AWS: $33/month
GCP: $16/month
Same code. Same architecture.
2x difference.
If you’re choosing a serverless platform and thinking “Lambda is the standard” - hold off on your budget calculations.
Tested identical Node.js API on AWS and GCP:
- 15M requests/month
- Average execution: 200ms
- Memory: 512MB
- No VPC, direct HTTP endpoints
Results
AWS Lambda: $33.63/month
- Requests: $2.80
- Compute: $23.33
- CloudWatch: $7.50
GCP Cloud Functions gen2: $15.95/month
- Invocations: $5.20
- Compute: $3.25
- Logging: $7.50
Difference: 2.1x
Why GCP is cheaper
GB-second pricing: $0.0000025 vs AWS $0.0000166667 - 6.7x lower.
Plus larger free tier (2M requests vs 1M), more generous limits (60min execution vs 15min, 32GB memory vs 10GB).
Where AWS pays off
Ecosystem. EventBridge, Step Functions, DynamoDB Streams - native integrations save development time.
GCP requires more glue code through Pub/Sub. What AWS does with built-in triggers requires separate services in GCP.
Real conclusion
Most overspending comes from architecture, not provider choice:
- 512MB memory where 256MB suffices - 2x overpay
- 1000 separate invocations instead of one batch - 1000x overpay
- Logs without retention policy - +$50/month in 6 months
- Bloated dependencies - +2 seconds cold starts
- Sync calls instead of async - double invocations on retry
Infrastructure audits find 30-40% savings without switching providers. Choosing between AWS and GCP gives you 2x savings on a PowerPoint slide, proper architecture - 3-5x on your actual bill.