Which is cheaper to run in the long term?

That depends on the workload, the services chosen, how well resources are sized and the agreements you sign, not on the provider's name. Model your own workload on each shortlisted cloud. Our cloud cost optimization page covers how running costs are reviewed.

Which is fastest to get started on?

The one your team already knows. Setting up a landing zone, identity and pipelines on an unfamiliar cloud takes longer than any difference between the services themselves.

Is it hard to switch clouds later?

Virtual machines, containers and open source databases move with modest effort. Provider-specific services, such as proprietary databases, identity integrations and serverless event wiring, take longer. Keep those at the edges of the design if leaving matters to you.

Does Kubernetes make us cloud neutral?

For compute it helps: EKS, AKS and GKE all run standard Kubernetes. Databases, identity, networking and storage still differ by provider, so a workload is only as portable as those layers.

Can Agnotic build on whichever cloud we choose?

Yes. Agnotic Technologies builds and runs software on AWS, Azure and Google Cloud, and recommends a provider based on your workload rather than a preferred vendor.

City lights spread across a continent at night, seen from orbit under a starry sky

AWS vs Azure vs Google Cloud

AWS vs Azure vs Google Cloud: which fits your workload?

AWS, Microsoft Azure and Google Cloud each publish documents that map their services to the others, because the core building blocks line up. That is why feature lists rarely settle the choice. This page compares the three on what does differ for a general product: the estate you already run, the services one workload needs, landing zone setup and your team's skills.

15-Minute Scoping Call All Comparisons

How do you choose between AWS, Azure and Google Cloud?

Quick answer

AWS vs Azure vs Google Cloud is settled by your workload and your estate, not by a ranking. All three offer comparable compute, database, storage and identity services. What decides it is the Microsoft licences and identity you already run, the data and AI services you need, regions and team skills. Agnotic Technologies tests the shortlist on one workload.

Criterion	AWS	Azure / Google Cloud
Virtual machines	Amazon EC2	Azure: Azure Virtual Machines. Google Cloud: Compute Engine.
Managed Kubernetes	Amazon EKS	Azure: Azure Kubernetes Service (AKS). Google Cloud: Google Kubernetes Engine (GKE).
Serverless containers and functions	AWS Fargate, AWS App Runner, AWS Lambda	Azure: Azure Container Apps, Azure Functions. Google Cloud: Cloud Run, Cloud Run functions.
Relational databases	Amazon RDS, Amazon Aurora	Azure: Azure SQL Database, Azure Database for PostgreSQL and for MySQL. Google Cloud: Cloud SQL, AlloyDB for PostgreSQL.
NoSQL databases	Amazon DynamoDB	Azure: Azure Cosmos DB. Google Cloud: Firestore, Bigtable.
Object storage	Amazon S3	Azure: Azure Blob Storage. Google Cloud: Cloud Storage.
Identity	AWS IAM, AWS IAM Identity Center	Azure: Microsoft Entra ID. Google Cloud: Cloud IAM, Cloud Identity.
Resource hierarchy	Accounts in AWS Organizations	Azure: subscriptions in management groups. Google Cloud: projects and folders in an organization.
Infrastructure as code	AWS CloudFormation; Terraform works on all three	Azure: Bicep and ARM templates. Google Cloud: Infrastructure Manager, which runs Terraform.
Landing zone	AWS Control Tower	Azure: Azure landing zone accelerators. Google Cloud: landing zone design and the enterprise foundations blueprint.

What they are

What is AWS

Amazon Web Services organizes resources in accounts grouped under AWS Organizations. Core services include Amazon EC2 for virtual machines, Amazon EKS for Kubernetes, AWS Lambda for functions, Amazon RDS and Aurora for relational data, DynamoDB for NoSQL and Amazon S3 for objects. AWS Control Tower sets up a governed multi-account landing zone.

What is Azure / Google Cloud

Microsoft Azure organizes resources in subscriptions under management groups and uses Microsoft Entra ID for identity, which matters to companies already on Microsoft 365. Google Cloud organizes resources in projects and folders under an organization, with Compute Engine, GKE, Cloud Run, Cloud SQL and BigQuery among its core services.

Key differences

Why doesn't the service map settle the choice?

Each provider publishes its own map. Microsoft's Azure for AWS professionals series compares Azure and AWS services in every category, and Google Cloud's comparison table maps its generally available services to similar or comparable offerings in AWS and Azure. When the building blocks line up this closely, the deciding factors sit outside the feature list.

Does the estate you already run decide it?

Often it does. A company on Microsoft 365 already signs staff in through Microsoft Entra ID and may hold Windows Server or SQL Server licences it can bring to Azure; our Azure migration page covers how those licences change the plan. A team with established AWS accounts, runbooks and skills carries them into every new workload. Count the cost of leaving what you have, not only the features of what you would join.

Which managed services does the workload need?

List the services the product will depend on: a specific database engine, a data warehouse, an event bus, AI services and regions close to your users. Check that each exists in every region you need and that your team can operate it. For AI-heavy products our cloud AI platform selection guide covers that choice, and for health data the healthcare cloud selection guide covers BAA scope.

How does landing zone setup differ by provider?

All three treat the landing zone as the first thing to build. AWS Control Tower orchestrates AWS Organizations, AWS Service Catalog and IAM Identity Center to set up a multi-account environment it calls a landing zone. Microsoft's Cloud Adoption Framework splits an Azure landing zone into a platform landing zone and workload landing zones. Google Cloud calls a landing zone a cloud foundation, often a prerequisite to deploying enterprise workloads.

Should you use more than one cloud?

Only for a reason you can name: a customer contract that requires a second provider, a service with no equivalent, or an acquisition that brought one in. Every extra cloud adds its own identity, network, security and cost tooling for your team to learn and run. Most products are better served by one primary cloud, with data kept in open formats and infrastructure kept in code.

When to use each

When to use AWS

Your team and existing workloads already run in AWS accounts.
The product depends on an AWS managed service you have already tested.
Your customers or partners require deployment in AWS.
Your engineers hold AWS skills and runbooks you would otherwise rebuild.

When to use Azure or Google Cloud

Azure: your company runs Microsoft 365, Entra ID, or Windows Server and SQL Server licences it wants to reuse.
Azure: your customers buy through Microsoft agreements or require Azure.
Google Cloud: the product centres on BigQuery analytics, GKE or Cloud Run.
Google Cloud: your team already runs Google Cloud projects and knows its tooling.

Which questions settle the choice for one workload?

1. Which cloud does the company already pay for, and who runs it today?

2. Which identity provider signs your staff in, and does the product's user sign-in need to connect to it?

3. Which licences could you bring, and do they apply to the services you would use?

4. Which managed services does the product depend on, and does each have an equivalent on the other clouds?

5. Are those services offered in every region your users and contracts require?

6. Which cloud do your engineers know, and who supports them out of hours?

7. Do your largest customers or partners require a specific cloud?

8. Which parts of the design would you have to rewrite to move?

With those answers on one page, we test the leading option with a short proof of concept on that one workload before anything else moves.

Which sources does this comparison rely on?

Cloud consulting services AWS migration services Azure migration services Healthcare cloud selection guide Cloud AI platform selection Cloud cost optimization

Frequently Asked Questions

None is best in general, and we do not rank them. Each offers comparable core services. The right choice depends on the estate you already run, the managed services your product needs, regions, customer requirements and your team's skills.