Pricing ledger · model input, not marketing copy
Every modeled rate, with its source attached.
The calculator starts from a checked-in price book, then loads the latest published normalized rates from the same ledger shown here. Ordinary changes publish daily; anomalous changes wait for an explicit decision.
Loading the latest automated source-check status…
Five comparison stacks
Price-book index
All figures are USD public list-price references. Regions and stack boundaries are part of the model, not footnotes. A verification date advances after a successful source check; a rate version advances only when a published value changes.
Workers Paid + R2 Standard
Cloudflare
Active Worker CPU time, requests, object storage operations and direct Internet delivery.
workers.paid_base workers.requests workers.cpu_ms r2.class_a r2.class_b r2.egress Stack assumptions
- Workers compute uses active CPU milliseconds rather than end-to-end wall time.
- R2 Standard is the only storage class in this reference stack.
- Direct R2 Internet egress remains an explicit zero-price line item.
- Recurring included usage is applied only in greenfield allowance mode.
Lambda + HTTP API or Function URL + S3
AWS serverless
Request-duration compute, optional HTTP API ingress, S3 Standard and tiered Internet transfer.
lambda.requests lambda.gb_seconds apigw.http.first_300m apigw.http.above_300m s3.standard_storage aws.egress.10tb aws.egress.next_40tb aws.egress.next_100tb aws.egress.next_350tb Stack assumptions
- Lambda is x86 on-demand in US East (N. Virginia).
- Compute follows configured memory and end-to-end billable request duration.
- Object delivery is direct from S3; CloudFront is not included.
- Account-level recurring allowances are available only in greenfield mode.
c6a.large + gp3 + ALB + S3
AWS EC2
A fixed-node application tier with root volumes, load balancing, public IPv4, object storage and transfer.
ec2.c6a_large ebs.gp3 alb.hours alb.lcu_hours vpc.public_ipv4 s3.standard_storage aws.egress.10tb aws.egress.next_40tb aws.egress.next_100tb aws.egress.next_350tb Stack assumptions
- Linux On-Demand c6a.large nodes run for the full 720-hour model month.
- Node throughput is a user benchmark, not an AWS performance guarantee.
- The selected node count must carry the modeled hourly peak before this stack can win.
- Snapshots, NAT Gateway, support and operating labour are not priced.
Cloud Run + Cloud Storage Standard
Google Cloud Run
Request-based CPU and memory, requests, standard object storage operations and Premium Tier transfer.
cloud_run.cpu cloud_run.memory cloud_run.requests gcs.standard_storage gcs.class_a gcs.class_b gcp.egress.1tb gcp.egress.10tb gcp.egress.over_10tb Stack assumptions
- Regional reference is Iowa (us-central1).
- Concurrency 1 and zero minimum instances are the complete v1 path.
- Entered wall time must include weighted startup overhead.
- Decimal workload storage and transfer inputs are converted where Google bills binary GiB.
CPX22 fixed nodes
Hetzner
A self-managed fixed-node reference with optional IPv4, included local disk and traffic overage.
hetzner.cpx22 hetzner.egress hetzner.operations Stack assumptions
- Germany / Finland list pricing excludes VAT.
- Node throughput is a user benchmark, not a Hetzner guarantee.
- The selected node count must carry the modeled hourly peak before this stack can win.
- Patching, monitoring, failover, managed load balancing and engineer time are not monetized.
Append-only observations
Price history
Every seeded, published and quarantined rate is retained in D1. Normal movements publish automatically; only a change below 0.5×, above 2×, or between free and paid is held.
Loading recorded price changes…
Read before deciding
List prices are evidence, not a quote.
The model uses public on-demand rates in USD. Unless a row says otherwise, it excludes tax or VAT, support plans, promotional credits, negotiated discounts, commitments, reservations, spot pricing, currency conversion and the value of operating labour.
Shared free allowances may already be consumed by other workloads. The modeler exposes greenfield and marginal views for that reason. Provider invoices, service terms and current official pricing remain authoritative.
Challenge the inputs
Now put a workload through them.
Change the traffic, runtime, storage and capacity assumptions. The result keeps every source and warning beside the bill it affected.
Model a workload