| Aspect | Amazon Web Services (AWS) | Microsoft Azure | Google Cloud (GCP) |
|--------|----------------------------|-----------------|--------------------|
| Starting Price | Usage-based; service-specific | Usage-based; service-specific | Usage-based; service-specific |
| Typical fit | Broad service catalog; established IaaS/PaaS; ecosystem-focused enterprises | Microsoft-centric enterprises (AD, Windows Server, M365); hybrid with Azure Stack/Arc | Data/analytics-centric organizations; Kubernetes/GKE-focused teams; certain managed data services |
| Key Strength | Extensive range of services; global footprint; mature primitives and partner ecosystem | Tight integration with Microsoft stack; enterprise agreements; hybrid tooling | Strong data/AI tooling reputation; some managed services and developer tooling strengths |
| Key Weakness | Operational complexity; bill predictability; support experience variability; egress/transfer costs | Pricing complexity and discount/contract opacity; portal complexity | GPU pricing and product change impacts in some scenarios |
- Direct competitors and differentiators
- Azure
- Strong enterprise procurement motion; formal vendor pricing calculator and integrations for Microsoft-centric environments.
- Claims about price parity with AWS should be interpreted cautiously—comparability and discounting vary by workload and contract.
- GCP
- Publishes SKU pricing and pricing table documentation; community feedback suggests changes to pricing presentation can complicate automation.
- Pricing comparison: value proposition vs alternatives
- Across major clouds, list pricing is only part of the story:
- Discount constructs (vendor-specific savings plans or committed-use discounts), enterprise agreements, support plan costs, and network egress can dominate total costs.
- Independent conclusions about which provider is cheaper vary widely by workload; workload-specific cost models are required.
- Customer sentiment comparing clouds (community)
- Some community discussions argue GCP can be less expensive for certain managed database sizes; others indicate GCP is more expensive for some GPU instances.
- The variability reinforces that a workload-specific model is required.
