FunctionPriceBook

Serverless platforms that scale to zero

15 of the 17 platforms we track scale to zero — they cost $0 for compute while idle and bill only for actual usage. That includes AWS Lambda, Cloudflare Workers, Google Cloud Run/Functions, Azure Functions/Container Apps, Vercel, Netlify, Deno Deploy, Supabase, Modal, Scaleway, DigitalOcean Functions, Render (free) and Fly.io Machines. The ones that keep billing while running are AWS Fargate and Railway (resource-time billing). The trade-off for scale-to-zero is cold-start latency on the first request after idle.

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

Scale-to-zero by platform

PlatformScale to zero?Pricing modelIdle / free behaviour
AWS LambdaYesFaaS (request + GB-second)1M requests + 400,000 GB-s every month, forever (not just 12 months).
Azure Container AppsYesServerless containers (consumption)Consumption plan free grant: 180,000 vCPU-s + 360,000 GiB-s + 2M requests/month per subscription.
Azure FunctionsYesFaaS (Consumption / Flex Consumption)Consumption plan: 1M executions + 400,000 GB-s/month free. Flex Consumption: 250,000 executions + 100,000 GB-s/month free.
Cloudflare WorkersYesEdge FaaS (request + CPU-ms)Free plan: 100,000 requests/day (~3M/month) with up to 10 ms CPU per invocation.
Deno DeployYesEdge FaaS (requests + CPU-time)Free tier: 1M requests/month + 20 GB egress + 15 CPU-hours/month.
DigitalOcean FunctionsYesFaaS (GB-second only, no per-request fee)90,000 GiB-seconds (25 GiB-hours) of compute per month free; no per-invocation charge.
Fly.io MachinesYesMicro-VMs (per-second, scale-to-zero)No free tier (the old free allowance ended in 2024). A shared-cpu-1x/256 MB Machine is ~$1.94/mo always-on; $0 when stopped.
Google Cloud FunctionsYesFaaS (now on Cloud Run model)2M invocations + 400,000 GB-s + 200,000 GHz-s + 5 GiB egress/month (1st-gen allowance); 2nd-gen runs on Cloud Run free tier.
Google Cloud RunYesServerless containers (vCPU-s + GiB-s + request)180,000 vCPU-s + 360,000 GiB-s + 2M requests/month free (Tier-1 regions: us-central1/us-east1/us-west1).
ModalYesServerless compute (per CPU-s + GiB-s, GPU)$30/month free compute credits on every account (not a perpetual free runtime).
Netlify FunctionsYesFaaS (invocations + GB-hour, credit-based)Free plan: 300 monthly credits ≈ 125,000 function invocations + ~30 GB-hours of compute.
RenderYesWeb services / containers (instance-hour)Free web service: 750 instance-hours/month, 512 MB / 0.1 CPU, spins down after 15 min idle (30-60 s cold start).
Scaleway Serverless FunctionsYesFaaS + containers (EU, request + GB-s)Functions: 1M requests + 400,000 GB-s/month free. Containers: 400,000 GB-s + 200,000 vCPU-s/month free.
Supabase Edge FunctionsYesEdge FaaS (per invocation, Deno)Free plan: 500,000 Edge Function invocations/month.
Vercel FunctionsYesFaaS / Fluid compute (invocation + active CPU + memory)Hobby plan (free): 1M invocations, 4 CPU-hours active compute, 360 GB-hours memory, 100 GB bandwidth/month (hard caps, no overage).
AWS FargateNoServerless containers (vCPU-hour + GB-hour)No free tier; billed per second (1-minute minimum) while tasks run.
RailwayNoUsage-based containers (per-minute vCPU + RAM)No always-free runtime; Hobby plan is $5/mo and includes $5 of usage. Trial credit for new accounts.

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

Snapshot June 2026 — verify on the vendor's pricing page.

Frequently asked questions

Which serverless platforms scale to zero?

Most FaaS and serverless-container platforms scale to zero: AWS Lambda, Cloudflare Workers, Google Cloud Run and Functions, Azure Functions and Container Apps, Vercel, Netlify, Deno Deploy, Supabase Edge Functions, Modal, Scaleway, DigitalOcean Functions, Render (free tier) and Fly.io Machines. The exceptions that keep billing while running are AWS Fargate and Railway (they bill resource-time, not per request). Snapshot captured June 2026.

Does scale-to-zero mean it is completely free when idle?

For compute, yes — a scaled-to-zero function or container costs $0 while it has no traffic. But fixed costs can remain: a platform minimum (Cloudflare Workers Paid $5/mo, Vercel Pro $20/seat), attached storage/volumes (Fly.io volumes still bill at $0.15/GB-mo when the Machine is stopped), or a managed database. Scale-to-zero saves on compute, not on everything.

What is the downside of scale-to-zero?

Cold starts. When a platform has scaled to zero, the first request after idle has to spin up an instance, adding latency — milliseconds for V8-isolate edge runtimes (Cloudflare Workers, Deno Deploy), a second or two for Lambda/Cloud Run containers, and up to 30-60 seconds for Render's free tier. For latency-sensitive apps you may pay to keep a minimum instance warm.

Keep exploring

Last updated: 2026-06-21