Contents
In short
“One language for frontend and backend” is a strong pitch — but production Node.js has a bill: throughput, idle memory, and the npm supply chain. A Dev.to piece frames that as the full-stack tax.
What happened
The author contrasts 2026 expectations with JavaScript’s original scope. Cited benchmarks: Express around 20k RPS, Go ~40k, Rust ~60k; on AWS Lambda, Node ~20 ms average latency vs Rust ~1.1 ms for comparable work.
Idle memory: Node 30–50 MB, Go 5–10 MB, Rust 1–2 MB — at scale that drives instance count and cold starts.
Security: a huge npm registry, packages running with full process rights, recurring supply-chain incidents.
Why it matters
A monolingual stack reduces friction for small teams: shared types, hiring, less context switching. For CPU-bound work, tight SLAs, and minimal footprint you pay in hardware and risk, not just “learning a second language.”
Node remains strong for I/O-bound APIs, streaming, BFFs, and prototypes.
In practice
- Load-test your endpoints, not generic blog numbers.
- Offload CPU-heavy work to workers, queues, or Go/Rust services.
- Lockfiles, dependency audit, least privilege at runtime.
- Public APIs with stable contracts often need OpenAPI/Protobuf more than “all TypeScript.”
Takeaway
JavaScript on the server is a trade-off, not a mistake. Use Node where delivery speed wins; use a systems language where milliseconds and megabytes are the product.