Total Cost of Ownership

Whoever buys something first sees the price on the tag. But the device also causes costs afterward: electricity, repairs, accessories, and eventually disposal. Total Cost of Ownership is the technical term for the sum of all these expenses over the entire period of use. In German, one says Gesamtbetriebskosten. A simple example is a car: the purchase price is often only half of what it really costs over ten years. Insurance, fuel, repairs, and depreciation make up the other half.

Why the purchase price is misleading

The purchase price is the only figure one knows immediately. That’s why it is overvalued. The follow-up costs are spread over years and barely stand out individually. In sum, however, they are often greater than the purchase itself.

With technology, this effect is especially strong. A cheap printer is a well-known example: the device costs 50 euros, a set of ink cartridges 40 euros. After two years, one has spent more on ink than on the printer. Anyone who compares only the purchase price systematically makes the worse decision here.

In companies, decisions are therefore rarely made based on the purchase price alone. The purchasing department calculates over a period, typically three to five years. All expected costs within this period are added up. Only this total is compared between the offers. The offer with the highest price can very well win.

Which items belong in the calculation

A distinction is usually made between direct and indirect costs. Direct costs are visible and appear on an invoice: purchase price, software license fees, maintenance contracts, electricity bill. Indirect costs are harder to grasp. These include work time for setup and training as well as downtime when the system doesn’t work.

A data center for artificial intelligence illustrates this well. A graphics processor, i.e. a specialized chip for AI computations, can quickly cost 30,000 euros to purchase. But it runs around the clock and requires considerable amounts of electricity. On top of that comes cooling, since the waste heat must be removed from the building. Over four years, the electricity can cost almost as much as the chip itself.

A common mistake is forgetting the end of use. Decommissioning, data deletion, and disposal also cost money. Conversely, a residual value can improve the calculation if hardware can be resold. Serious TCO calculations always state the period they refer to. Without this information, the figure is worthless.

TCO in cloud offerings and AI news

You most often encounter the term in the context of cloud computing. Cloud means renting computing power from a provider instead of buying one’s own servers. Providers like Amazon or Microsoft advertise that they lower total operating costs. Their argument: you pay for no hardware, no server room, no maintenance staff. Critics counter that rental costs become more expensive than owning a computer when utilization is consistently high.

TCO also appears in reports about AI models. A model with lower computing requirements can be cheaper to operate, even if it responds somewhat worse. For a company with millions of requests per day, this difference determines profitability. That’s why technical articles today often compare models by cost per response rather than just by quality.

It’s important not to confuse TCO with Return on Investment. TCO only counts what something costs. Return on Investment weighs these costs against the benefit. A purchase with high total cost of ownership can still be worthwhile if it generates enough return. And if a manufacturer advertises particularly low TCO figures, it’s worth taking a closer look at which items they left out.