R2 Egress
$0
Zero egress fees, always. S3 charges $0.09/GB.
Workers Requests
$0.30/M
10M requests included on the $5/mo plan.
Workers AI
$0.011/neuron
10,000 free neurons/day. Run LLMs at the edge.
Platform Fee
$5/mo
One plan unlocks Workers, KV, R2, D1, Queues & more.
Cloudflare
AWS
Azure
GCP
Serverless Compute
Workers vs Lambda vs Azure Functions vs GCP Cloud Functions
| Metric |
Cloudflare Workers |
AWS Lambda |
Azure Functions |
GCP Cloud Functions |
| Platform Fee | | | | |
| Requests Included | | | | |
| Price/M Requests | | | | |
| Compute Duration | | | | |
| Memory Model | | | | |
| Max CPU Time | | | | |
| Cold Start | | | | |
| Edge Deploy | | | | |
| Static Assets | | | | |
| Egress | | | | |
Key Insight: Cloudflare Workers run on V8 isolates with near-zero cold starts, deployed to 300+ cities automatically. AWS Lambda, Azure Functions, and GCP Cloud Functions are region-bound by default and exhibit cold starts of 100ms–3s+.
Key-Value Storage
Workers KV vs DynamoDB vs Cosmos DB vs Firestore
| Metric |
Workers KV |
DynamoDB |
Cosmos DB |
Firestore |
| Reads | | | | |
| Writes | | | | |
| Deletes | | | | |
| Storage | | | | |
| Egress | | | | |
| Global Distribution | | | | |
Key Insight: Workers KV includes global distribution at no extra cost. DynamoDB Global Tables add $1.725/M WRU, and Cosmos DB multi-region writes carry a significant premium.
Durable Objects
Cloudflare Durable Objects vs nearest hyperscaler equivalents (composite services)
| Metric |
Cloudflare Durable Objects |
AWS (Lambda + DynamoDB + API GW) |
Azure (Functions + Cosmos + SignalR) |
GCP (Cloud Run + Firestore) |
| Requests | | | | |
| Duration | | | | |
| WebSocket | | | | |
| Storage | | | | |
| Hibernation | | | | |
Key Insight: Durable Objects are unique to Cloudflare — providing in-memory state, WebSocket coordination, and transactional storage in a single primitive. Hyperscaler equivalents require stitching together 2–3 separate services at higher combined cost.
Object Storage
R2 vs S3 vs Azure Blob Storage vs Google Cloud Storage
| Metric |
Cloudflare R2 |
AWS S3 |
Azure Blob Storage |
Google Cloud Storage |
| Storage/GB-mo | | | | |
| Class A Ops/M | | | | |
| Class B Ops/M | | | | |
| Egress | | | | |
| Free Tier | | | | |
| IA Storage | | | | |
| Min Duration | | | | |
Key Insight: R2 eliminates egress fees entirely — a game-changer for data-heavy workloads. At 10 TB of egress, S3 would cost $900/mo vs. $0 with R2. R2 also offers 10 GB free storage plus generous free Class A/B operations.
Video Streaming
Cloudflare Stream vs MediaConvert+CloudFront vs Azure Media vs GCP Transcoder
| Metric |
Cloudflare Stream |
MediaConvert + CloudFront |
Azure Media + CDN |
Transcoder + CDN |
| Storage | | | | |
| Encoding | | | | |
| Delivery | | | | |
| Ingress | | | | |
| Live Streaming | | | | |
| Pricing Model | | | | |
Key Insight: Cloudflare Stream offers radically simple pricing — minutes stored + minutes delivered. Encoding is free. Hyperscaler video stacks require assembling 3–4 services (storage + encoding + CDN + egress) with complex, unpredictable billing.
Workers AI
Workers AI vs AWS Bedrock vs Azure AI vs Vertex AI — prices per 1M tokens (USD)
| Model |
Cloudflare Workers AI |
AWS Bedrock |
Azure AI |
GCP Vertex AI |
| Llama 3.1 8B — Input | | | | |
| Llama 3.1 8B — Output | | | | |
| Llama 3.3 70B — Input | | | | |
| Llama 3.3 70B — Output | | | | |
| DeepSeek R1 32B — Input | | | | |
| DeepSeek R1 32B — Output | | | | |
| Model |
Cloudflare Workers AI |
AWS Bedrock |
Azure AI |
GCP Vertex AI |
| BGE Small | | | | |
| BGE Large | | | | |
| Model |
Cloudflare Workers AI |
AWS Bedrock |
Azure AI |
GCP Vertex AI |
| FLUX Schnell |
Neuron-based ($0.011/neuron) |
N/A (Titan Image ~$0.01/img) |
DALL-E 3: $0.04/img |
Imagen: $0.02/img |
| Model |
Cloudflare Workers AI |
AWS Transcribe |
Azure Speech |
GCP Speech-to-Text |
| Whisper / STT | | | | |
| Feature |
Cloudflare Workers AI |
AWS Bedrock |
Azure AI |
GCP Vertex AI |
| Free Tier |
10,000 neurons/day |
None (some limited trials) |
Limited free credits |
$300 free credits (90 days) |
| GPU Provisioning |
Automatic (serverless) |
On-Demand or Provisioned |
Provisioned Throughput |
Provisioned Throughput |
| Edge Inference |
Yes — global edge |
No (region-bound) |
No (region-bound) |
No (region-bound) |
Key Insight: Workers AI runs inference at the edge with zero cold starts and a generous free tier (10,000 neurons/day). Hyperscaler AI services are region-bound and typically require provisioned throughput for consistent performance. DeepSeek R1 32B is currently exclusive to Workers AI among these providers.
Additional Cloudflare Services
D1, Queues, Vectorize, Hyperdrive — with hyperscaler equivalents
| Service / Metric |
Cloudflare |
AWS Equivalent |
Azure Equivalent |
GCP Equivalent |
| D1 (SQLite at the Edge) |
| Rows Read | | Aurora Serverless: $0.20/M ACU-s | SQL DB Serverless: $0.000145/vCore-s | Cloud SQL: Instance-based |
| Rows Written | | Aurora: $0.20/M ACU-s | SQL DB: $0.000145/vCore-s | Cloud SQL: Instance-based |
| Storage | | Aurora: $0.10/GB-mo | SQL DB: $0.115/GB-mo | Cloud SQL: $0.17/GB-mo |
| Queues |
| Operations | | SQS: $0.40/M + $0.09/GB transfer | Service Bus: $0.05/M ops (Basic) | Pub/Sub: $40/TiB |
| Vectorize |
| Queried Dimensions | | OpenSearch Serverless: $0.24/OCU-hr | AI Search: $0.10/unit-hr | Vertex Vector Search: $0.35/node-hr |
| Stored Dimensions | | OpenSearch: $0.024/GB-mo | AI Search: Included in unit | Vertex: Included in node |
| Hyperdrive |
| Queries | | RDS Proxy: $0.015/vCPU-hr | N/A (connection pooling via app) | Cloud SQL Proxy: Free (compute cost) |
Key Insight: D1 offers 25 billion rows read/month included — a remarkable free tier for edge-native SQLite. Queues, Vectorize, and Hyperdrive are all included in the $5/mo Workers Paid plan with generous included usage.