Serverless API compute · Last verified 2026-08-25
Serverless API compute
How is a serverless invocation billed?
A billed invocation is a request plus a duration unit that is not the same across clouds. Lambda and Azure Functions use GB-seconds. Workers Paid uses requests and CPU-milliseconds plus a platform fee.
serverless lambda request
$0.2000 source
serverless lambda gb-second
$0.00001667 source
serverless functions request
$0.2000 source
serverless functions gb-second
$0.00001600 source
[01] How each vendor meters this
| Vendor | Unit | Free allowance | Tiers | Special rule | Source |
|---|---|---|---|---|---|
| AWS Lambda | Requests + GB-seconds | Monthly request and GB-second grants (1000000 million-requests) | On-demand flat in this snapshot | Provisioned concurrency omitted | Source |
| Azure Functions | Requests + GB-seconds | Consumption grants on the SKUs (1000000 million-requests) | Consumption flat in this snapshot | Premium plan omitted | Source |
| Cloudflare Workers | Requests + CPU-ms + platform fee | Included allotments on Paid (10000000 million-requests) | Paid platform fee is a monthly SKU | No extra Worker internet throughput meter | Source |
[02] Monthly total at the hub usage
Rank uses a published-SKU bundle on this hub and the same usage for every vendor. It is not an invoice.
| Vendor | Monthly total at the hub usage |
|---|---|
| Cloudflare | $0.00 |
| AWS | $1.80 |
| Azure | $1.80 |
| Google Cloud | $3.20 |
| SKU / Service | Unit price (USD) | Unit | Observed | Source |
|---|---|---|---|---|
| aws lambda request (us-east-1) | $0.2000 source | million-requests | 2026-08-25 | Source |
| aws lambda gb-second (us-east-1) | $0.00001667 source | GB-second | 2026-08-25 | Source |
| azure functions request (eastus) | $0.2000 source | million-requests | 2026-08-25 | Source |
| azure functions gb-second (eastus) | $0.00001600 source | GB-second | 2026-08-25 | Source |
| gcp cloud-run request (us-central1) | $0.4000 source | million-requests | 2026-08-25 | Source |
| gcp cloud-run vcpu-second (us-central1) | $0.00002400 source | vCPU-second | 2026-08-25 | Source |
| cloudflare workers request (global) | $0.3000 source | million-requests | 2026-08-25 | Source |
| cloudflare workers cpu-ms (global) | $0.02000 source | million-cpu-ms | 2026-08-25 | Source |
[Judgment]
Do not add Lambda GB-seconds to Workers CPU-milliseconds and call the sum a fair comparison. Lambda and Azure Functions Consumption multiply allocated memory by wall-clock; Cloud Run request-based billing splits vCPU-seconds and GiB-seconds; Workers Paid bills inbound requests and CPU time, not duration, and adds a monthly platform fee on the Paid plan with no extra charge for internet egress from the Worker itself. Idle fetch() time is expensive on Lambda and cheap on Workers. Provisioned concurrency and Cloud Run min instances are absent from this snapshot, so any always-warm estimate is a lower bound. Walk the same request count and duration through the serverless calculator and read the unit column. Sources: https://aws.amazon.com/lambda/pricing/ https://azure.microsoft.com/pricing/details/functions/ https://cloud.google.com/run/pricing https://developers.cloudflare.com/workers/platform/pricing/
Sources
Last verified 2026-08-25