Cloudflare Workers: real pricing and costs in 2026

At a glance

• Workers Free: 100,000 requests/day, 10 ms CPU per request. Enough for a prototype or small site.
• Workers Paid: $5/month, 10 million requests included then $0.30/million. No egress fees.
• R2 (storage): $0.015/GB/month, egress free, unlike S3 (~$0.09/GB).
• D1 (SQL edge): 25 billion rows read included/month, then $0.001/million. Storage from $0.75/GB.
• In production, a typical SMB app runs for $5-20/month on Cloudflare, where AWS can exceed $50 due to egress.

The free tier (Workers Free)

Cloudflare offers a generous free tier, no credit card needed:

• Workers: 100,000 requests/day, 10 ms CPU per invocation.
• R2: 10 GB storage, 1 million write operations (Class A), 10 million reads (Class B) per month.
• D1: 5 million rows read/day, 100,000 rows written/day, 5 GB storage.
• KV: 100,000 reads/day, 1,000 writes/day, 1 GB stored.

Plenty for a prototype, a dynamic showcase site, or a low-traffic API. Beyond that, you need the paid plan.

The paid plan (Workers Paid: $5/month)

The paid plan costs $5 USD/month, flat. It unlocks much larger limits and adds features (Durable Objects, Vectorize, Queues, Logpush). Pricing is usage-based, with no egress or bandwidth fees.

Workers (compute)

• Requests: 10 million included/month, then $0.30/million additional.
• CPU time: 30 million CPU ms included/month, then $0.02/million CPU ms. Max 5 minutes CPU per invocation (30 s default).
• Egress: free, unlimited. This is the biggest difference from AWS.

Example: a Worker serving 15 million requests/month with 7 ms CPU average costs $8/month ($5 subscription + $1.50 requests + $1.50 CPU). A site with 100 million requests/month costs around $45/month.

R2 (file storage, S3 alternative)

• Storage: $0.015/GB/month (Standard). 10 GB included.
• Class A operations (writes, uploads): $4.50/million. 1 million included.
• Class B operations (reads, downloads): $0.36/million. 10 million included.
• Egress: free. This is the key point: on S3, 1 TB of outbound data costs ~$90; on R2, it is $0.

Concrete example: serving 1 TB of files per month from R2 costs around $15 (storage only, no egress). The same on S3 costs $90+ in egress fees alone. For a site serving images, PDFs or videos, the difference measures in hundreds of dollars per month.

D1 (serverless SQL database at the edge)

• Rows read: 25 billion included/month, then $0.001/million rows.
• Rows written: 50 million included/month, then $1.00/million rows.
• Storage: 5 GB included, then $0.75/GB/month.
• Egress: free.

D1 bills by rows read/written, not by query count. A SELECT on a 5,000-row table without an index counts 5,000 rows read. With an index on the filtered column, it may only read a few dozen. Good indexing divides your D1 bill. This site (buildoto.com) reads all its content from D1 at the edge, and the bill stays negligible.

Workers AI and Vectorize (AI at the edge)

• Vectorize (vector database): 50 million queried dimensions included/month, then $0.01/million. 10 million stored dimensions included, then $0.05/100 million.
• Workers AI: billed per neuron (AI compute unit), 10,000 neurons/day free, then per-model pricing.

For a RAG (document chatbot) in production at an SMB, AI costs stay low: a few thousand vectors, a few hundred queries/day, and you run for under $10/month of Workers AI + Vectorize. The biggest line item remains the generation LLM API (Mistral, OpenAI), not the Cloudflare infrastructure.

What it costs in production: 3 real examples

1. Dynamic showcase site (like buildoto.com)

A Next.js site deployed via OpenNext, content read from D1 on each request:

• Workers: ~300,000 requests/month = $5/month (all within the plan).
• D1: ~2 million rows read/month = $0 (well within the 25 billion included).
• R2: ~500 MB of assets = $0 (within the 10 GB included).
• Total: $5/month (~4.60 EUR).

2. B2B SaaS API with auth and storage

• Workers: ~5 million requests/month, 5 ms CPU average = $5/month (within the plan).
• R2: 50 GB of files, 500,000 reads/month = ~$1/month (storage).
• D1: ~10 million rows read/month = $0 (included).
• Total: ~$6/month (~5.50 EUR).

3. Document RAG (chatbot on 1,000 documents)

• Workers: ~100,000 requests/month = $5/month (plan).
• Vectorize: 10,000 vectors x 768 dimensions, 1,000 queries/month = ~$0.30/month.
• R2: 2 GB of documents = $0 (included).
• LLM API (Mistral): ~500 queries/month = ~$5-15/month (depending on model).
• Total: $10-20/month (~9-18 EUR). The dominant cost is the LLM API, not Cloudflare.

Cloudflare Workers vs AWS Lambda: the comparison that matters

Both are serverless, but the bill looks completely different once data leaves.

• Requests: Workers $0.30/M (10M included) vs Lambda $0.20/M (1M included). Lambda is cheaper per request, but includes 10x less free volume.
• Egress: Workers $0 vs Lambda/S3 ~$0.09/GB. This is where Cloudflare wins: 1 TB/month of outbound = $0 on R2, ~$90 on S3.
• Cold start: Workers ~0 ms (V8 isolates) vs Lambda 200-1000 ms cold. Not a cost, but a direct UX impact.
• Storage: R2 $0.015/GB vs S3 $0.023/GB, and R2 has no egress fees.

When Lambda still makes sense: if you are already on AWS, using VPC, RDS, or deep AWS services (SQS, EventBridge), and your egress is internal (between AWS services). But as soon as you serve content to the Internet, AWS egress dominates the bill.

Developing on Cloudflare Workers

Knowing the pricing is one thing, making the platform work efficiently is another. Classic pitfalls: D1 without indexes (row-read bill explodes), Workers running too long in CPU, R2 misconfigured. This is what I do on beforbuild.com (my B2B SaaS in production) and what I adapt to your needs.

If you have a Cloudflare project (edge API, RAG, AWS migration, full app), let us talk: my Cloudflare Workers service, fixed price from 490 € HT, free 30-min audit.

Sources

• Cloudflare Workers · official pricing
• Cloudflare R2 · official pricing
• Cloudflare D1 · official pricing