Cloud Pricing Decoded: How Cloudflare, AWS, Azure, and Google Cloud Actually Charge You
A deep analysis of cloud pricing philosophies — flat-rate vs metered vs committed-use models, hidden costs (egress, cross-AZ, API calls, support), free tier comparison, and total cost of ownership modeling across Cloudflare, AWS, Azure, and GCP.
Frequently Asked Questions
Find answers to common questions
Hyperscaler pricing has many dimensions: compute (per-hour or per-second), storage (per-GB/month), data transfer (per-GB, varying by source, destination, and region), API calls (per-request, varying by operation type), and feature-specific charges (per-rule, per-health-check, per-log-entry). A single AWS service like S3 has charges for storage class, PUT/GET/LIST operations, data transfer out, data transfer between regions, lifecycle transitions, and optional features. This dimensionality makes prediction difficult without detailed usage monitoring.
The most commonly underestimated costs are: (1) data transfer/egress fees — often the largest line item at scale, (2) cross-AZ data transfer on AWS (/bin/sh.01/GB each way), (3) NAT Gateway charges on AWS (/bin/sh.045/GB processed), (4) CloudWatch/monitoring costs that grow with resources, (5) support plans (9-5,000+/month), (6) cost of idle resources (oversized instances, unused EBS volumes), and (7) API call costs for high-frequency operations. Cloudflare eliminates many of these by bundling features into flat-rate plans.
Cloudflare uses primarily flat-rate, plan-based pricing. The free plan includes unlimited CDN bandwidth, DNS, DDoS protection, and basic WAF. Paid plans (0-00/month per zone) add features, not capacity. Usage-based services (Workers, R2, D1) have generous free tiers and simple per-unit pricing. Hyperscalers use primarily metered pricing — you pay for every GB transferred, every API call, every compute-second, often with tiered rates. Cloudflare's model is predictable; hyperscaler pricing rewards optimization but punishes unpredictability.
For stable, predictable workloads: yes. AWS Savings Plans offer 30-72% savings over on-demand for 1-3 year commitments. Azure Reserved Instances offer similar discounts. Google Committed Use Discounts offer 37-55% savings. The risk: you pay regardless of usage. If your workload shrinks, you waste the commitment. If it changes shape (different instance type, different region), you may not be able to apply the discount. For variable or evolving workloads, on-demand or spot/preemptible instances often provide better value.
AWS: Developer (9/mo), Business (greater of 00/mo or 3-10% of bill), Enterprise On-Ramp (,500/mo), Enterprise (5,000+/mo). Azure: Developer (9/mo), Standard (00/mo), Professional Direct (,000/mo), Premier (custom). Google: Standard (included), Enhanced (00/mo), Premium (2,500/mo). Cloudflare: Email support (free), Priority support (Business 00/mo plan), Premium support (Enterprise). Support costs are frequently omitted from pricing comparisons but can be significant — 3-10% of a large AWS bill is substantial.
On AWS, data transfer between availability zones in the same region costs /bin/sh.01/GB in each direction (/bin/sh.02/GB round trip). This affects any architecture where services communicate across AZs — which is most high-availability architectures. A database replica receiving 1TB of replication traffic across AZs costs 0/month just for the data transfer. Multiply by multiple services and the cost becomes material. Cloudflare and Google do not charge for equivalent internal data movement.
Cloudflare offers the most generous permanent free tier: unlimited CDN bandwidth, unlimited DNS, DDoS protection, basic WAF, 100K Workers requests/day, 10GB R2 storage. AWS Free Tier includes 12-month trials (750 hours EC2/RDS, 5GB S3, 1TB CloudFront) plus always-free services (1M Lambda requests, 25GB DynamoDB). Azure offers 00 credit for 30 days plus 12-month trials plus always-free services. Google offers 00 credit for 90 days plus always-free services. Cloudflare's free tier is the only one that remains fully useful indefinitely for production workloads.
No. Cloudflare is dramatically cheaper for bandwidth-heavy, security-critical, edge-compute workloads. But for compute-heavy regional workloads (ML training, data processing, large applications), hyperscaler compute is more capable and often cheaper per-compute-unit, especially with reserved pricing. For storage, R2 is cheapest when egress is high, but S3 Glacier is far cheaper for archival storage. The total cost depends entirely on your workload profile.
Include: (1) compute costs at projected usage, (2) storage costs including all tiers, (3) data transfer — egress, cross-region, cross-AZ, CDN, (4) database costs including IOPS/throughput, (5) security services (WAF, DDoS, bot management), (6) monitoring and logging, (7) support plan, (8) DNS and domain costs, (9) SSL/TLS certificate costs, (10) operational cost of managing the platform (engineer hours). Model at 3 traffic levels: current, 3x, and 10x. The cost winner often changes with scale.
The Bandwidth Alliance is a group of cloud and hosting providers that have agreed to waive or reduce data transfer fees for shared customers with Cloudflare. Members include Google Cloud, Microsoft Azure, DigitalOcean, Vultr, Linode, and others. Notably, AWS is NOT a member. If you use Cloudflare with a Bandwidth Alliance partner, egress from that partner to Cloudflare may be free or discounted — further reducing the total cost of a Cloudflare-fronted architecture.
Is your cloud secure? Find out free.
Get a complimentary cloud security review. We'll identify misconfigurations, excess costs, and security gaps across AWS, GCP, or Azure.