Serverless function pricing is rarely just a request fee. A bill can combine requests, runtime, memory or CPU, plan fees, and usage allowances. Free tiers can dominate at low volume. Execution and overage charges matter more as traffic grows.
The Serverless Functions Pricing Comparison gives each provider the same workloads at the same request volumes. Plan floors, included usage, and metered execution stay separate so you can see what drives the monthly cost.
Benchmark workloads
| Workload | Duration | Active CPU | Memory | Request | Response |
|---|---|---|---|---|---|
| Short API requests | 100 ms | 10 ms | 64 MiB | 1 KiB | 1 KiB |
| CPU-intensive functions | 1 s | 1 s | 512 MiB | 1 KiB | 1 KiB |
Short API requests put more weight on request charges, allowances, and plan floors. CPU-intensive functions put more weight on sustained CPU, memory allocation, and product limits.
Each workload runs at 1M, 10M, and 100M synchronous HTTP invocations per calendar month. These are comparison workloads, not estimates of a typical customer’s traffic.
Pricing scope
The comparison uses the current public products and configurations listed in the source table below. Regional products use
aligned eastern-US reference locations: AWS us-east-1, Azure East US 2, Google us-east4, and Vercel iad1. Global
products use published global pricing. This aligns rate scope, not infrastructure or latency.
Prices use USD and public pay-as-you-go pricing or the lowest eligible public self-serve plan. Monthly allowances and plan credits are treated as available to the modeled workload at the start of the billing period. Taxes, unrelated account usage, negotiated pricing, enterprise contracts, annual-prepay discounts, promotions, and commitment discounts are excluded.
How prices are calculated
Each provider keeps its published billing model. The calculation applies allocation and minimum-duration rules, provider-native meters, included allowances, pricing tiers or packages, plan selection, credits, and required rounding or aggregation in the order documented by the provider.
Requests, executions, CPU time, memory over duration, wall duration, packages, fixed plans, and credits remain separate. They are not collapsed into a synthetic universal compute unit.
Some providers require a paid plan before usage charges begin. We count that plan in the monthly total even when an existing customer might already subscribe to it. This keeps the comparison on a consistent new-customer basis. An exact estimate includes:
- Fixed monthly spend: the recurring self-serve amount required before workload-driven usage or overage.
- Incremental usage: scenario-dependent spend after applicable allowances, credits, plan rules, and other usage charges.
The comparison shows the total first, followed by the selected plan and fixed-versus-usage breakdown where applicable.
When a price is shown
We show a price only when the reviewed public terms support one exact monthly total for the workload. Missing quantities, conflicting billing rules, or product limits are not replaced with guessed prices or ranges.
Required adjacent services count only when the workload defines their usage. Azure Flex Consumption requires a Storage account. The workload defines no Storage usage, so Azure shows a runtime-only estimate and calls out the Storage requirement separately. Optional egress, databases, queues, observability, and other adjacent services stay outside the headline estimate unless the workload requires them.
If a workload exceeds a reviewed memory or runtime limit, we do not invent an estimate to keep the table complete.
Official pricing sources
Each row fixes the product, plan, configuration, and region used for the comparison. Pricing, configuration, and service-limit evidence comes from the official sources below. The comparison page shows the current pricing verification date.
| Provider | Product / plan | Configuration / SKU | Region | Official sources |
|---|---|---|---|---|
AWS Lambda | LambdaOn-Demand | x86 | us-east-1 | PricingMemory and CPU allocation |
Azure Flex Consumption | Flex ConsumptionOn Demand | FC1512 / 2,048 / 4,096 MBRuntime only | East US 2 | Retail prices API · Regional availabilitySKU · Flex planStorage requirement |
Cloudflare Workers | Workers StandardPaid | 128 MB per-isolateMemory cap | Global | PricingLimits |
Deno Deploy | Deno DeployFree / Pro | 768 MB default4,096 MB documented Pro upper | Global | PricingRuntime configuration |
Fastly Compute | ComputeUsage-based | 128 MBHeap cap | Global | PricingBilling modelResource limits |
Google Cloud Run functions | Cloud Run functionsRequest-based | x86CPU selected by memory/workload | us-east4 | Cloud Run pricingCPU configuration |
Netlify Functions | Credit-based FunctionsSelf-serve | 1 GB | Global | Usage metersCredit-plan ladder |
Supabase Edge Functions | Edge FunctionsPro | 256 MB memory cap2 s CPU cap | Global | PricingLimits |
Vercel Functions | Fluid computePro | 2 GB1 vCPU | iad1 | PricingMemory configuration |
What’s excluded
These prices do not score cold-start performance, latency, developer experience, runtime quality, or custom optimization. Private enterprise terms are excluded. GPU inference and products without enough comparable public pricing or workload support do not receive synthetic function-price estimates.