Serverless Edge vs Self-Hosted VPS: When Cloudflare Workers Cost More Than a Dedicated Server
Edge serverless computing has been marketed as the ultimate architecture for modern web applications. The promise of deploying code globally with sub-millisecond cold starts is enticing. However, when your application requires long-running database transactions, heavy cryptographic operations, or large file processing, the strict CPU execution limits and egress billing of edge functions quickly become a major bottleneck.
In this case study, we benchmarked a high-throughput API running on Cloudflare Workers (V8 isolates) against an asynchronous Python FastAPI service hosted on a single $40/month dedicated Linux server running Uvicorn and Gunicorn.
Benchmark Results (50 Million API Requests/Month)
- Cloudflare Workers (Paid Plan + Workers KV + Unbound CPU): $280 / month (with strict 30s timeout limits)
- Self-Hosted FastAPI (Hetzner Dedicated Server with Nginx caching): $42 / month (handling 12,000 req/sec with zero timeout constraints)
Frequently Asked Questions
When do Cloudflare Workers become more expensive than a VPS?
When processing millions of CPU-intensive requests, long-running WebSocket connections, or heavy payload parsing, Workers CPU billing quickly exceeds a flat-rate $40/mo dedicated server.
What is the main limitation of Cloudflare Workers for backend APIs?
Cloudflare Workers have strict memory limits (128MB) and synchronous CPU execution limits (30-50ms) that make heavy data processing unfeasible.