AWS vs Azure vs Google Cloud: how a mid-size company should choose
For a mid-size company, AWS, Azure and Google Cloud can all run the same web applications, databases and internal systems, so the choice is rarely decided by features. It is decided by what you already own and know: your identity system, your Microsoft licences, your team's skills, your data and analytics plans, and where your customers and regulators require data to sit. Pick one primary provider on those factors, set up accounts and access properly from day one, and add a second provider only for a specific workload that needs it.

What the three providers have in common
Amazon Web Services, Microsoft Azure and Google Cloud sell the same basic building blocks: virtual machines, managed databases, object storage, private networks, load balancers, container platforms and serverless functions. All three bill by use, all three operate regions around the world, and all three offer discounts when you commit to a level of spend or capacity for a period.
All three also work on a shared responsibility model. The provider secures the buildings, the hardware and the platform. You are responsible for what you put on it: who has access, how the network is opened, whether backups exist and whether the data is encrypted. A badly configured account is equally exposed on any of them.
For a mid-size company, that means the question “which cloud has the best technology” has no useful answer. The useful question is which provider fits what you already have.
Six factors that decide the choice
1. Your identity system and Microsoft estate
If staff sign in with Microsoft 365, your directory is already Microsoft Entra ID. Azure uses the same directory for access to cloud resources, so joiners, leavers, multi-factor authentication and conditional access are managed in one place. Microsoft also lets customers with eligible Windows Server and SQL Server licences apply them to Azure through the Azure Hybrid Benefit, which changes the cost comparison for Windows-heavy estates. Check the current terms with Microsoft or your licence provider before relying on it.
A company that runs on Google Workspace has the same kind of head start with Google Cloud, where Workspace identities and groups can be granted roles directly.
AWS connects to either directory through IAM Identity Center, so an outside identity system is not a blocker. It is one more thing to set up.
2. The skills you have or can hire
A provider your engineers already know is worth more than a slightly better service on a provider they do not. Each platform has its own vocabulary and its own traps: AWS accounts and IAM policies, Azure subscriptions and role assignments, Google Cloud projects and IAM bindings. Mistakes made while learning are usually security or cost mistakes.
If you have no cloud skills in-house, look at who you can hire locally and which platform your software vendors and integrators work on.
3. The workloads themselves
- Line-of-business applications on Windows Server and SQL Server. Azure is the natural first candidate, for the licensing and identity reasons above.
- A custom web or mobile product. All three fit. AWS has the widest catalog of services and the largest pool of engineers who have used it. Google Cloud’s Cloud Run and Azure Container Apps run containers with very little operations work, and AWS offers ECS for the same purpose.
- Analytics on large datasets. Google Cloud’s BigQuery is a common reason companies choose Google Cloud for a data workload, including companies whose applications run elsewhere.
- Kubernetes. All three offer a managed control plane: EKS on AWS, AKS on Azure and GKE on Google Cloud. Whether you need a cluster at all is a separate question, covered in do you need Kubernetes.
4. Where data must live
Customers, regulators and contracts sometimes require data to stay in a country or region. Check each provider’s current region list for the locations you need, and check that the specific services you plan to use are available in that region. Not every service is offered everywhere.
5. What stays on-premises
Few mid-size companies move everything. Production-line systems, file servers at remote sites and software tied to specific hardware often stay. Each provider has a way to connect a site by VPN or private line and to extend management to servers you keep. The pattern is explained in what is hybrid cloud. If a large share of the estate will stay on site, weigh how well the provider’s identity and management tools reach those servers.
6. Commercial terms
All three sell commitments: Savings Plans and Reserved Instances on AWS, reservations and savings plans on Azure, committed use discounts on Google Cloud. All three run credit programs for new customers and negotiate on larger contracts. Ask each for a quote on your real workload, including support plans and data transfer, and compare totals over the same period. We do not quote provider prices here because they change and depend on region and commitment.
AWS vs Azure vs Google Cloud side by side
| AWS | Azure | Google Cloud | |
|---|---|---|---|
| Top-level structure | Accounts grouped under AWS Organizations | Subscriptions grouped under management groups | Projects grouped in folders under an organization |
| Identity | IAM roles and policies, staff sign-in through IAM Identity Center | Microsoft Entra ID, the same directory as Microsoft 365 | Cloud IAM, with Google Workspace or Cloud Identity accounts |
| Managed Kubernetes | EKS | AKS | GKE |
| Containers without a cluster | ECS, Lambda for functions | Azure Container Apps, App Service | Cloud Run |
| Commitment discounts | Savings Plans, Reserved Instances | Reservations, savings plans, Azure Hybrid Benefit for eligible licences | Committed use discounts |
| Vendor’s design guidance | AWS Well-Architected Framework, six pillars | Azure Well-Architected Framework | Google Cloud Well-Architected Framework |
| Often the first candidate when | The product team already knows it, or you need a broad service catalog | The company runs on Microsoft 365, Windows Server and SQL Server | The main workload is analytics, or the company runs on Google Workspace |
The table shows equivalents, not winners. Each row exists on all three platforms under a different name.
A scoring method you can run in an hour
- List your workloads in a spreadsheet: the application, its operating system, its database, who uses it and where its data must sit.
- Add one column per provider.
- For each workload, score each provider from 0 to 2 on four questions: does our identity system connect directly, does our team know it, do we own licences that apply, and is the service available in the region we need.
- Add the scores. Weight the workloads by how important they are to the business.
- Note any workload where one provider scores far above the rest. That is a candidate for a second provider later, not a reason to split everything now.
In most mid-size companies one provider comes out clearly ahead, and the reason is usually identity, licences or skills.
One provider or several
Multi-cloud sounds like protection against lock-in. In practice it doubles the surface you have to secure and understand. Two identity models, two network designs, two bills and two sets of monitoring are a real cost for a team of two or three people.
A better protection against lock-in is portability inside one provider: applications packaged in containers, standard databases such as PostgreSQL, and infrastructure described as code. Those choices keep a later move affordable without paying for a second platform today.
A second provider is justified when a specific workload needs it: an analytics platform, a customer who requires their data on a named cloud, or a company you acquire.
Mistakes that cost more than the choice of provider
The setup matters more than the logo. These are the problems a review finds on every platform.
- Everything in one account, subscription or project. Production, staging and experiments share one space, so a test mistake can reach live data.
- Administrator rights for everyone. Long-lived access keys with full rights sit on laptops and in CI secrets.
- Resources created by hand in the console. Nobody can say why a firewall rule exists or rebuild the environment after a loss.
- Backups kept beside the data. A compromised account loses both.
- A bill nobody reads. No tags, no owners and no budget alerts, so a cost increase is noticed months later.
Each has the same fix on all three providers: separate environments, single sign-on with least privilege, infrastructure as code, backups held outside the production account, and tagging with budgets.
What we do, and what we do not claim
Goldenroz is not an AWS, Microsoft or Google Cloud partner. We hold no vendor certifications and have no published cloud case studies. We do not resell cloud capacity or licences, so we earn nothing from the platform you pick.
What we offer is a paid review followed by fixed-price phases, led by Shai Goldenberg, our CTO. If you have not chosen a provider, cloud consulting services produce a written target architecture with at least two options compared on cost and operating effort. If you have chosen, the platform pages describe the work on each: AWS consulting covers account structure, IAM, networking and deploy pipelines; Azure consulting covers management groups, Entra ID access, networking and pipelines as code; and Google Cloud consulting covers project and IAM structure, CI/CD, runtime choice and cost review.
What to do next
Build the workload list and run the scoring method above. If one provider wins clearly, open the organization and the separate production and staging environments before the first workload moves. If the result is close, or the decision carries a multi-year commitment, our cloud consulting services start with an architecture review and end with a written design your team can run. Projects start at $25,000 at a fixed price per phase, and the first one-hour scoping call is free.