Why This Comparison in 2024?
The public cloud market is dominated by three hyperscalers that together account for more than 65% of global spending: Amazon Web Services (32%), Microsoft Azure (23%), and Google Cloud Platform (12%). Every year, CIOs and cloud architects face the same question: which cloud to choose — or how to distribute workloads across several?
This comparison does not seek to crown a "universal winner" — there isn't one. It aims to give you objective criteria to choose the cloud that fits your context, technical stack, regulatory constraints, and business objectives.
Overview of the Three Players
| Criterion | AWS | Azure | GCP |
|---|---|---|---|
| Market share (2024) | ~32% | ~23% | ~12% |
| Number of regions | 33 | 60+ | 40 |
| Available services | 300+ | 200+ | 150+ |
| Security certifications | ISO 27001, SOC 2, HDS | ISO 27001, SOC 2, HDS | ISO 27001, SOC 2 |
| French sovereignty | AWS Local Zones Paris | Azure France Central | GCP Paris (europe-west9) |
Compute: Instances and Containers
AWS EC2 vs Azure VMs vs GCP Compute Engine
AWS has the broadest instance catalogue, with over 750 types across families (general purpose, compute, memory, storage, GPU-accelerated). Azure follows closely with its D, E, F, and N series. GCP offers a different model with "custom" instances (independently configurable vCPU and RAM), which can be very economical for atypical workloads.
- AWS Graviton3 (ARM): best price/performance ratio for Linux workloads, 40% cheaper than Intel equivalents at equal performance
- Azure Spot VMs: pricing up to 90% lower for interruption-tolerant workloads
- GCP Preemptible / Spot VMs: equivalent, with a maximum duration of 24h
- GCP Sustained Use Discounts: automatic discounts with no commitment from 25% monthly usage — a unique advantage versus AWS/Azure which require explicit reservations
Managed Kubernetes
- EKS (AWS): most widely used in enterprise, native IAM/VPC/ALB integration, but control plane costs $0.10/h
- AKS (Azure): free control plane, native Azure AD integration, ideal for Microsoft environments
- GKE (Google): the most technically mature (Google created Kubernetes), Autopilot for zero-ops management, free on a standard cluster
Serverless and Functions
| Criterion | AWS Lambda | Azure Functions | GCP Cloud Run |
|---|---|---|---|
| Max duration | 15 min | Unlimited (Premium) | 60 min |
| Max memory | 10 GB | 14 GB | 32 GB |
| Cold start | ~100–500 ms | ~200–800 ms | ~50–200 ms |
| Free tier | 1M req/month | 1M req/month | 2M req/month |
Managed Databases
This is often the most structuring criterion in a cloud choice, as databases create the strongest lock-in.
- AWS RDS / Aurora: the undisputed leader for managed PostgreSQL and MySQL. Aurora is up to 5x faster than standard PostgreSQL, with auto-scaling storage. DynamoDB remains the serverless NoSQL reference.
- Azure SQL / Cosmos DB: Azure SQL benefits from perfect integration with the Microsoft ecosystem (SSMS, SSIS, Power BI). Cosmos DB is the most complete multi-model alternative (SQL, MongoDB, Cassandra, Gremlin APIs).
- GCP Cloud Spanner / BigQuery: Cloud Spanner offers globally distributed strong consistency — unique on the market. BigQuery is the serverless analytics reference, with no direct rival at AWS or Azure for petabyte-scale analysis.
Artificial Intelligence and ML
AI has become the main differentiating ground for hyperscalers in 2024.
- AWS: Amazon Bedrock (access to Claude, Llama, Mistral, Titan models), SageMaker for end-to-end ML, CodeWhisperer for code assistance
- Azure: Azure OpenAI Service (exclusive access to GPT-4 and o1 models from OpenAI with enterprise SLA), Azure ML, Copilot integrated across the Microsoft 365 ecosystem
- GCP: Vertex AI (Gemini 1.5 Pro natively, Llama access), proprietary TPUs for model training, BigQuery ML for ML on analytical data
Pricing and Cost Optimisation
| Mechanism | AWS | Azure | GCP |
|---|---|---|---|
| Discount without commitment | ❌ | ❌ | ✅ Sustained Use (auto) |
| 1-year reservation | ~40% (Reserved) | ~36% (Reserved) | ~37% (Committed Use) |
| 3-year reservation | ~60% (Savings Plans) | ~55% (Reserved) | ~55% (Committed Use) |
| Spot / preemptible instances | ~90% off | ~90% off | ~80% off |
Decision Matrix
- Choose AWS if: you want the most complete service catalogue, your team holds AWS certifications, you need the richest ISV marketplace ecosystem, or you are deploying a complex microservices architecture on Kubernetes
- Choose Azure if: your organisation is on the Microsoft ecosystem (Active Directory, Office 365, SQL Server, .NET), you need GPT-4 access with an enterprise SLA, or you are subject to regulations requiring Microsoft tooling integration
- Choose GCP if: your needs are data and AI-oriented (BigQuery, Vertex AI), you have a native Kubernetes application (GKE Autopilot is the most advanced), or you want automatic savings without complex reservation management
- Multi-cloud: AWS + Azure is the most common enterprise combination — AWS for cloud-native workloads, Azure for Microsoft integration. Expect a 20–30% operational overhead.
Conclusion
There is no universally superior cloud in 2024. AWS remains the leader with the most complete catalogue, Azure gains ground through Microsoft integration and OpenAI access, GCP stands out on data and Kubernetes. The real question is not "which is the best cloud?" but "which cloud is best for my context?"
A pragmatic approach is to start by analysing your existing stack, internal skills, regulatory constraints, and AI requirements before choosing. Move2Cloud supports its clients through this analysis with a structured cloud assessment methodology, independent of any hyperscaler.