AWS vs Azure vs Google Cloud Comparison
How the three major clouds differ in strengths, networking models, identity, data and AI platforms, with a service mapping and a structured way to choose.
Use the interactive Cloud Service Comparison tool for the full mapping of 75 capabilities.
Core service mapping
| Capability | AWS | Azure | Google Cloud |
|---|---|---|---|
| VMs | EC2 | Virtual Machines | Compute Engine |
| Kubernetes | EKS | AKS | GKE |
| Serverless containers | Fargate / App Runner | Container Apps | Cloud Run |
| Functions | Lambda | Functions | Cloud Run functions |
| Object storage | S3 | Blob Storage | Cloud Storage |
| Relational | RDS / Aurora | Azure SQL / Database for PostgreSQL | Cloud SQL / AlloyDB |
| Global strongly consistent SQL | Aurora DSQL | (Cosmos DB, NoSQL) | Spanner |
| NoSQL | DynamoDB | Cosmos DB | Firestore / Bigtable |
| Warehouse | Redshift | Fabric / Synapse | BigQuery |
| Messaging | SQS / SNS / EventBridge | Service Bus / Event Grid | Pub/Sub / Eventarc |
| Streaming | Kinesis / MSK | Event Hubs | Pub/Sub / Managed Kafka |
| API management | API Gateway | API Management | Apigee |
| Identity | IAM / Identity Center | Entra ID | Cloud IAM / Cloud Identity |
| Keys | KMS | Key Vault | Cloud KMS |
| Monitoring | CloudWatch | Azure Monitor | Cloud Monitoring |
| GenAI | Bedrock | Azure OpenAI / Foundry | Vertex AI |
Meaningful differences (not just names)
| Topic | AWS | Azure | Google Cloud |
|---|---|---|---|
| Network scope | VPC is regional | VNet is regional | VPC is global, subnets regional |
| Resource organization | Accounts in Organizations (account = isolation boundary) | Management groups → subscriptions → resource groups | Organization → folders → projects |
| Identity center of gravity | IAM per account + Identity Center | Entra ID (shared with Microsoft 365) | Google identity + IAM inheritance |
| Load balancing | Regional ALB/NLB + CloudFront/Global Accelerator | Front Door (global) + App Gateway | Global anycast L7 LB native |
| Signature strengths | Breadth and maturity of services, largest ecosystem | Enterprise/Microsoft integration, hybrid (Arc), Windows/SQL licensing | Data and analytics (BigQuery), Spanner, Kubernetes heritage, AI/TPUs |
How to choose (decision framework)
- Existing estate and skills: Microsoft-heavy enterprises often favour Azure; teams with deep AWS skills move faster on AWS.
- Workload fit: global consistent databases, analytics at scale, specific AI models or hardware.
- Commercials: enterprise discount commitments, licence portability (Windows/SQL on Azure).
- Regions and compliance: available regions, sovereign cloud offerings, certifications.
- Ecosystem and partners: marketplace products, SI partners, managed services.
- Exit and concentration risk: regulated industries need exit plans.
[!CTO] A primary cloud with selective use of others for specific strengths is usually more effective than spreading every workload across clouds. Multi-cloud multiplies skills, tooling and security effort.
Interview questions and answers
A company asks you to pick a cloud for a new data platform. How do you decide?
Run a structured evaluation: requirements (data volumes, latency, governance, AI needs), proof-of-concept on representative workloads, total cost over 3 years including people, skills and hiring market, integration with existing identity and tools, and contractual leverage. Present a weighted decision matrix and an ADR.
What is the biggest conceptual difference when moving from AWS to GCP?
The resource hierarchy and IAM inheritance through folders/projects instead of account-centric isolation, and the global VPC networking model with Shared VPC.