The pricing and total cost of ownership calculators

Dana Whitcombe's question about the October invoice had a second half. "Fine, I understand why October was high," she told Renata a few days later. "But the board approved Skylight on a business case that said Azure would cost less than replacing the SAN and the server room. I need to know that is still true, and I need a monthly number I can put in next year's budget." Those are two different questions. The first is about total cost of ownership: is running the workloads in Azure cheaper over several years than running them on premises? The second is about an estimate: what will a specific set of Azure resources cost each month? Microsoft provides tools for both, and Theo spent the last week of October learning to use them properly.

The Azure pricing calculator

The Azure pricing calculator is a free web tool for estimating what Azure resources will cost. You do not need an Azure subscription to use it. You choose products from a list of Azure services, add them to an estimate, and configure each one: the region, the tier or size, the number of instances, the hours per month, the storage capacity and redundancy, the expected data transfer and any licensing options. The calculator multiplies the published rates by your inputs and shows a monthly total for each item and for the whole estimate.

The calculator models the cost factors from the previous lesson directly. For a virtual machine, you pick the region, the operating system, the size and the number of hours, then choose between pay-as-you-go, a savings plan or a one- or three-year reservation, and whether Azure Hybrid Benefit applies. For storage, you pick the redundancy, the access tier, the capacity and the expected number of operations. For networking, you enter the gigabytes of outbound data transfer. Changing any one input shows immediately how much that factor matters, which makes the calculator a useful teaching tool as well as an estimating tool.

Estimates can be saved, named, duplicated, shared through a link and exported to a spreadsheet. Theo built one estimate called StockLine production with the application servers, the database, the backup vault, the disks and the expected outbound data. He then copied it as StockLine production with reservations so that Dana could see both versions side by side. He built a third for the nonproduction environment with business-hours schedules on every VM.

The calculator also offers settings for the whole estimate: the currency, which lets Harrow & Finch show figures in Canadian dollars, the support plan to include, and whether to show dev/test pricing. Customers with certain agreements who sign in may be able to see prices closer to their own negotiated rates; the public prices are shown otherwise. Either way, the result is an estimate, not a quote. Actual invoices depend on what is actually deployed and how it is used.

Making an estimate realistic

An estimate is only as good as its inputs. Renata had seen many cloud projects fail their budgets not because the calculator was wrong but because the estimate left things out. She gave Theo a checklist. First, include everything that runs, not only the obvious servers: disks, backup storage, the VPN gateway, public IP addresses, monitoring data, Key Vault operations and so on. Second, size from measurements, not from the size of the old hardware; a server bought five years ago with sixteen cores may use four. Third, include data transfer, especially anything that leaves Azure regularly. Fourth, separate production from nonproduction, because their schedules and purchase options differ. Fifth, add a margin for growth and for things nobody has thought of yet.

Measuring real usage is where many organizations get their biggest savings. Theo pulled CPU and memory figures for the 38 servers from the existing monitoring system for the previous three months. Eleven of them averaged less than one tenth of their CPU, and several were sized for peaks that had not happened in years. Sizing those servers to their measured load in the estimate reduced the compute total substantially, and Renata noted that Azure makes resizing later easy if a guess turns out to be too small.

Theo also documented his assumptions on the first tab of the exported spreadsheet: the region, the hours, the growth rate, which servers were reserved and which were not, and the date the prices were checked. "An estimate without assumptions is just a number," Renata said. "When the invoice differs, the assumptions tell you why."

Total cost of ownership

Total cost of ownership (TCO) is the full cost of owning and running something over its life, not just its purchase price. For an on-premises server room, TCO includes the servers, storage and network hardware; software licences and support contracts; the power and cooling the equipment uses; the space it occupies, whether rented or owned; the staff time spent installing, patching, monitoring and replacing it; and the cost of the backup site. Many of these costs are hidden in other budgets. The power bill for the head office building, for example, does not show which part is the server room.

A cloud TCO comparison puts the on-premises total next to the cost of running the same workloads in Azure over the same period, usually three or five years. It is a key part of a migration business case because it compares like with like: capital purchases spread over time against operating costs that start when migration begins. It also makes the case visible for things that are easy to overlook, such as the SAN support contract that Harrow & Finch would otherwise renew in June 2027, or the cost of replacing the backup tape library.

From the TCO calculator to the Azure Migrate business case

For many years Microsoft offered a separate TCO calculator on its website, where you entered your on-premises servers, storage and networking and it produced a comparison with Azure. Microsoft has since retired the TCO calculator, and its training material now says so. The capability has moved into Azure Migrate, the service used to discover, assess and migrate on-premises workloads. In Azure Migrate, a business case estimates total cost of ownership and potential savings for the servers and applications you plan to move. Older study material and practice questions may still mention the TCO calculator, so it is worth knowing both the name and what replaced it.

A business case starts with discovery data. The most accurate source is the Azure Migrate appliance, a lightweight virtual machine that discovers servers and collects their configuration and performance data over time. For a quick first look, you can import an inventory from a spreadsheet file instead. You then choose a migration preference, such as moving servers as they are to virtual machines or modernizing suitable workloads to platform services, and set cost assumptions: the target region, the pace of migration, expected growth, savings options such as Azure Hybrid Benefit, and a comfort factor that adds headroom to the sizes.

The result compares the estimated on-premises cost with the estimated Azure cost, broken down by category and shown year by year as a cash flow. The on-premises side includes compute and storage hardware, licences, network equipment, security tools, facilities such as rent and power, management software and labour. The Azure side includes the Azure services, security services, management services such as monitoring and backup, and the labour still needed to run them. The business case also compares the current state with keeping servers on premises but managing them through Azure Arc, a topic for lesson 9.

Theo ran a business case on October 27 using a spreadsheet of the 38 servers with their measured utilization. The illustrative result showed that Azure would cost more than on premises in the first year, because the company would pay for both environments while migrating and would still be paying off the old hardware, but less from the second year onward. Dana was not surprised. "Every capital project looks like that," she said. "What I need is the year it pays back, and what happens if we are slower than planned."

Dana's question about a slower migration is the kind of thing a short script can answer. Theo's version models the cumulative cost of each option year by year and shows when the Azure option catches up, using illustrative figures and a delay that keeps both environments running longer.

def cumulative(first_year, later_years, years=5):
    total, out = 0, []
    for y in range(1, years + 1):
        total += first_year if y == 1 else later_years
        out.append(total)
    return out

on_prem = cumulative(600_000, 210_000)
azure_plan = cumulative(330_000, 190_000)
azure_slow = [c + 60_000 for c in azure_plan]   # a slower migration: both environments run longer

for y in range(5):
    print(f"Year {y + 1}: on-prem {on_prem[y]:>9,}  Azure {azure_plan[y]:>9,}  Azure (slow) {azure_slow[y]:>9,}")
print("Five-year saving on plan:", f"{on_prem[-1] - azure_plan[-1]:,} dollars")
print("Five-year saving if slow:", f"{on_prem[-1] - azure_slow[-1]:,} dollars")

Year 1: on-prem   600,000  Azure   330,000  Azure (slow)   390,000
Year 2: on-prem   810,000  Azure   520,000  Azure (slow)   580,000
Year 3: on-prem 1,020,000  Azure   710,000  Azure (slow)   770,000
Year 4: on-prem 1,230,000  Azure   900,000  Azure (slow)   960,000
Year 5: on-prem 1,440,000  Azure 1,090,000  Azure (slow) 1,150,000
Five-year saving on plan: 350,000 dollars
Five-year saving if slow: 290,000 dollars

Using estimates responsibly

Estimates support decisions; they do not make them. A good TCO comparison is honest about what it includes and what it leaves out, and both the on-premises and cloud sides should be built with the same care. It is easy to make the cloud look cheap by forgetting labour, monitoring and data transfer, and equally easy to make it look expensive by ignoring the hardware refresh the company would otherwise face. The business case also cannot measure some benefits directly, such as faster delivery of new features, better resilience or the ability to scale for a seasonal peak, which matter a great deal to a retailer heading into spring.

Prices also change. Azure adjusts prices from time to time, new virtual machine families appear, and exchange rates move. Renata scheduled a review of the Skylight estimates every quarter. Each review compares the estimate with real spending from Cost Management, updates the assumptions and explains the differences to Dana, so that the forecast gets better as the project goes on.