Managed compute / CPU time vs wall time

Workers vs Lambda,
without false parity.

Describe the request once, but keep active CPU and wall-clock duration separate. Compare Workers, Lambda, and Cloud Run using the meter each platform publishes.

Use the calculator Methodology

03managed compute stacks

02separate duration inputs

100%source-linked rates

0server-side calculations

theserverless.dev

Generated: 09 Aug 2026
Prices checked: 2026-07-24

Cloud cost calculator · itemized list-price estimate

Managed compute cost comparison

Cloudflare Workers · AWS Lambda · Google Cloud Run

Lowest comparable monthly estimate$23.40Workers Paid

WorkloadSaaS APIRecurring allowances applied

Inputs and scope

The exact provider-neutral workload used for every estimate below.

Monthly requests: 50,000,000 / month
Active CPU time: 7 ms / request
Wall-clock duration: 35 ms / request
Working memory: 128 MiB
Inbound request: 2 KB / request
Public response: 25 KB / response
AWS request ingress: API Gateway HTTP API
Allowance view: Greenfield
Public delivery modeled: 1,250 GB / month

Monthly comparison

Totals are sums of the itemized list-price lines in each provider section.

Provider stack	Coverage	Top cost driver	Monthly total
Workers PaidCloudflare	Comparable	Worker requests	$23.40
Lambda x86 + HTTP APIAWS	Comparable	Internet egress (first 10 TB tier)	$163.30
Cloud Run request-basedGoogle Cloud	Comparable	Premium Tier egress (up to 1 TiB)	$273.73

03 · Itemized provider bill

Workers Paid

Global edge runtime; active CPU billing

Estimated total / month$23.40Comparable

Line item and source	Quantity	Pricing rule	Cost
Workers Paid account minimumM-CF-001 · checked 2026-07-24https://developers.cloudflare.com/workers/platform/pricing/	1 account-month	1 Workers Paid account minimumOne account-level minimum; not a per-application platform fee.	$5.00
Worker requestsM-CF-001 · checked 2026-07-24https://developers.cloudflare.com/workers/platform/pricing/	50,000,000 requests	40,000,000 requests after 10,000,000 account allowance	$12.00
Worker CPU timeM-CF-001 · checked 2026-07-24https://developers.cloudflare.com/workers/platform/pricing/	350,000,000 CPU-ms	320,000,000 active CPU-ms after 30,000,000 account allowanceWorkers bills active CPU, not wall-clock time spent waiting on I/O.	$6.40
Internet egressM-CF-003 · checked 2026-07-24https://developers.cloudflare.com/r2/pricing/	1,250 GB	1250.000 GB × $0 internet egressExplicit zero-dollar line: R2 internet egress is not charged.	$0.00
Monthly total			$23.40

Assumptions

Workers Paid request and active CPU tariffs.
Public response transfer is represented explicitly at $0.

Model notes and limits

shared allowance Greenfield view assumes recurring account-wide free allowances are unused. Promotional credits are excluded.

04 · Itemized provider bill

Lambda x86 + HTTP API

us-east-1 list price; wall-time compute billing

Estimated total / month$163.30Comparable

Line item and source	Quantity	Pricing rule	Cost
Lambda requestsM-AWS-LAMBDA-001 · checked 2026-07-24https://aws.amazon.com/lambda/pricing/	50,000,000 requests	49,000,000 invocations after 1,000,000 shared allowance	$9.80
Lambda x86 computeM-AWS-LAMBDA-001 · checked 2026-07-24https://aws.amazon.com/lambda/pricing/	218,750 GB-seconds	50,000,000 × 35 ms (1 ms rounded) × 0.125 GB; 400,000 GB-s allowanceLambda bills allocated memory for wall-clock execution, including I/O wait.	$0.00
API Gateway HTTP APIM-AWS-APIGW-001 · checked 2026-07-24https://aws.amazon.com/api-gateway/pricing/	50,000,000 metered requests	50,000,000 incoming requests × 1 request unit(s) at 512 KB eachThe 512 KB multiplier applies to inbound request payload, not response egress.	$50.00
API Gateway HTTP API (above 300M)M-AWS-APIGW-001 · checked 2026-07-24https://aws.amazon.com/api-gateway/pricing/	50,000,000 metered requests	0 metered requests above the first 300M	$0.00
Internet egress (first 10 TB tier)M-AWS-NET-001 · checked 2026-07-24https://aws.amazon.com/ec2/pricing/on-demand/	1,250 GB	1250.000 GB total − 100 GB shared allowance; 1150.000 GB in this tier	$103.50
Internet egress (next 40 TB tier)M-AWS-NET-001 · checked 2026-07-24https://aws.amazon.com/ec2/pricing/on-demand/	1,250 GB	0.000 GB in this tier	$0.00
Internet egress (next 100 TB tier)M-AWS-NET-001 · checked 2026-07-24https://aws.amazon.com/ec2/pricing/on-demand/	1,250 GB	0.000 GB in this tier	$0.00
Internet egress (next 350 TB tier)M-AWS-NET-001 · checked 2026-07-24https://aws.amazon.com/ec2/pricing/on-demand/	1,250 GB	0.000 GB in this tier	$0.00
Monthly total			$163.30

Assumptions

Lambda x86 in us-east-1, wall-clock duration with 1 ms rounding.
Endpoint selection and public response transfer are included; object storage is excluded.

Model notes and limits

shared allowance Greenfield view assumes recurring account-wide free allowances are unused. Promotional credits are excluded.

05 · Itemized provider bill

Cloud Run request-based

us-central1; concurrency 1, min instances 0

Estimated total / month$273.73Comparable

Line item and source	Quantity	Pricing rule	Cost
Cloud Run request CPUM-GCP-RUN-001 · checked 2026-07-24https://cloud.google.com/run/pricing	5,000,000 vCPU-seconds	50,000,000 × 0.1 s (100 ms rounded) × 1 vCPU; 180,000 vCPU-s allowanceRequest-based billing allocates CPU for the billable request lifetime.	$115.68
Cloud Run request memoryM-GCP-RUN-001 · checked 2026-07-24https://cloud.google.com/run/pricing	625,000 GiB-seconds	50,000,000 × 0.1 s × 0.125 GiB; 360,000 GiB-s allowance	$0.66
Cloud Run requestsM-GCP-RUN-001 · checked 2026-07-24https://cloud.google.com/run/pricing	50,000,000 requests	48,000,000 requests after 2,000,000 allowance	$19.20
Premium Tier egress (up to 1 TiB)M-GCP-NET-001 · checked 2026-07-24https://cloud.google.com/vpc/network-pricing	1,164.15 GiB	1164.153 GiB total − 1 GiB allowance; 1024.000 GiB in this tier	$122.88
Premium Tier egress (1–10 TiB)M-GCP-NET-001 · checked 2026-07-24https://cloud.google.com/vpc/network-pricing	1,164.15 GiB	139.153 GiB from 1–10 TiB	$15.31
Premium Tier egress (above 10 TiB)M-GCP-NET-001 · checked 2026-07-24https://cloud.google.com/vpc/network-pricing	1,164.15 GiB	0.000 GiB above 10 TiB	$0.00
Monthly total			$273.73

Assumptions

Cloud Run request-based billing in us-central1, concurrency 1 and min instances 0.
Public response transfer is included; Cloud Storage is excluded.

Model notes and limits

shared allowance Greenfield view assumes recurring account-wide free allowances are unused. Promotional credits are excluded.

Method and limitations

List-price estimate, not a quote or architecture recommendation. Taxes, support, observability, credits, negotiated discounts, request geography, migration, operations labor, and account eligibility can materially change a production bill.

Methodology: https://theserverless.dev/methodology
Pricing data: https://theserverless.dev/pricing-data

Why the bills diverge

7 ms CPU is not 35 ms wall time.

This request spends 5× as long end-to-end as it does actively executing code.

WorkersActive CPUI/O wait does not become CPU time.
LambdaMemory × wall timeEnd-to-end invocation duration is billed.
Cloud RunvCPU + memory × wallRequest-based, scale-to-zero reference.

01 / Workload

Describe one request

Monthly list price · USD

Monthly requestsProvider-neutral invocations.requests / monthActive CPU timeCode execution, excluding I/O wait.ms / requestWall-clock durationEnd-to-end time including I/O.ms / requestRequired working memoryProvider hard limits determine eligibility.MiBInbound requestAverage payload entering the endpoint.KB / requestPublic responseAverage payload sent to the internet.KB / response

Allowance view

GreenfieldRecurring allowances available.MarginalShared allowances already consumed.

AWS request ingress

HTTP APIAPI Gateway request fee included.Function URLNo API Gateway request fee.

03 / Audit the result

Workers Paid

Global edge runtime; active CPU billing

Modeled total$23.40Cloudflare · list price

Line item / source	Quantity	Pricing rule	Cost
Workers Paid account minimumM-CF-001 · checked 2026-07-24	1 account-month	1 Workers Paid account minimumOne account-level minimum; not a per-application platform fee.	$5.00
Worker requestsM-CF-001 · checked 2026-07-24	50M requests	40,000,000 requests after 10,000,000 account allowance	$12.00
Worker CPU timeM-CF-001 · checked 2026-07-24	350M CPU-ms	320,000,000 active CPU-ms after 30,000,000 account allowanceWorkers bills active CPU, not wall-clock time spent waiting on I/O.	$6.40
Internet egressM-CF-003 · checked 2026-07-24	1.25K GB	1250.000 GB × $0 internet egressExplicit zero-dollar line: R2 internet egress is not charged.	$0.00
Monthly total			$23.40

List-price model, not a quote. Taxes, support, observability, credits, negotiated discounts, request geography, and account eligibility can materially change a production bill.

Model the full architecture

The asymmetry that matters

I/O wait is time.
It is not always CPU.

An API may execute code for 5 ms and finish in 250 ms because it waits on a database. Treating those as one duration quietly favors whichever billing model the calculator was built for.

Compare with EC2 and a VPS

What is actually compared

One request.
Three published meters.

Request volume, payload, response size, and allowance mode stay constant. Only the provider-specific mapping changes.

01 / workers

Active CPU billing

Workers Paid request and active CPU tariffs are applied. The account minimum, included amounts, and zero-dollar response transfer stay explicit.

02 / lambda

Memory × wall time

x86 Lambda in us-east-1 uses configured memory and wall-clock duration. Choose an HTTP API or Function URL for request ingress.

03 / cloud run

vCPU + memory × wall

Request-based Cloud Run in us-central1 is modeled at concurrency 1 with scale-to-zero. Public transfer stays a separate line.

Included

Monthly request volume
Active CPU and wall duration
Configured memory and payload
HTTP API or Function URL ingress

Not invented

Cold-start rate or synthetic latency
Database and queue charges
Reserved capacity and discounts
Logging, support, and operations labor

Model boundaries

Workers vs Lambda calculator FAQ

The live bill links every rate and methodology rule. These answers explain how to enter an honest workload.

01 Why do Workers and Lambda need different duration inputs?

Workers bills active CPU time, while Lambda bills configured memory multiplied by wall-clock invocation duration. A request waiting on a database can therefore have much less CPU time than wall time.

02 Does the AWS total include API Gateway?

You choose. HTTP API includes API Gateway HTTP API requests. Function URL removes that line. Both options retain Lambda request, compute, and direct internet-transfer charges.

03 Why include Google Cloud Run?

Cloud Run is a credible managed-container alternative. This model prices a narrow request-based configuration: us-central1, concurrency 1, minimum instances 0, and entered duration including weighted startup overhead.

04 Does the calculator predict cold starts or latency?

No. Enter observed or expected weighted wall duration, including startup overhead when material. The calculator maps a workload to published billing meters; it does not invent a benchmark for your code.

Workers vs Lambda,without false parity.

Managed compute cost comparison

Assumptions

Model notes and limits

Assumptions

Model notes and limits

Assumptions

Model notes and limits

Method and limitations

7 ms CPU is not 35 ms wall time.

I/O wait is time.It is not always CPU.

Active CPU billing

Memory × wall time

vCPU + memory × wall

Workers vs Lambda calculator FAQ

Workers vs Lambda,
without false parity.

I/O wait is time.
It is not always CPU.