How to calculate TCO in IT projects: total cost of ownership for CRM, ERP, infrastructure and AI services

In the late 1980s, Gartner analysts noticed a common pattern: companies budgeted only for hardware purchases, while the main costs arose during operation. As a result, projects went over budget and required additional investment.

To account not only for the acquisition price but also for all subsequent costs, the concept of _Total Cost of Ownership (TCO)_ was introduced - the total cost of ownership. TCO- is the calculation of all costs for an asset over its entire useful life: from the purchase decision to decommissioning. For IT and finance leaders, this is a way to see the project's true financial burden rather than relying only on the initial price.

Businesses calculate TCO to: 1. Compare solutions by total cost, not by entry price-TCO makes it possible to evaluate alternatives by accounting for all costs over several years: on-premises infrastructure or cloud, perpetual license or SaaS, buying a solution or outsourcing. Often, the option with the lowest upfront cost turns out to be the most expensive to operate.

For a portfolio of cloud services, use a separate company SaaS subscription audit: it ties cost to users, pricing plans, activity, and renewal dates. 2. Forecast the budget 3-5 years ahead-the calculation includes support, updates, scaling, additional staff, and other recurring expenses. This allows the company to see the financial burden in advance and reduce the risk of cash flow gaps.

3. Back investments with concrete numbers- if you include specialist salaries, downtime costs, maintenance, and possible fines, the real cost of the project becomes clear. Decisions are made based on calculations, not on a subjective judgment of whether it is "expensive or cheap." Let's look at an example: the company buys a server for 1 million rubles.

Additional annual costs: - electricity and cooling - 150,000-300,000 rubles; - administrator salary - from 1-1.5 million rubles per year; - maintenance and component replacement - 100,000-200,000 rubles; - backups and licenses - 100,000-300,000 rubles. Over 3-5 years of operation, the server costs 2-3 times more expensive initial cost. TCO shows that full amount.

Companies often calculate total cost of ownership superficially: they gather numbers in Excel but miss important expense items or compare different models incorrectly. The result is a calculation without a management decision. The table below shows five common mistakes and how to avoid them.

When calculating: - _Account for the time value of money._Payments today and in three years are not equal in value. For projects with a 3-5 year horizon, discount future expenses to present value so the comparison is financially sound. - Include a contingency buffer for unforeseen expenses. During a project, additional tasks almost always appear: enhancements, requirement changes, and growing load.

To keep the budget from falling apart, add 10-15% to the total estimate. - _Recalculate TCO regularly._Rates change, the business scales, and new regulatory requirements emerge. Update the calculation at least once a year or whenever conditions change significantly. - Check the source data. Even a precise formula will not produce the right result if the inputs are wrong. Verify actual specialist rates, real support costs, downtime statistics, and contract terms.

TCO calculation rules are changing. Below are the main changes to consider when planning a budget. 1. Artificial intelligence as a new variable cost item Artificial intelligence has become a permanent part of the IT budget, and the cost structure has changed. Instead of fixed licenses, companies pay for usage - tokens, API requests, and compute, while demand can fluctuate significantly.

At the same time, AI requires not a one-time implementation but regular fine-tuning and support, which makes the costs both recurring and variable. What this changes in TCO: 1. Costs become variable. Peak loads, hypothesis testing, and service scaling can sharply increase the budget. 2. A split appears between model training and model usage. On-premises infrastructure requires investment in GPUs and specialists. Cloud APIs are more convenient, but at high volumes they may be more expensive.

3. _Consumption control is needed._ Companies are starting to track AI costs by department and product. 2. Integrating TCO with ROI and FinOps Companies are actively adopting AI-FinOps practices: they track token consumption, set limits, and generate reports by department. TCO is calculated separately for model training and model usage, since the cost structure differs. Scenario modeling is also used to assess how the budget changes if load grows by 2, 5, or 10 times.

1. TCO is no longer calculated in isolation - it is tied to return on investment. The modern model looks like this:
2. We calculate the total cost of ownership.
3. We forecast the economic impact.
4. We compare costs and expected benefits.
5. We approve the project only when the economics are clear.

Companies that adopt FinOps discipline achieve: - transparent allocation of costs across departments; - reduced non-core spending; - regular budget reviews based on actual consumption. Example: if marketing uses AI services, the costs are charged to its budget. This increases accountability for results and makes TCO calculation practical rather than formal. 3. Hybrid infrastructure models The choice between "on-prem or cloud" is gradually becoming a thing of the past.

Most companies use a hybrid model: - mission-critical, latency-sensitive systems - on-premises infrastructure; - scalable and experimental workloads - in the cloud; - AI workloads - depending on economic feasibility. When calculating TCO, the business accounts for: - the ability to move workloads between environments; - integration costs for a hybrid architecture; - the cost of fault tolerance and redundancy.