Depreciation Cycle

When a company buys something expensive that will be used for many years, the purchase price doesn’t show up entirely in the annual results. Instead, it’s spread across the years the item is expected to be in use. This period is called the depreciation cycle. A delivery van costing 60,000 euros with six years of expected use therefore burdens profits with 10,000 euros per year, not 60,000 in the first year. The money is still gone immediately — only the accounting pretends the value is disappearing slowly. How long this period is set is decided by the company itself, within certain rules.

Why one year more or less shifts billions for AI chips

Large tech corporations are currently buying data centers full of graphics cards for sums in the hundreds of billions. These chips are the actual cost driver in operating AI systems. At such sums, the assumed useful life is no longer a mere formality. It directly determines how profitable a corporation appears in its quarterly figures.

A calculation example makes this clear. A company buys hardware worth 30 billion euros. With a three-year depreciation period, this burdens profit with 10 billion per year. With six years, it’s only 5 billion. The same purchase, the same machines — but the reported profit differs by 5 billion per year.

This is exactly why investors watch this figure very closely. Several major providers have extended their assumed useful lives for server hardware in recent years. Critics suspect cosmetic motives behind this: profits would then look better than the business actually performs. The companies argue that older chips continue to be used sensibly for simpler tasks.

From acquisition to a figure on the balance sheet

The process is essentially simple. The company determines the purchase price, estimates the useful life, and estimates the residual value at the end. The difference is spread across the years. In the simplest method, straight-line depreciation, the amount is the same every year.

The estimate must be justified, but it remains an estimate. No one knows for certain how long a graphics card will hold up under continuous operation or when it becomes technically obsolete. Both factors can shorten the cycle: wear and tear, and obsolescence due to newer models. For AI chips, the second factor dominates, since significantly faster generations appear every 18 to 24 months.

It’s important to distinguish this from the actual cash flow. Depreciation costs not a single additional cent in the year it is booked. Payment was made at the time of purchase. That’s why analysts often look at both figures separately: profit after depreciation and the pure cash flow. If the two figures diverge widely, it’s worth taking a closer look at the assumed cycles.

Where the figure appears in quarterly reports

In annual reports, this line item usually appears as “Depreciation” in the income statement. The assumed useful lives can be found in the notes, often in a brief table. There it might say, for example: server hardware, four to six years. Anyone reading news about cloud and AI providers regularly comes across reports about changes to these figures.

Even outside the tech industry, this principle is everyday practice. Phones, company cars, and machinery are treated the same way. Even the self-employed calculate this way: a laptop costing 1,500 euros is typically depreciated over three years. A common misconception is that a fully depreciated device is worthless or must be replaced. In accounting terms it stands at nearly zero, but in practice it can keep running.