FunctionPriceBook

What does a 1M requests/month API cost on serverless?

A small JSON API handling 1,000,000 requests a month, each running about 100 ms at 256 MB of memory. That is about 25,000 GB-seconds/month. After each platform's free tier, the cheapest of the GB-second-priced platforms is Google Cloud Run at roughly $0/month (compute + 5 GB egress). Most providers' free tiers swallow a workload this small entirely, so the realistic cost for many is $0 (plus any platform minimum). The figures below show the on-paper compute cost before free-tier credits.

Source: Provider pricing pages. Data as of June 2026.

Estimated monthly cost by platform

PlatformCompute (after free tier)Egress (5 GB)Total / month
Google Cloud Run$0varies$0+
Google Cloud Functions$0varies$0+
Azure Functions$0varies$0+
Azure Container Apps$0varies$0+
Scaleway Serverless Functions$0varies$0+
DigitalOcean Functions$0varies$0+
AWS Lambda$0$0.45$0.45

Source: Provider pricing pages. Data as of June 2026.

Assumptions: 1,000,000 requests/month, 256 MB memory, 100 ms duration, 5 GB egress. Totals are after each platform's free tier and exclude platform minimums and regional variation. "+" marks platforms whose egress rate varies. Snapshot June 2026 — verify on the vendor's pricing page.

How this is calculated

GB-seconds = requests x (duration/1000) x (memory/1024)
compute = max(0, requests - free) / 1M x req-fee + max(0, GB-s - free) x GB-s-rate
total = compute + egress GB x egress-rate

These are transparent calculations over the committed June 2026 snapshot rates — see the methodology. Change the numbers for your own workload in the calculator, and read the full breakdown on each provider page.

Frequently asked questions

What does a 1M requests/month API cost on serverless?

For 1,000,000 requests/month at 256 MB and 100 ms each (about 25,000 GB-seconds) plus 5 GB egress, the cheapest of the GB-second-priced platforms is Google Cloud Run at roughly $0/month after its free tier. Most providers' free tiers swallow a workload this small entirely, so the realistic cost for many is $0 (plus any platform minimum). The figures below show the on-paper compute cost before free-tier credits. snapshot captured June 2026.

What assumptions does this estimate make?

Compute = max(0, requests - free requests)/1M x per-1M-request fee + max(0, GB-seconds - free GB-seconds) x per-GB-second rate (CPU-metered platforms add a vCPU-second charge assuming 1 vCPU). Egress = 5 GB x the platform's egress rate. It excludes platform minimums (e.g. Cloudflare Workers Paid $5/mo, Vercel Pro $20/seat), cold-start nuances and regional price differences. See the methodology page.

Which platforms are not shown here?

Two groups are excluded. Resource-time platforms - AWS Fargate, Fly.io Machines, Render, Railway - bill by vCPU-hour, per-minute or instance-hour rather than per request, so a per-request figure is not comparable; for steady, high-utilisation traffic they can be cheaper, while for spiky low-volume traffic the request-priced platforms usually win. The others - Cloudflare Workers, Vercel Functions, Netlify Functions, Deno Deploy, Supabase Edge Functions, Modal - are request-priced but do not publish a clean per-GB-second rate (they bill CPU-time, CPU-ms, per-invocation or a multi-meter model), so we do not estimate their GB-second compute here; see each provider page for its own model.

More scenarios

Last updated: 2026-06-21