Microsoft Azure: Pros, Cons, and Services

Choose Azure if your business already runs on Microsoft licences. That is the honest version of a decision most comparison articles bury three thousand words down, under a feature list nobody reads. If your staff sign in with Microsoft 365, your main line-of-business application sits on SQL Server, and your finance team already holds an Enterprise Agreement, Azure removes work the other clouds hand straight back to you. If none of that is true, the case gets thinner, and we would rather tell you that now than after you have signed a three-year reservation.

Why Azure wins on paper

The market numbers are not the reason to pick Azure, but they do tell you the platform is not going anywhere. Synergy Research Group, now part of TechInsights, put second-quarter 2026 enterprise spending on cloud infrastructure services at $143 billion, a jump of more than 43% on the same quarter a year earlier. Microsoft took 20% of that spend. AWS held 28% and Google Cloud reached 15%, its highest share on record. Between them the three providers account for 67% of everything enterprises spend on cloud infrastructure.

What that share does not explain is why businesses land on Azure. In our experience it is rarely a bake-off won on price per vCPU. It is that the company already pays Microsoft. Entra ID is already holding every staff identity. Conditional Access policies already decide who can read the finance folder from an unmanaged laptop. Group Policy equivalents already exist in Intune. Putting workloads in Azure means those controls extend to the servers instead of being rebuilt in a second identity provider.

That matters more than it sounds. The expensive part of a cloud migration is almost never the virtual machines. It is the identity plumbing, the network routes back to the office, the compliance evidence and the retraining. Azure shortens three of those four for a Microsoft-centred business. For a shop running Linux, Postgres and Okta, it shortens none of them, and you should read our walk through AWS before committing.

The service groups that actually matter

Azure publishes hundreds of services. Most businesses we work with use eight. These are the groups that carry real weight in a small or mid-sized estate:

• Compute. Virtual Machines for anything you are lifting across as-is, App Service for web applications you would rather not patch, Azure Kubernetes Service when you genuinely have containers in production, and Functions for the glue code.
• Data. Azure SQL Database is the one that pulls Microsoft shops over, because it takes an existing SQL Server schema with very little argument. Azure Database for PostgreSQL Flexible Server and Cosmos DB cover the rest.
• Identity. Microsoft Entra ID. If you only adopt one Azure service, this is usually it, and it often arrives through a Microsoft 365 subscription rather than a cloud project.
• Hybrid. Azure Arc projects on-premises and other-cloud servers into the Azure control plane so policy and patching work in one place. ExpressRoute gives you a private circuit when the internet is not good enough.
• Storage. Blob Storage with hot, cool, cold and archive tiers, plus managed disks for the VMs. Microsoft designs managed disks for 99.999% availability by keeping three replicas of your data.
• AI. Azure AI Foundry and the Azure OpenAI models, which is where a lot of 2026 budget quietly went.

What Azure costs when the invoice lands

Here is the comparison people actually want: a plain 2 vCPU, 8 GiB virtual machine, which is the shape of most small business application servers. The figures below are published list prices for East US captured on 3 September 2026 by the pricing tracker Economize, cross-checked against Microsoft’s and Amazon’s own pricing pages. They move, so treat them as a shape rather than a quote.

Two things jump out. The Windows licence more than doubles the hourly rate, and that is where Azure Hybrid Benefit earns its keep. Microsoft lets you apply Windows Server and SQL Server licences you already own to Azure compute, so you pay only the base Linux rate. Microsoft’s own pricing page claims up to 76% off pay-as-you-go for Linux through the equivalent benefit, and says that stacking Hybrid Benefit with a Reserved VM Instance can save up to 80% against the standard pay-as-you-go rate for Windows Server. For comparison, buying Windows Server 2025 on a pay-as-you-go basis from Microsoft lists at $33.58 per CPU core per month, or $0.046 per core-hour.

The other cost line is bandwidth, and it is the one that ruins forecasts. Data going into Azure is free. Traffic inside a single availability zone is free. Everything else meters. These are Microsoft’s published rates for internet egress out of North America or Europe, routed over the Microsoft premium global network:

Cross-region transfer is cheaper but still real: $0.02 per GB between regions inside Europe, $0.08 per GB between regions inside the Middle East and Africa, and $0.16 per GB between regions inside South America. Moving data from North America or Europe to another continent costs $0.05 per GB. If your architecture chats across regions on every request, you will find out about it in month two.

Regions, zones and data residency

Microsoft’s documentation states that Azure provides over 70 regions globally, more than any other provider publishes. Each region sits inside a single geography that acts as a fixed data residency boundary, and Microsoft says every geography has at least one region with availability zones. That combination is the quiet reason Azure wins regulated tenders: you can usually keep data where a contract says it must stay without giving up zone redundancy.

Availability zones are independent sets of datacentres inside one region with their own power, cooling and network. Microsoft says a zone-enabled region connects at least three physically distinct datacentres, chosen against more than 30 viability criteria and risk factors, and holds a latency perimeter of under two milliseconds between zones. That last number is what makes synchronous replication across zones practical rather than theoretical.

The SLA follows the architecture, not the brochure. Microsoft commits to 99.99% uptime for virtual machines only when you run two or more instances across two or more availability zones in a region. Put two VMs in an availability set inside one datacentre and the commitment drops to 99.95%. Run one VM and you are relying on the single-instance commitment, which is a different and much weaker promise. If somebody has sold you “four nines” on a single server, the SLA does not say that.

Residency is getting more attention, not less. Gartner forecast worldwide sovereign cloud infrastructure-as-a-service spending to reach $80 billion in 2026, a 35.6% increase on 2025. And from 12 January 2027 the EU Data Act prohibits providers from charging switching fees, including data egress charges, for moving to another provider. If you are negotiating a long contract now, that date is worth putting in the room.

Where Azure frustrates people

Azure is not a tidy platform, and pretending otherwise does nobody any favours. The portal is enormous and inconsistent; two services built by two teams will ask you the same question in two different shapes. Product names change often enough that documentation and blog posts go stale while still ranking well in search. Quota limits are set per subscription and per region, so the first time you try to scale on a Friday afternoon you may be filing a support request instead.

Service quality across the portfolio is uneven. The core, Virtual Machines, Blob Storage, Azure SQL, Entra ID, is solid and well documented. Further out, you will meet services where the portal, the CLI and the Terraform provider each believe something slightly different. Budget time for that. Reserved instances and savings plans add another layer: the savings are genuine, but they are committed spend, and an unused reservation is just an expensive subscription.

That 29% figure is not an Azure problem specifically, it is an everyone problem, and the same report found 73% of respondents now run hybrid estates. But it does mean the difference between a well-run Azure subscription and a neglected one is roughly a third of the bill. Tagging, budgets, alerts and a quarterly review are not optional extras.

Who should pick Azure, and who should not

Pick Azure when the Microsoft estate is already the centre of gravity: Microsoft 365 for email and files, Entra ID for identity, SQL Server or Windows Server licences you own outright, a Windows-heavy desktop fleet, or a compliance requirement that a named geography can satisfy. Pick it also when you need a hybrid control plane, because Azure Arc is genuinely good at pulling servers you cannot move into the same policy and patching view as the ones you did.

Look elsewhere when your stack is Linux, Postgres and containers with no Microsoft licences to redeploy, when you want the widest managed-service catalogue and the deepest third-party tooling, or when your team already knows another platform well. Operator familiarity beats a 12% price difference almost every time. If your market is Asia or Africa and you care about regional presence and price, Huawei Cloud is worth a look. If your spend is dominated by an Oracle database estate, Oracle Cloud can be cheaper than running the same licences anywhere else.

Whichever way you go, do not let the decision stay abstract. Price one real workload on two platforms, including storage, bandwidth, backup and the licences, then compare the totals. Eudora runs that exercise remotely for clients worldwide as part of our cloud solutions work, and the answer is not always the one the client expected.

Weighing Azure against another platform and want the numbers run on your actual workload rather than a brochure? Get in touch with Eudora Technology to talk about your project.