Say, you hired 200 remote analysts. Now, instead of commissioning 200 high-spec laptops, you just get them to log into a cloud-hosted desktop from their own device.
All sensitive data and programs remain safe within Azure, and your IT team stays on top of security, updates, and accessibility. Sounds like a convenient, productive decision.
But then the bill arrives. No numbers make sense in terms of desktop usage or cost per employee. Instead of actual business activity and AVD usage patterns contributing to costs, you see charges for:
- Virtual machines (VM)
- Managed disks
- Azure Files storage
- Network bandwidth
- Snapshots and backups
Well, that’s the reality of Azure cost management. It doesn’t tell you where the money is going or who is spending it, which makes tracking (let alone controlling) AVD costs very difficult.
But this article will help.
By the end, you’ll know what’s contributing to the cost, where the money’s leaking, and how to fix it.
But first, understand what makes AVD cost management so complex
AVD costs don’t come from a single item. It accumulates from the charges of the underlying Azure resource consumption.
So, when you look at native reports, you see charges for components like:
- VM compute
- Storage
- Networking
- Backup
- Licensing
And while we love that AVD billing scales up and down with its pay-as-you-go model, it also makes it volatile and unpredictable.
Let’s discuss the cost drivers behind these complex bills.
The big cost drivers of your AVD bills
When you break down any AVD bill, the following four levers commonly stand out as major contributors.
Session host VMs
The largest contributor is of session-host VMs (compute), which account for 35-75% of spend. Yes, we know it’s a wide range, but it depends on the type and size of the Azure Virtual Machines you’re using, the purchased model, and how well they’re optimized.
For example, a D8 VM can support 20 users, but if only 8 users are usually active, you’re still paying for all 20.
AVD session hosts stay on even without active sessions. So, you might be paying for costs from idle VMs, which charge you for CPU, RAM, or storage even when no one is actively using the desktop.
Storage
You’re paying for storage in two places in every AVD environment:
VM operating system (OS) disks: Stores Windows OS, system files, and installed applications.
In every session, the host VM has one operating system disk.
Therefore, if you’re running 20 session hosts, you will pay for 20 OS disks.
Note: If your AVD environment uses Premium SSD managed disks for better performance, it’ll cost more than standard disks.
User profile storage (Azure Files/NetApp): AVD stores each user profile as a virtual disk file (FSLogix profile container) containing their desktop settings, files, and other data.
The size of virtual disk files increases with the profile, which in turn grows with user data (Outlook, OneDrive sync files, browser caches, and temporary application data). As per the Configuration Setting Reference, FSLogix profile containers can grow up to 30GB.
So, say your 200 new remote analysts’ profiles average at 5-10 GB (low-balling it, here). The storage footprint will quickly grow to several terabytes, significantly increasing your monthly storage costs.
Networking
Networking charges aren’t as significant as compute and storage, which is why they go unnoticed as the environment scales. These charges often come from data egress (the cost of moving data out of the Azure environment) and fixed VPN/ExpressRoute cost.
Networking costs grow quickly in Azure environments where:
- Teams need to download or upload files often
- Applications transfer a lot of data
- AVD workloads are graphics or video-heavy
- Employees all over the world connect
Licensing
This one’s often the most confusing one.
There’s no AVD license fee, and its access rights often come with many Microsoft 365 and Windows Enterprise licenses. This means your employees can log in to AVD without a software license.
Now, you might wonder, “We already have the license, so AVD should be cheap.” But what you’re missing out on is that the licensing doesn’t cover infrastructure costs.
In other words, while you may not pay extra for Windows OS, costs accumulate for every hour of VM compute, storage space, and network used by your desktops.
Cracks through which AVD money quietly slips
If your AVD bills are soaring but you aren’t sure how the above-mentioned cost drivers are draining your money, then the following practices might be the cracks that need to be fixed.
- Always-on VMs and idle sessions: If many of your users are not logging off at the end of the workday but simply disconnecting, then VMs are costing you money without doing any work. Some of these sessions hang in host memories, lasting through the nights and even weekends.
- Improper auto-scaling configurations: If you haven’t implemented autoscaling correctly, you’re either paying for empty, idle host pools or spinning up extra VMs that no one is actually using.
- Overprovisioning VMs: Engineers often use larger VMs “just in case.” If that’s the case with your employees, you’re actually paying for resources that you don’t even need.
- No cost allocation visibility: Unassigned expenses often grow when not checked regularly. If your teams aren’t tagging resources, you don’t know who’s contributing to the spend. That means you lack accountability.
Where native Azure cost management falls short
Azure’s cost management tools are a good start to managing cloud costs. However, it isn’t helpful for tracking and controlling AVD costs. Here’s why:
- It doesn’t offer granularity: While Azure Cost Management shows how much you spent at the subscription level and at the resource group total, it won’t show who spent it. So, no cost per-user session or per department.
- Reactive nature: The Azure native suite only reports what you spent during the billing period, but it won’t do anything to prevent waste during it. Azure Advisor can show you idle VMs or suggest reserved instances, but it can’t automatically trigger resizing or session host shutdowns based on usage patterns.
- Complex learning curve: The effort required to obtain AVD-related insights from Azure’s native tools is significant. Tools like Azure Monitor and Log Analytics give only raw infrastructure data. To make that data useful, your team needs to:
- Run custom KQL queries
- Create dashboards
- Implement extensive tagging strategies
What we’re getting at: These tools don’t have AVD-specific behavior awareness. It can’t tell a session host from a regular VM, and it doesn’t automatically account for patterns such as disconnected user sessions, logon spikes, or host draining during maintenance.
What you really need to optimize AVD spend
By now, you know that AVD spend can’t be optimized by Azure’s native tools alone. But here’s another reality: If your organization’s AVD spend is all over the place, it’s not just the finance team’s responsibility to fix it. Nor can the IT team work alone and get it in line.
Traditionally, finance teams would conduct a monthly review or generate an ad hoc report in the hope of identifying AVD budget drainers and fixing them. But cloud workloads are dynamic. By the time a monthly report recognizes idle instances, you’ve already incurred its cost.
To implement FinOps effectively for cloud spend, your engineering and finance teams need to work together and think beyond Azure-native tools.
You’ll need:
Real-time cost visibility: Track costs down to the exact cost centers. For example, the X department spends $Y on IT desktops.
What you need to do: All teams using cloud resources must tag cloud resources with a standardized labelling practice. These tags can be cost centers, departments, projects, environments, owners, etc.
Accurate, context-rich forecasting: Use AI/ML predictive capabilities that consider usage patterns, planned changes, and seasonality to create more accurate budgets. So, when those 200 new analysts start using AVD, you should already anticipate the numbers that show up on the bill.
What you need to do: Use a proper FinOps platform that analyzes historic trends and aligns them with future business realities.
Automations: Triggers that schedule actions to optimize spends like VM shutdowns after hours or rightsizing recommendations as per resource usage, etc.
What you need to do: Create a system with triggers that act on insights.
Cost accountability: Every dollar spent must be tied back to the company’s goals and objectives.
What you need to do: Track metrics like cost per active user, cost per transaction, or cost per business unit.
Bridging the gap left by the native Azure tools
Azure-native tools pose limitations, creating a gap that requires a FinOps layer to bridge and truly manage AVD costs. That’s where Turbo360 comes into the picture.
Turbo360 is the only platform you need for holistic Azure cost management. It ingests Azure usage data and applies business-level cost allocation to present you with optimization opportunities that native tools can’t surface on their own.
Here’s what makes Turbo360 a value addition for FinOps:
- Get real-time visibility across subscriptions and environments: Get granular data of Azure spend across tenants, subscriptions, and resource groups under a single dashboard.
- Make cost allocation easy: Track all the cloud spend back to products, teams, environments, or customers, and make unit economics measurable. Get a clear understanding of how it contributes to your business needs.
- Stop anomalies before they escalate: Get alerts for unusual spending patterns and investigate costs early, before they appear on your bills.
- Get an accurate forecast: Use historical data and trend-based forecasting to better anticipate runway and revenue alignment.
- Continuously optimize: Identify idle resources, underutilized infrastructure, and pricing inefficiencies early.
Turbo360 helps your organization to move from reactive reports to proactive FinOps. CFOs get the dashboards they need, and your IT gets all the tools to implement optimization strategies. Book a Turbo360 demo to see how you can turn your AVD spend into a strategic investment rather than a cost variable.
