Skip to content
Graph of Plan

Cloudflare Cost Structure Works

Cloudflare’s pricing model (Workers, Durable Objects, R2) will remain cost-competitive for our workload patterns and unit economics.

The Assumption

Our entire technical architecture is built on Cloudflare primitives. We’re betting that:

  1. Current pricing remains stable (no >50% increases)
  2. Our workload patterns align with their pricing model
  3. Costs scale sub-linearly with usage (economies of scale, not diseconomies)
  4. No hidden costs emerge at scale (egress, CPU limits, etc.)

This is a supplier concentration risk we’re explicitly accepting in exchange for developer experience and time-to-market.

Evidence

Current evidence:

  • Dev costs tracking within projections (~$50/month at current scale)
  • Workers pricing stable for 2+ years (no significant changes)
  • Cloudflare’s business model incentivises aggressive pricing (they want ecosystem lock-in)
  • No egress fees (unlike AWS/GCP)—reduces cost surprises

Pricing model analysis:

  • Workers: $0.50/million requests (paid plan) — predictable
  • Durable Objects: $0.15/million requests + $0.15/GB-month storage — potentially expensive at scale
  • R2: $0.015/GB-month storage, no egress — very competitive
  • D1: $0.75/million reads, $1.00/million writes — reasonable

Risk factors:

  • Durable Objects could become expensive with high-frequency access patterns
  • CPU time limits (50ms free, 30s paid) could limit compute-heavy agents
  • Cloudflare’s pricing power increases as we become more dependent

Counter-Evidence

What would prove this wrong:

  • Cloudflare raises prices >50% on core primitives
  • Our workload patterns hit expensive tiers unexpectedly
  • Gross margin under 50% at projected 1,000-customer scale
  • Egress fees introduced (breaking a key advantage)

Warning signs to monitor:

  • Cloudflare quarterly earnings mentioning pricing changes
  • Developer community complaints about cost surprises
  • Our own bills growing faster than usage

Impact If Wrong

Products affected: All products—SmartBoxes, Nomos Cloud, Murphy, P4gent

Technical impact: Would need to:

  • Re-architect for AWS/GCP (6+ months of work)
  • Accept higher baseline costs
  • Lose edge-first deployment model
  • Rebuild around different primitives (Lambda, DynamoDB, etc.)

Financial impact:

  • At 1,000 customers, 50% cost increase = ~£2,000/month additional COGS
  • Could flip unit economics from profitable to unprofitable
  • Runway consumed faster if we need to migrate

Mitigation strategies:

  • Keep architecture portable where possible
  • Monitor costs weekly, not monthly
  • Set alerts for cost anomalies
  • Maintain relationship with Cloudflare (startup credits, communication)

Testing Plan

Ongoing monitoring:

  1. Weekly cost review: Track spend per primitive
  2. Unit economics model: Update with real data as usage grows
  3. Cloudflare communications: Monitor announcements, earnings calls
  4. Competitive pricing: Track AWS/GCP/Fly.io/Modal alternatives

Stress testing:

  • Model 10x usage scenarios before they happen
  • Identify which primitives become expensive at scale
  • Have migration plan documented (even if we never use it)

Review frequency: Quarterly (or immediately if pricing announced)

Kill criteria: If costs exceed 30% of revenue at target scale, evaluate migration.

Creates risk:

Enables decision:

Affects products:

Assumption

Cloudflare’s pricing model (Workers, Durable Objects, R2) will remain cost-competitive for our workload patterns and unit economics.

Depends On

This assumption only matters if these are true:

Enables

If this assumption is true, these become relevant:

How To Test

Track actual costs as usage scales; model break-even points; monitor Cloudflare pricing announcements.

Validation Criteria

This assumption is validated if:

  • Gross margin over 70% at scale
  • No pricing changes that break model
  • Costs scale sub-linearly with usage

Invalidation Criteria

This assumption is invalidated if:

  • Cloudflare raises prices by over 50%
  • Workload patterns hit expensive tiers
  • Gross margin under 50% at projected scale

Current Evidence

  • Current dev costs within projections
  • Workers pricing stable for 2+ years

Dependent Products

If this assumption is wrong, these products are affected:

Decisions Depending On This