they're not interchangeable. Serverless runs in specific regions with cold starts measured in hundreds of milliseconds. Edge runs on a global network with cold starts under 10ms. The trade-off: edge functions have stricter runtime constraints — smaller memory limits, no native Node.js APIs, and limited execution time.
When to go edge
Use edge functions for anything latency-sensitive that doesn't need heavy computation: authentication checks, A/B test routing, geolocation-based redirects, header manipulation, and personalized content delivery. A user in Tokyo hitting an edge function gets a response in 20ms instead of 200ms from a US-East Lambda. For global products, that difference is the gap between 'fast' and 'instant.'
When to go serverless
Use serverless for compute-heavy workloads: image processing, PDF generation, database migrations, complex API orchestration, and anything that needs full Node.js compatibility. Serverless functions can run for up to 15 minutes, use up to 10GB of memory, and access any npm package. The cold start penalty matters less for background jobs and API endpoints that aren't user-facing.
The hybrid approach
The smartest architecture combines both. Edge functions handle the request layer — auth, routing, caching, personalization — and proxy to serverless functions for heavy lifting. Next.js middleware runs at the edge by default, while API routes can be configured as either edge or serverless per-route. This gives you global latency for the user-facing layer and full compute power for the backend. At steezr, this hybrid approach is our default for any project that serves users across multiple regions.
