Cloud hosting cost estimator

Cloud hosting cost estimator — AWS, GCP, Azure, and the cheap providers

Cloud costs are one of the top three expenses for most SaaS companies and the fastest-growing for product-led apps. But pricing is intentionally complex — each provider has dozens of SKUs with tier-based pricing, regional variance, and bundle discounts. This calculator gives you a rough monthly estimate across the main providers so you can see order-of-magnitude differences before architecting.

The five cost categories

1. Compute: EC2, GCE, Azure VM, Droplets, Fly machines, Lambda/Functions. Billed per compute-hour or per-invocation.
2. Bandwidth (egress): Outbound data transfer. AWS $0.09/GB, GCP $0.12/GB, Azure $0.087/GB, DO $0.01/GB, Fly $0.02/GB. The biggest variance across providers.
3. Database: Managed Postgres/MySQL, DynamoDB, Firestore, Cosmos DB. $0.12-0.28/GB-mo plus compute for the DB instance.
4. Object storage: S3, GCS, Blob. $0.02-0.025/GB-mo. Cheap but accumulates. Glacier / archive tiers drop to $0.004/GB-mo.
5. Request fees: Per-million requests on Lambda ($0.20/M), API Gateway ($1.00/M), DynamoDB requests, S3 GET/PUT. Usually small but adds up at scale.

Provider reality check

Rough pricing spread for the same workload (50k MAU, modest app):

• AWS: $500-2000/mo baseline. Premium bandwidth ($0.09/GB) dominates at scale. Most feature-rich, most expensive for the same compute.
• GCP: $450-1900/mo. Cheaper compute than AWS. Egress is the mostexpensive at $0.12/GB. Best BigQuery + ML services.
• Azure: $550-2100/mo. Similar to AWS on compute. Best enterprise integration (AD, Office 365). Lowest preference among startups.
• DigitalOcean: $80-400/mo. Droplets at fixed prices, bandwidth at $0.01/GB. No RDS-equivalent (has Managed Databases but fewer features). Best for < $5k/mo workloads.
• Fly.io: $100-500/mo. Modern edge-first deploy model. Pay-per-second compute. Strong for apps that need global edge compute.
• Hetzner: $50-250/mo. German provider. Bare-metal and cloud. 10-20x cheaper than AWS but requires more DevOps. Growing fast in 2024+.

Where unexpected costs hide

The line items that surprise teams:

• Inter-AZ bandwidth: AWS charges $0.01/GB for data between availability zones inside your VPC. Multi-AZ databases + distributed apps rack this up quickly.
• NAT Gateway: $32.40/mo per AZ + $0.045/GB processed. For a 3-AZ setup: ~$100/mo baseline just for the gateway before any traffic.
• Load balancers: ALB $22/mo baseline + LCU costs. 3 environments = $66/mo minimum.
• CloudWatch logs: $0.50/GB ingested. Verbose app logging + retained for 30 days can hit $500+/mo on a modest app.
• VPN / VPC peering: $0.05/hr per VPN connection + bandwidth. Multi-region setups accumulate.
• Support plans: Business support is 10% of spend with $100/mo minimum. Enterprise 7%. Often forgotten in budget.

Cost optimization patterns

1. Reserved Instances / Savings Plans: 30-72% off compute for 1-3 year commitments. Start here for stable workloads. Lose flexibility.
2. Spot instances: 60-90% off for interruptible workloads (batch jobs, ML training). Not for production web traffic.
3. Right-sizing: Most teams overprovision by 30-50%. AWS Compute Optimizer, GCP Recommender flag underutilized instances. Run quarterly.
4. S3 Intelligent-Tiering: Auto- moves rarely-accessed objects to cheaper tiers. Typical 20-40% savings on storage bills for mature data sets.
5. CloudFront / CDN: Caches responses, reduces egress bandwidth 60-80% for static + cacheable dynamic content. $0.085/GB CloudFront vs $0.09/GB direct — small, but combined with caching cuts total egress meaningfully.
6. Compression: Gzip/Brotli on responses cuts bandwidth 60-80% for text/JSON. Usually already on but worth auditing.
7. Database indexes + connection pooling: DB often biggest cost driver. Proper indexes can cut needed RDS class by 2-4 sizes. Connection pooling (pgbouncer, RDS Proxy) reduces needed instance size.

When to migrate providers

Rough triggers for considering a switch:

• Bandwidth > 40% of bill, app is mostly static/ cacheable: consider Cloudflare R2 + Workers
• Total bill < $5k/mo and not using managed services: DO / Fly / Hetzner often 5-10x cheaper
• Heavy compute, tolerable eventual consistency: Hetzner + bare-metal orchestration can drop 50-80%
• Global edge needs (low latency, 20+ regions): Fly.io, Cloudflare Workers beat hyperscaler multi-region setups
• Enterprise compliance required (HIPAA, FedRAMP, SOC 2 bundled): stay with AWS/GCP/Azure

The FinOps discipline

Any company spending > $10k/mo on cloud should have FinOps discipline:

• Tag every resource with owner + project + env
• Monthly cost review by team/service
• Budget alerts at 80%, 100%, 120% of forecast
• Quarterly right-sizing audit
• Annual RI / Savings Plan renewal decision

Teams without FinOps typically overspend by 25-45%. Teams with disciplined FinOps often reduce cloud costs year-over-year even as usage grows.