Serverless
A cloud model where the provider runs your code on demand without you managing servers.
Last reviewed: July 25, 2026
What is serverless?
Serverless (often FaaS — Functions as a Service) means you upload functions or containers, and the cloud provider scales, patches, and bills based on invocations and duration rather than reserved VMs.
Characteristics
- No server management — no SSH into boxes
- Event-driven — HTTP requests, queue messages, cron schedules, file uploads
- Scale to zero — pay nothing when idle (aside from storage)
- Short-lived — functions have timeout limits (seconds to minutes)
Examples
AWS Lambda, Google Cloud Functions, Azure Functions, Cloudflare Workers, Vercel Functions.
Cold starts
When a function hasn’t run recently, the platform may need to cold start — provision runtime, load code. Latency-sensitive apps mitigate with provisioned concurrency or always-warm tiers.
Serverless vs containers vs VMs
| Model | Control | Ops burden | Best for |
|---|---|---|---|
| VM | Full | Highest | Legacy, stateful systems |
| Container | High | Medium | Microservices, long-running |
| Serverless | Low | Lowest | Spiky, small units of work |
Limitations
Execution time caps, stateless design expectations, vendor-specific integrations, and debugging distributed traces across many tiny functions.
Cost reality
Brilliant at low or spiky volume, brutal at sustained high volume: per-invocation pricing crosses over container pricing at roughly sustained double-digit requests/second — do the math before growth forces a migration under load. Watch the sidecar costs too: API Gateway fees, per-invocation logging, and NAT charges often exceed the function compute itself.
What people get wrong
- Long chains of functions calling functions. Each hop pays latency and cold-start risk; step-function orchestration or a plain service is often simpler.
- Opening database connections per invocation. Traditional connection pools break under function fan-out; use provider data-API proxies or connection poolers.
- Ignoring the event model lock-in. Containers move between clouds easily; code written against a provider’s triggers and services is the stickiest lock-in in compute.
See the full decision framework in Serverless vs Containers.
What “Serverless” Actually Means
Despite the name, serverless computing still runs on physical servers — the term refers to the fact that the application developer never provisions, patches, or manages those servers directly, with the cloud provider handling all underlying infrastructure and automatically scaling capacity up or down (including to zero) based on incoming demand. This model shifts the pricing structure fundamentally: rather than paying for provisioned capacity that sits idle during low-traffic periods, serverless billing is typically based on actual execution time and resource consumption, which can be dramatically more cost-effective for spiky or unpredictable workloads, though potentially more expensive than provisioned infrastructure for consistently high, predictable traffic where the overhead-free efficiency of always-on servers wins out.
Historical figures and technical concepts for informational purposes only. Not technical, professional, legal, or financial advice. Sources: Official Documentation.