AWS vs AZURE vs GCP
Hyperscaler-uppgörelsen. Vi jämför beräkning, lagring, ML, prissättning och ekosystem över de tre molnplattformarna som driver det moderna internet.
Senast uppdaterad: 4 april 2026 - av Simon Axelsson
Cloud Market Share (2026)
AWS
Amazon Web Services
Den bredaste tjänstekatalogen och djupaste ekosystemet. Förstahandsfördelar syns i mognad och marknadsandel.
Pay-as-you-go | Free tier 12 months
Marknadsandel: 31%
Azure
Microsoft Azure
Företagshybridmolnets ledare. Oslagbar om du kör Windows, Active Directory eller Microsoft 365.
Pay-as-you-go | Free tier 12 months
Marknadsandel: 25%
GCP
Google Cloud Platform
Marknadsledande för data och ML-arbetsbelastningar. BigQuery ensam motiverar plattformen för analysteam.
Pay-as-you-go | $300 free credit
Marknadsandel: 12%
Betygsjämförelse
Funktionsjämförelse
| Funktion | AWS | Azure | GCP |
|---|---|---|---|
| Virtual Machines (EC2 / VMs / GCE)Compute | |||
| Kubernetes (EKS / AKS / GKE)Compute | |||
| Arm-based InstancesCompute | |||
| Spot / Preemptible InstancesCompute | |||
| Object Storage (S3 / Blob / GCS)Storage | |||
| Block StorageStorage | |||
| File Storage (NFS)Storage | |||
| Archive / Cold StorageStorage | |||
| Managed SQL (RDS / SQL / Cloud SQL)Database | |||
| Managed NoSQL (DynamoDB / CosmosDB / Firestore)Database | |||
| Managed Redis / MemcachedDatabase | |||
| Data Warehouse (Redshift / Synapse / BigQuery)Database | |||
| AutoML / No-code MLML & AI | |||
| GPU Instances (H100 / A100)ML & AI | |||
| Foundation Model APIsML & AI | |||
| ML Ops / Model RegistryML & AI | |||
| Functions (Lambda / Functions / Cloud Functions)Serverless | |||
| Container-based ServerlessServerless | |||
| Event-driven ArchitectureServerless | |||
| CDN (CloudFront / Front Door / Cloud CDN)CDN & Edge | |||
| Edge Functions / Edge ComputeCDN & Edge | |||
| Multi-factor Auth (MFA)IAM & Security | |||
| Active Directory IntegrationIAM & Security | |||
| SOC 2 / ISO 27001 / HIPAAIAM & Security | |||
| Zero Trust ArchitectureIAM & Security | |||
| Hybrid / On-prem IntegrationHybrid |
Prisjämförelse
| Kategori | AWS | Azure | GCP |
|---|---|---|---|
| Free Tier Duration | 12 months + always-free services | 12 months + always-free services | $300 credit (90 days) + always-free |
| Compute (basic VM/mo) | ~$8/mo (t4g.micro) | ~$10/mo (B1s) | ~$6/mo (e2-micro free) |
| Object Storage (GB/mo) | $0.023/GB (S3 Standard) | $0.018/GB (Hot) | $0.020/GB (Standard) |
| Data Egress (GB) | $0.09/GB (first 10TB) | $0.087/GB (first 5TB) | $0.12/GB (first 1TB) |
| Managed SQL (basic/mo) | ~$15/mo (db.t4g.micro) | ~$5/mo (Basic DTU) | ~$7/mo (db-f1-micro) |
| Serverless (per M invocations) | $0.20/M (Lambda) | $0.20/M (Functions) | $0.40/M (Cloud Functions) |
| Support Plan (Business) | $100/mo + 10% of spend | From $100/mo (Standard) | $250/mo (Enhanced) |
| Enterprise Agreement | Custom (1-3 year commits) | Custom (EA with Microsoft) | Custom (CUDS / SUDs) |
Vinnare per användningsfall
AWS: 2 vinster | Azure: 3 vinster | GCP: 5 vinster
Startups
$300 free credit, simplest pricing, GKE Autopilot is magic
Enterprise
Active Directory, hybrid cloud (Arc), M365 integration
ML / AI Workloads
Vertex AI, TPUs, BigQuery ML, and Gemini APIs lead the field
Serverless
Lambda maturity, Step Functions, and broadest event source integration
Data & Analytics
BigQuery is the best serverless data warehouse, period
Hybrid Cloud
Azure Arc and deep Windows Server integration
Service Breadth
200+ services - more than any competitor
Kubernetes
GKE set the standard - built by the team that created k8s
Security & Compliance
Most compliance certifications, Entra ID is industry standard
Cost Optimization
Per-second billing, sustained use discounts, simpler pricing
Simons bedömning
Running infrastructure for a 50-person org with an Active Directory domain, my cloud choice is driven by practical reality: Azure for identity and hybrid, GCP for data, AWS when you need specific services nobody else has.
Azure wins for us because of Entra ID (formerly Azure AD) and the seamless bridge to vår on-prem AD-domän. When your company runs Microsoft 365, Azure is not just convenient - it's the only choice that doesn't create identity management headaches.
GCP is where our data lives. BigQuery is genuinely best-in-class for analytical workloads, and the pricing model (pay per query, not per cluster) means our data platform costs a fraction of what Snowflake or Redshift would charge at our scale.
AWS has the broadest catalog and the strongest ecosystem. When I need a very specific managed service - like SES for transactional email, or a particular database engine - AWS almost always has it. But the console UX and billing complexity remain painful.
Multi-cloud is not just a buzzword for us - it's operational reality. The key is picking the right workload for the right cloud, not trying to abstract away the differences.
- Simon Axelsson, April 2026