Total Cost of Ownership

Anyone buying something usually looks first at the price tag. But the matter is rarely settled with the purchase. A car needs fuel, insurance, tires, and workshop visits. Over ten years, it may end up costing three times as much as at purchase. This is exactly what Total Cost of Ownership, or TCO for short, refers to: the sum of all costs from acquisition to disposal. The German term for this is Gesamtbetriebskosten.

Why the purchase price is misleading

Companies constantly face decisions between two offers. One is cheaper to acquire, the other more expensive. Anyone who looks only at the purchase price seems to choose correctly – and pays for it later. A cheap printer with expensive cartridges is the best-known example of this.

In IT, the effect is particularly strong. Software often costs little or nothing, but someone has to set it up, maintain it, and update it. This work time doesn’t appear in any quote. Studies suggest that the purchase price of IT systems often accounts for only twenty to thirty percent of the total costs. The rest is spread out over years.

That’s why procurement departments today usually demand a TCO calculation before signing. It forces uncomfortable questions to be asked early. Who trains the staff? What happens if the provider raises prices? How expensive is a later switch to a competitor?

What belongs in the calculation

A rough distinction is made between direct and indirect costs. Direct costs are visible and appear on invoices: acquisition, licensing fees, electricity, spare parts, maintenance contracts. Indirect costs are harder to grasp. These include onboarding time, productivity losses during outages, and the effort of transferring data to a new system at the end.

A common method is to define a time period – say, five years – and list all expected expenses within that window. The totals of the alternatives are then compared. Some calculations additionally account for the fact that money today is worth more than money in five years. This is referred to as a present value analysis.

A typical mistake is confusing TCO with Return on Investment. TCO only counts what something costs. Return on Investment relates these costs to the benefit, that is, to the money saved or earned. A system with high TCO can still be worthwhile if it brings in enough.

Data centers, cloud, and AI models

In news about artificial intelligence, TCO almost always comes up when data centers are discussed. A graphics processor for AI training costs tens of thousands of euros to acquire. Over its lifetime, however, electricity, cooling, buildings, and personnel add up. In large facilities, these ongoing costs significantly exceed the hardware price.

The same question arises when choosing between owning hardware and using the cloud. Cloud providers rent out computing power, you pay by the hour and avoid the purchase. This initially seems cheaper. However, anyone running the machines around the clock often does better with their own hardware. Several AI companies have therefore started building their own data centers again.

TCO is also the deciding argument when choosing a particular AI model. An openly available model has no licensing fees but requires its own servers and specialists. A model accessed via a provider’s interface costs per request, eliminating operational overhead. Which variant is cheaper depends solely on the volume of use – and that is exactly what TCO calculates.