You need to put a new platform in the cloud, or move an old one there, and someone has to pick a provider. The comparison articles you’ll find count services and regions. That matters less than you think. For an enterprise, the choice is mostly about what you already own: your licences, your identity system, your team’s skills and your contracts.

AWS vs. Azure compared

Where does Azure win?

Microsoft licences you already pay for. With Azure Hybrid Benefit you can reuse Windows Server and SQL Server licences that have Software Assurance. On a large Windows and SQL estate, that is often the single biggest line in the cost comparison. Moving the same servers to AWS usually means paying for those licences again.

One identity system. Most enterprises already manage staff in Microsoft Entra ID (formerly Azure Active Directory). On Azure, the same accounts, groups and conditional access rules control access to cloud resources. You can connect Entra ID to AWS too, but you then run two permission models and need people who understand both.

Data and BI in the Microsoft stack. If your analysts live in Power BI and Excel, Azure keeps data, reports and permissions in one place. Azure Databricks is sold and supported by Microsoft and billed on your Azure invoice. Our Microsoft Fabric vs. Databricks guide covers which data platform fits.

Old Windows servers. Extended security updates for older Windows Server and SQL Server versions come free on Azure. That buys time to modernise instead of paying for support you’ll throw away.

Where does AWS win?

Breadth and maturity of services. AWS has been in the market longest and offers the widest range of managed services. If you need a niche database, a specific IoT or media service, or very large scale, AWS usually has a well-documented option.

Linux and open-source teams. If your applications run on Linux, containers and open-source databases, Microsoft licences give Azure no advantage. Many engineers learned cloud on AWS, so hiring and onboarding can be easier.

Ecosystem. Many SaaS vendors build on AWS first and offer AWS Marketplace billing, private networking to their service and reference architectures that assume AWS. If your critical vendors sit there, staying close to them cuts network cost and complexity.

Is AWS or Azure cheaper?

Neither, as a rule. List prices for comparable virtual machines and storage are close, and both providers change them often. Four things move your bill far more:

1. Licences. Reusing Microsoft licences on Azure can cut Windows and SQL Server costs sharply. With no Microsoft licences, this advantage disappears.

2. Commitments. Both offer one and three year commitments with discounts that dwarf the list price gap. Commit only after you’ve measured real usage for a few months.

3. Data transfer. Both charge for data leaving the cloud. Architectures that move data between providers, or out to users at volume, pay for it every month. Both now waive these fees if you move all your data out to switch providers.

4. Architecture. A lift-and-shift of oversized servers costs more on either cloud than a right-sized, autoscaling design. This is where most cloud bills go wrong, regardless of provider.

Ask both providers, or a partner, for a priced estimate of your actual workloads, with your licences and expected commitments included. Never compare on list prices.

What about AI workloads?

The gap has narrowed. Azure got an early lead through its OpenAI partnership, and AWS built Amazon Bedrock around a choice of model vendors. Today both platforms offer models from several vendors, so the model you want is rarely the deciding factor. Choose the cloud where your data already lives. Moving large datasets to reach a model costs more, and creates more compliance work, than the model itself.

What about EU compliance and data residency?

Both providers run multiple regions in the EU, hold the common certifications (ISO 27001, SOC 2, C5) and sign data processing agreements that cover GDPR. Both have answered European sovereignty concerns: Microsoft with its EU Data Boundary, AWS with its European Sovereign Cloud.

For regulated industries, the provider matters less than how you configure it. Private networking, key management, logging and access control decide whether you pass an audit. If NIS2 or DORA applies to you, document your exit plan and concentration risk whichever cloud you pick.

Two examples from our work

Azure. A U.S. human services organization ran case management on fragmented manual tools and needed a secure platform that could handle sensitive health data. Their staff already used Microsoft accounts. We built a serverless platform on Azure Functions, Azure SQL and Cosmos DB, with Azure AD B2C for sign-in and role-based access control. Staying on Azure kept identity, security and billing in one place. A related project for the same sector put a HIPAA-aligned data platform on Azure and Databricks, fully private, with no public endpoints.

AWS. A Nordic consumer audio brand needed a cloud backend for its wireless speakers and headphones, with no Microsoft estate to consider. We built it on AWS Lambda, API Gateway, S3, SNS and CloudFront, with over-the-air updates for every hardware generation. The team was Linux and Node.js based, and AWS’s managed services covered everything the product needed.

Same consultancy, different clouds. The existing estate decided both.

How do you decide?

Answer these five questions:

1. How much of your estate runs on Windows Server, SQL Server or .NET? A large share points to Azure.

2. Where do your users and permissions live? Entra ID and Microsoft 365 point to Azure. A separate identity provider makes this neutral.

3. What does your team already know? Retraining a whole team costs more than most price differences. Count certifications and production experience, not preferences.

4. Where do your data and key vendors sit? Keep workloads close to the data and the SaaS products they talk to.

5. Do you have an enterprise agreement or committed spend? An existing Microsoft Enterprise Agreement or AWS commitment often settles the question on price alone.

If the answers split, a multi-cloud setup is possible, but it doubles the skills, security policies and tooling you need. Most mid-sized enterprises are better off choosing one primary cloud and using the other only for a clear reason.

How RUBICON helps you choose and build

We build on both clouds. We’re a Microsoft Solutions Partner for Cloud & AI Platforms and a Databricks Partner, and we’ve delivered production systems on AWS for IoT and consumer products. We’ll tell you which cloud fits your estate, even when it isn’t the one we’d find easier to sell.

A typical cloud engagement starts with a two to four week assessment: your workloads, licences, identity setup and compliance needs, priced on both providers where it’s a close call. Then we build the landing zone with Terraform and migrate or build workloads in waves. We’re about 55 people in Sarajevo, 40+ of them engineers, ISO 27001:2022 and ISO 9001:2015 certified.

Frequently asked questions