If Azure Storage is the most-confused topic on the AZ-900, compute is the most foundational. It sits in the “Describe Azure architecture and services” domain — the largest slice of the exam at roughly 35–40% of your marks — and the questions almost never ask you to configure anything. They ask you to choose: given a workload, which compute service is the right home for it? A team wants to run a Windows Server application unchanged; a developer wants to deploy a web API without touching a server; an event needs to trigger a tiny piece of code that runs a few thousand times a day. Each has one best answer, and the exam is built from these matches.
The trick is to stop treating the services as an unrelated list and see them as points on a single spectrum of control versus management. At one end you run a full virtual machine and manage almost everything; at the other you write a function and Azure manages almost everything. This guide walks that spectrum end to end — Virtual Machines, Scale Sets, App Service, Containers (ACI and AKS), Azure Functions, and Azure Virtual Desktop — then adds the pricing models and a decision table so you can map any scenario to the right service under exam pressure. Everything here is fundamentals-level: what each service is and when to choose it, not how to administer it. (If you want the deeper, hands-on administration angle, that’s the AZ-104 exam’s territory, and we’ll point you there at the end.)
The Compute Spectrum: From Full Control to Fully Managed
Every Azure compute choice is a trade between how much control you want and how much you’re willing to manage. The more Azure manages for you, the less flexibility and responsibility you have — and usually the less you pay for idle capacity.
| Model | You manage | Azure manages | Example services |
|---|---|---|---|
| IaaS (Infrastructure as a Service) | OS, runtime, application, patching | Physical hardware, virtualization, datacenter | Virtual Machines, Scale Sets |
| PaaS (Platform as a Service) | Your application and data | OS, runtime, patching, scaling infrastructure | App Service, AKS (managed control plane) |
| Serverless / FaaS | Just your code | Everything else, including scaling to zero | Azure Functions, Container Apps |
This single table answers a surprising number of exam questions. When a scenario stresses full control or running existing software unchanged, you’re at the IaaS end (VMs). When it stresses no infrastructure to manage or just deploy the app, you’re in PaaS or serverless. Keep the spectrum in mind as we walk each service.
Azure Virtual Machines (IaaS)
An Azure Virtual Machine is a computer in the cloud that you control as if it were your own server: you choose the operating system (Windows or Linux), install whatever you like, and you’re responsible for patching, configuration, and security inside the VM. Azure handles the hardware, the hypervisor, and the datacenter beneath it.
VMs are the maximum-control, maximum-responsibility option, and the exam’s cues for them are consistent:
- Lift-and-shift migrations — moving an existing on-premises server to the cloud with no code changes.
- Full OS-level control — you need to install custom software, specific drivers, or legacy dependencies.
- Any workload that doesn’t fit a managed platform — the general-purpose fallback.
You size a VM by choosing a series (general purpose, compute optimized, memory optimized, and so on), and you improve availability by spreading VMs across availability zones (physically separate datacenters in a region) or, within a datacenter, an availability set. For the AZ-900 you don’t need to configure any of this — just recognize that a single VM is not highly available, and that spreading instances across zones is how you survive a datacenter failure.
Azure Virtual Machine Scale Sets
A Virtual Machine Scale Set (VMSS) is a group of identical, load-balanced VMs that Azure can automatically scale in and out based on demand or a schedule. Instead of hand-building and managing a fleet, you define one configuration and let Azure add instances when traffic rises and remove them when it falls.
The exam cue is simple: if a scenario needs many identical VMs that must scale automatically to handle variable load — a web tier during a sale, say — that’s a Scale Set, not a pile of individually managed VMs.
Azure App Service (PaaS)
Azure App Service is a fully managed platform for hosting web apps, REST APIs, and mobile backends. You bring your code (in .NET, Java, Node.js, Python, PHP, and more) and Azure runs it — no operating system to patch, no web server to configure. Built-in features include autoscaling, custom domains and TLS certificates, deployment slots (staging environments you can swap into production), and integrated CI/CD.
App Service is the textbook answer when the requirement is “deploy a web application quickly without managing servers.” Its position on the spectrum is squarely PaaS: you own the app and data; Azure owns everything underneath.
A frequent AZ-900 distinction: choose App Service when you want a managed platform for a standard web app or API, and choose Virtual Machines only when you need OS-level control that App Service doesn’t give you.
Azure Containers: Container Instances (ACI) and AKS
Containers package an application with its dependencies so it runs consistently anywhere. Azure offers two compute homes for them that the AZ-900 tests:
- Azure Container Instances (ACI) is the fastest, simplest way to run a single container without managing any servers or orchestration. You give it an image, it runs, and you’re billed per second. ACI suits simple, isolated, or bursty tasks — a batch job, a build step, a small API — where you don’t need orchestration.
- Azure Kubernetes Service (AKS) is managed Kubernetes for running and orchestrating many containers at scale. Azure manages the Kubernetes control plane for you (the control plane itself is free; you pay for the worker nodes), while you deploy and scale containerized microservices on top. AKS suits complex, multi-container applications that need orchestration, self-healing, and scaling.
The exam cue: a single container or a simple task → ACI; orchestrating many containers / microservices at scale → AKS. (Azure also offers Container Apps, a serverless container platform for microservices; for AZ-900, recognizing ACI and AKS is what matters most.)
Azure Functions (Serverless / FaaS)
Azure Functions is Azure’s serverless, event-driven compute. You write a small piece of code — a function — and it runs in response to a trigger: an HTTP request, a message on a queue, a timer, a new blob in storage, and so on. On the Consumption plan you pay only for the time your code actually runs, and it scales automatically, even down to zero when nothing is happening, so there’s no cost for idle capacity.
This is the far, fully-managed end of the spectrum: you manage only your code. The exam signature is unmistakable — event-driven, runs occasionally, pay only when it executes, no servers to manage. Think image thumbnail generation when a file is uploaded, or a scheduled cleanup job.
Azure Virtual Desktop
Azure Virtual Desktop (AVD) is a desktop and application virtualization service running in Azure. It delivers a full Windows desktop or specific applications to users from the cloud, accessible from almost any device — the classic virtual desktop infrastructure (VDI) use case. If a scenario describes giving remote employees a secure, centrally managed Windows desktop experience without shipping laptops, that’s Azure Virtual Desktop.
Compute Pricing Models You Should Recognize
The AZ-900 also touches how compute is priced. You don’t need numbers — you need to match a pricing model to a goal:
| Pricing model | What it is | Best for |
|---|---|---|
| Pay-as-you-go | Per-second/-minute billing, no commitment | Short-term, unpredictable, or bursty workloads |
| Reserved Instances | 1- or 3-year commitment for a big discount | Steady, predictable, always-on workloads |
| Spot Virtual Machines | Deep discount on Azure’s surplus capacity, can be evicted anytime | Interruptible, fault-tolerant work (batch, dev/test) |
| Azure Hybrid Benefit | Reuse existing on-prem Windows Server / SQL Server licenses | Cutting cost when you already own eligible licenses |
The reliable cues: predictable, always-on → Reserved; interruptible and cheap → Spot; already own licenses → Hybrid Benefit; unpredictable, no commitment → Pay-as-you-go.
Decision Table: Which Compute Service?
| Scenario cue | Answer |
|---|---|
| Migrate an existing server to the cloud, no code changes | Virtual Machine |
| Need full OS-level control / custom software | Virtual Machine |
| Many identical VMs that must autoscale with demand | Virtual Machine Scale Set |
| Deploy a web app or REST API without managing servers | App Service |
| Run a single container quickly, no orchestration | Azure Container Instances |
| Orchestrate many containers / microservices at scale | Azure Kubernetes Service (AKS) |
| Event-driven code, pay only when it runs, scale to zero | Azure Functions |
| Deliver Windows desktops/apps to remote users | Azure Virtual Desktop |
| Predictable, always-on workload, lowest cost | Reserved Instances |
| Interruptible batch work, deepest discount | Spot VMs |
Common AZ-900 Scenarios
Scenario 1 — Legacy line-of-business app. A company wants to move a Windows Server application to Azure without rewriting it. Full OS control, no code changes → Virtual Machines (lift and shift).
Scenario 2 — New customer-facing API. A developer needs to publish a Node.js REST API fast, with autoscaling and staging slots, and doesn’t want to manage servers → App Service.
Scenario 3 — Occasional image processing. Whenever a photo is uploaded to storage, a small function should generate a thumbnail; it runs sporadically and should cost nothing when idle → Azure Functions.
Scenario 4 — Microservices platform. A team is deploying dozens of containerized microservices that need orchestration, self-healing, and scaling → Azure Kubernetes Service (AKS).
Common Mistakes to Avoid
- Defaulting to Virtual Machines for everything. VMs are the most work to manage. If the scenario is a standard web app or API, App Service is usually the better (PaaS) answer.
- Confusing ACI and AKS. A single container or simple task is ACI; orchestrating many is AKS.
- Missing the serverless cue. Event-driven, pay-per-execution, scale to zero points to Azure Functions, not a VM or App Service.
- Forgetting a single VM isn’t highly available. High availability comes from spreading instances across availability zones (or a Scale Set), not from one bigger VM.
- Mixing up the pricing models. Reserved = commit for a discount on steady workloads; Spot = cheap but evictable for interruptible work.
Practice Until the Trade-offs Are Automatic
Reading about the compute spectrum is enough to answer a clean question, but the real AZ-900 mixes cues — a scenario might combine a workload description, a management preference, and a cost constraint in one paragraph and expect you to weigh all three in well under a minute. That fluency comes from repetition on exam-style questions.
That’s where full-length practice pays off. Sailor.sh’s AZ-900: Azure Fundamentals Mock Exam Bundle gives you timed, scenario-based mock exams that mirror the real exam’s format and passing bar, with a detailed explanation on every question so a wrong answer becomes a lesson. Use the free concepts here to build the model; use timed exams to make the compute choice reflexive. Pair this with the sibling guide on Azure Storage for AZ-900 to round out the architecture-and-services domain, start with the AZ-900 exam guide for 2026 for logistics, and follow the 30-day AZ-900 study plan to schedule your prep. When you’re ready to go deeper into hands-on administration of these compute services, the AZ-104 exam guide is the natural next step.
Frequently Asked Questions
What is the difference between IaaS, PaaS, and serverless in Azure?
They differ in how much you manage. With IaaS (Virtual Machines) you manage the OS, runtime, and application. With PaaS (App Service, and AKS’s managed control plane) Azure manages the OS and infrastructure and you manage just the app. With serverless (Azure Functions) you manage only your code, and Azure handles everything else including scaling to zero.
When should I use App Service instead of a Virtual Machine?
Choose App Service when you want to deploy a web app or API quickly on a fully managed platform — no OS to patch, built-in autoscaling and deployment slots. Choose a Virtual Machine only when you need OS-level control that App Service can’t provide, such as installing custom software or running a legacy dependency.
What’s the difference between Azure Container Instances and AKS?
Azure Container Instances (ACI) is the fastest way to run a single container without managing servers or orchestration — ideal for simple or bursty tasks. Azure Kubernetes Service (AKS) is managed Kubernetes for orchestrating many containers at scale, with self-healing and scaling; Azure manages the control plane while you run the worker nodes.
What makes Azure Functions “serverless”?
With Azure Functions you deploy only your code, it runs in response to a trigger (HTTP, timer, queue message, new blob, etc.), and on the Consumption plan it scales automatically and bills only for actual execution time — even scaling to zero when idle so you pay nothing for unused capacity.
What are Spot Virtual Machines used for?
Spot VMs run on Azure’s surplus capacity at a steep discount but can be evicted whenever Azure needs the capacity back. They suit interruptible, fault-tolerant workloads such as batch processing, rendering, or dev/test — never workloads that must stay up.
Do I need to configure these compute services for the AZ-900?
No. AZ-900 is a fundamentals exam: you’re expected to know what each service is and when to choose it, not how to deploy or administer it. The hands-on configuration and administration of VMs, App Service, AKS, and the rest is assessed on the role-based AZ-104 (Azure Administrator) exam.
Conclusion
Azure’s compute portfolio looks broad, but it lines up on one axis: how much you control versus how much Azure manages. Virtual Machines and Scale Sets give you IaaS control for lift-and-shift and custom workloads; App Service is managed PaaS for web apps and APIs; Container Instances run single containers fast while AKS orchestrates many at scale; Azure Functions is serverless, event-driven code that scales to zero; and Azure Virtual Desktop delivers cloud desktops. Layer the pricing models on top — pay-as-you-go, Reserved, Spot, and Hybrid Benefit — and you can answer both the “which service?” and “which price?” halves of a scenario. Anchor every question to the control-versus-management spectrum, watch for the serverless and containers cues, and Azure compute becomes one of the most reliable sources of points on the AZ-900 — and a mental model you’ll reuse every time you design something real in Azure.