Price the same 3-tier architecture (web + app + database) on AWS, Azure, GCP, and DigitalOcean using each provider's official calculator. Document every assumption, normalize the SKUs across clouds, and produce a reproducible markdown cost table that another builder could rebuild from your notes alone.
AWS — use https://calculator.aws. Pick "Add service", then EC2, RDS, S3, ELB, Data Transfer. Save as a "share link" at the top right.
Azure — use https://azure.microsoft.com/en-us/pricing/calculator/. Add Virtual Machines, Azure Database for PostgreSQL Flexible Server, Storage Accounts, Load Balancer, and Bandwidth. Export to Excel or save with a Microsoft account.
GCP — use https://cloud.google.com/products/calculator. Compute Engine, Cloud SQL for PostgreSQL, Cloud Storage, Cloud Load Balancing, Network egress. Click "Share" to get a permalink.
DigitalOcean — use https://www.digitalocean.com/pricing/calculator. Droplets, Managed Databases, Spaces, Load Balancer, Bandwidth. Flat pricing means fewer knobs — that's the point.
Normalization tip: pick a vCPU + RAM target (e.g. "2 vCPU, 4 GB, ~$30-40/mo") and find the closest SKU on each cloud rather than chasing exact equivalence.
```markdown
| Provider | Compute | Database | Storage | Egress | LB | Total (USD) | Total (INR) |
|--------------|---------|----------|---------|--------|-------|-------------|-------------|
| AWS | $62.40 | $51.10 | $2.30 | $90.00 | $18.40| $224.20 | INR 18,712 |
| Azure | $58.40 | $48.50 | $2.10 | $86.70 | $18.25| $213.95 | INR 17,856 |
| GCP | $54.10 | $46.80 | $2.00 | $80.00 | $18.00| $200.90 | INR 16,775 |
| DigitalOcean | $48.00 | $30.00 | $5.00 | included| $12.00| $95.00 | INR 7,929 |
*Exchange rate used: 1 USD = 83.46 INR (RBI reference, 2025-04-30)*
*Region: us-east (AWS/Azure/GCP us-east-1/eastus/us-east1; DO NYC3)*
*Workload: 2x 2vCPU/4GB web, 1x 4vCPU/8GB app, 1x 2vCPU/4GB Postgres single-AZ, 100GB object storage, 1TB egress*
## Stretch goals
1. Re-price the same architecture under 1-year reserved / committed-use on AWS, Azure, and GCP and add a "Reserved (1yr)" column — note the breakeven utilization.
2. Add a fifth provider (Hetzner, Linode/Akamai, or OVH) to the table and see how the hyperscalers compare to a budget/EU provider.
3. Model 10x scale (20 web instances, 10 app instances, 10TB storage, 100TB egress) and observe which provider's pricing curve is most punishing — egress is usually the answer.
4. Build a small Python script that calls each provider's pricing API directly and regenerates the table without you re-clicking the calculator.
5. Add a "Total Cost of Ownership" column that includes one engineer-hour per month for ops on each provider — make a defensible estimate of whether the cheapest cloud is actually cheapest end to end.