AWS vs Azure: how to actually choose between them

Ask a room of engineers about AWS vs Azure and you will get a technical argument. Watch how the decision is actually made in most organisations and it turns on three things: what your identity system already is, what software you are already licensed for, and who your procurement team has an agreement with.

That is not cynicism. Those factors have larger cost and effort consequences than the differences between the platforms, which have converged considerably.

Where they are genuinely the same

Both offer virtual machines across general purpose, compute optimised, memory optimised, storage optimised and accelerated families, on Intel, AMD and Arm silicon. Both offer object storage with tiered pricing, managed relational and NoSQL databases, managed Kubernetes, serverless functions, message queues, CDN, global load balancing and a large security tooling estate.

Both bill compute per second with a minimum, both discount heavily for commitments, both charge for outbound data transfer, and both enforce per account quotas that start conservative. If you are comparing a virtual machine to a virtual machine, you are comparing very similar things at very similar prices.

Where they genuinely differ

Identity. This is the biggest single difference in practice. Azure is built on Microsoft Entra ID, which is already the identity provider for a large share of organisations. If your staff already sign in with Microsoft accounts, Azure inherits that and the integration is deep rather than bolted on. AWS has IAM Identity Center, which federates well, but it is federating to something rather than being the something.

Licensing. If you run Windows Server or SQL Server, Azure Hybrid Benefit lets you apply licences you already own and remove the licence component from the virtual machine rate. On a Windows estate this is frequently the largest single line on the comparison, and it has no equivalent of the same weight on AWS.

Service catalogue depth. AWS is broader and generally first to a category. If you want a specific managed service for an unusual requirement, AWS is more likely to have it. The counterpoint is that most teams use fifteen services, not two hundred, and both platforms have the fifteen.

Account structure. AWS uses accounts as the isolation boundary, with Organizations grouping them. Azure separates the identity tenant from billing subscriptions from resource groups. The Azure model is more granular; the AWS model has a harder boundary. Neither is better, but they lead to different governance designs.

Documentation and community. AWS has more third party material, more Stack Overflow answers and more tooling written against it first. Microsoft Learn is well structured and improving quickly. This gap is smaller than it was and still real.

Pricing structure, not price

Comparing list rates is close to useless because the discount mechanics differ:

• AWS offers savings plans, which commit to an hourly spend, and reserved instances, which commit to specific capacity. Spot covers interruptible work.
• Azure offers reservations for specific series and a savings plan for compute that applies flexibly, plus spot virtual machines and Hybrid Benefit on top.

Both reward committing to the baseline and floating the peak. The detail is in how Azure pricing works and cloud pricing models compared.

Egress is the other variable that moves the total, and it moves it a long way for content heavy workloads. Both charge for it, both include a small monthly allowance, and both are far more expensive than the European providers. That comparison is in data transfer costs.

The three questions that usually settle AWS vs Azure

1. What is your identity provider today? If the answer is Microsoft Entra ID and it is not changing, Azure removes an integration project.
2. Do you hold Windows or SQL Server licences? If yes, run the Hybrid Benefit numbers before anything else. They are often decisive.
3. Does the workload need a service only one of them has? Rare, but when true it ends the discussion. Check the specific service rather than the marketing comparison.

If all three come back neutral, pick the one your team already knows. Familiarity is worth more than a few percent on the invoice, and the operational cost of a platform nobody understands is paid every week.

Where each one is the clear answer

Choose Azure for a Microsoft centric organisation, a Windows and SQL Server estate, hybrid deployments with on premises Active Directory, or where procurement already has an enterprise agreement.

Choose AWS for the broadest service catalogue, the largest talent pool, the deepest third party tooling support, or where a specific service you need only exists there.

Choose neither if your workload is a handful of Linux servers with predictable traffic. Both platforms carry complexity you are paying for, and the providers covered in the Hetzner guide and the DigitalOcean guide will do that job for a fraction of the cost and a fraction of the learning curve.

For the third of the big three, see AWS vs Google Cloud and Azure vs Google Cloud.