Cloud Infrastructure as a Service (IaaS) Explained: AWS, Azure, GCP Compared
Country Manager, Sweden
AI, DevOps, Security, and Cloud Solutioning. 12+ years leading enterprise cloud transformation across Scandinavia
Cloud Infrastructure as a Service (IaaS) Explained: AWS, Azure, GCP Compared
Infrastructure as a Service is the cloud layer that turns physical hardware into programmable resources you rent by the second. The IaaS layer underpins almost everything else in the cloud — PaaS, container platforms, serverless runtimes — and the choices made here propagate through every later architectural decision. This article compares the IaaS offerings of the three major hyperscalers in the dimensions that matter for enterprise procurement: compute, storage, networking, identity, and pricing.
What IaaS Actually Includes
IaaS gives you the virtualised data centre. The customer manages everything from the operating system up; the provider manages the physical hardware, hypervisor, networking fabric, and storage backplane. The standard scope:
- Compute — virtual machines, dedicated hosts, GPU instances, ARM instances
- Block storage — VM-attached disks (SSD, HDD)
- Object storage — bulk data tiers
- Networking — VPC/VNet, subnets, load balancers, NAT, VPN, peering
- Identity — IAM, KMS, secrets management
- DNS and CDN — managed DNS, edge caching
Above this layer, IaaS gradually merges into PaaS as managed databases, container services, and serverless runtimes get layered on. The boundary is fuzzy — Azure App Service is PaaS but runs on Azure VMs that are themselves IaaS.
Compute Comparison
All three providers offer comparable VM families. The naming and the pricing differ, but the broad shape is the same.
| Use case | AWS | Azure | GCP |
|---|---|---|---|
| General-purpose | m6i, m7g (Graviton) | D-series | n2, n2d, t2d |
| Compute-optimised | c6i, c7g | F-series | c2, c2d |
| Memory-optimised | r6i, r7g, x2gd | E-series, M-series | n2-highmem, m1, m2 |
| GPU | p5, g6, g5 | NDv5, NCv5 | A2, G2, A3 |
| ARM-native | Graviton (m7g, c7g, r7g) | Cobalt (Dpsv6, Dpdsv6) | Axion (c4a, n4a) |
| Spot / preemptible | Spot Instances | Spot VMs | Spot VMs |
For general-purpose workloads, the price-per-performance gap between providers is small (5-15%). For ARM-native workloads, AWS Graviton has a 12-18 month head start and a meaningful lead in mature workload coverage. For GPU workloads, availability and quota matter more than per-hour pricing — the cheapest GPU you cannot allocate is infinitely expensive.
Need expert help with cloud infrastructure as a service (iaas) explained?
Our cloud architects can help you with cloud infrastructure as a service (iaas) explained — from strategy to implementation. Book a free 30-minute advisory call with no obligation.
Storage Comparison
Block storage is broadly equivalent across providers. Object storage is where the differentiation matters most.
- AWS S3 — most mature, widest ecosystem support, six storage classes from Standard to Glacier Deep Archive
- Azure Blob Storage — strong integration with Microsoft estate, Hot/Cool/Archive tiers, Azure Data Lake Gen2 layered on top
- GCS — Google's object storage, four classes, strong consistency since 2020, integrates tightly with BigQuery
For data-warehousing patterns where the same object store powers analytics, GCS and S3 are roughly equivalent. For Microsoft-centric estates, Blob Storage's tighter integration with Synapse, Fabric, and Power BI is meaningful.
Networking Comparison
The networking primitives are similar in capability but different in detail in ways that affect day-1 architecture.
| Concept | AWS | Azure | GCP |
|---|---|---|---|
| Virtual network | VPC (per region, per account) | VNet (per region, per subscription) | VPC (global by default) |
| Subnet boundary | Per AZ | Per region | Per region |
| Default routing | Per-subnet route tables | System routes + UDRs | Auto routes per subnet |
| L4 load balancer | NLB | Standard LB | Internal/External TCP/UDP LB |
| L7 load balancer | ALB | Application Gateway, Front Door | HTTPS LB |
| Service mesh option | App Mesh, customer-installed Istio | Service Mesh (preview / customer Istio) | Anthos Service Mesh, customer Istio |
The biggest model difference: GCP VPCs are global by default; AWS and Azure are regional. For global-by-default architectures, GCP's model is conceptually simpler but creates IP-management implications worth understanding before committing.
Identity Comparison
IAM is where each provider's history shows most. AWS IAM emerged in 2010 with policy-based access control attached to resources. Azure AD (now Entra ID) evolved from on-prem Active Directory and integrates with the broader Microsoft identity estate. GCP IAM is the youngest and most consistent, modelled on resource hierarchy with inheritance.
- AWS — most powerful policy language, steep learning curve, well-documented patterns
- Azure — most enterprise-friendly via Entra ID and conditional access, RBAC layered on resource scope
- GCP — cleanest model, simplest mental model, less expressive than AWS at scale
Pricing Comparison
List pricing for equivalent workloads sits within 10-15% across providers. Real cost depends on:
- Reserved / committed-use discounts (typically 30-60% savings on 1- or 3-year commits)
- Egress costs — historically a key cost driver and a key lock-in lever. EU Data Act-driven changes from 2024-25 reduced this somewhat
- Premium-support contracts (often the difference between paying list and paying enterprise rates)
- Per-service quirks — some services are markedly cheaper on one provider than others (e.g., GCP BigQuery vs. AWS Redshift / Azure Synapse for serverless data warehouse workloads)
The Three Patterns We See in Customer Procurement
- AWS-primary, Azure-secondary — most common pattern in cloud-native organisations. AWS for application infrastructure, Azure for productivity / Microsoft estate
- Azure-primary, AWS-secondary — common in enterprise IT-led adoption. Azure for application infrastructure (because the Microsoft contract was already there), AWS for specific workloads
- GCP-primary, AWS-secondary — common in data-heavy organisations. GCP for analytics and ML, AWS for everything else
Pure single-cloud is rare and usually deliberate (small startups; some heavily regulated workloads). Multi-cloud-by-design with strict workload isolation is increasingly common in mid-market and large enterprise.
How Opsio Helps
Opsio is an accredited managed-service partner across all three major clouds — how Opsio delivers aws managed, Opsio's azure managed, and google cloud managed services. Our cloud infrastructure delivery service maps customer workloads to the right provider portfolio, designs the landing zones, and operates the platforms long-term. Customers running across providers also use our finops cloud for enterprise service to keep the multi-cloud bill predictable and the commitment portfolio rationalised.
For hands-on delivery, see end-to-end managed service.
For hands-on delivery, see google cloud for enterprise.
For hands-on delivery, see zero-downtime datadog monitoring.
For hands-on delivery, see Opsio's azure backup.
For hands-on delivery, see how Opsio delivers azure disaster.
For hands-on delivery, see azure sentinel delivery.
For hands-on delivery, see google cloud devops services.
For hands-on delivery, see azure managed services.
For hands-on delivery, see azure infrastructure services.
For hands-on delivery, see cloud scalability services.
For hands-on delivery, see managed cloud hosting.
For hands-on delivery, see cloud solutions services.
For hands-on delivery, see end-to-end aws cloud.
For hands-on delivery, see cloud migration delivery.
For hands-on delivery, see managed infrastructure service.
For hands-on delivery, see aws cloud for enterprise.
For hands-on delivery, see azure cloud services.
For hands-on delivery, see azure cloud cost management services.
For hands-on delivery in India, see how Opsio delivers cloud migration.
About the Author
Country Manager, Sweden at Opsio
AI, DevOps, Security, and Cloud Solutioning. 12+ years leading enterprise cloud transformation across Scandinavia
Editorial standards: This article was written by a certified practitioner and peer-reviewed by our engineering team. We update content quarterly to ensure technical accuracy. Opsio maintains editorial independence — we recommend solutions based on technical merit, not commercial relationships.