Zach Albertson

Behind Depreciation and Interest

3/22/2026

If in reading financial statements there is one number I have learned to be suspicious of, it is any measure of earnings that begins with the word “before.”

We will leave aside taxes and the amortization of intangibles for another day. What concerns us here is simpler and, for businesses that own real physical assets, more consequential: the habit of ignoring depreciation and interest when measuring what a business of any nontrivial capital intensity actually earns.

Let us start with depreciation. When a business buys a piece of equipment, a building, a fleet of trucks, or a data center full of servers, that asset starts aging the moment the check clears. The accountants spread the original cost across the asset's useful life, which is what depreciation is. The cash is gone, but the income statement recognizes the cost gradually. For a business that is neither growing nor shrinking, this spreading works out to a reasonable approximation of what the business must spend each year simply to stay in the same place. Call that maintenance capital expenditure. It is real money, spent on real things, for a real capability. Depreciation, imperfect as it is, is the accountant's best effort to tell you what that number is.

So when someone hands you earnings before depreciation and calls it a measure of what the business earns, what they are really handing you is a number that assumes the assets do not wear out. The trucks stay road-worthy without new engines. The servers keep humming without replacement. The factory floor stays current without a dollar of reinvestment. None of that is true of any business I have ever seen.

Now, depreciation is calculated on historical cost, and there is a good reason for that. We cannot know the replacement price until we face the charge, and we can only recognize the cost we incurred. Say we bought a locomotive engine for $3 million and the same locomotive now costs $4 million new. We cannot write up the book value and take larger charges going forward, because $4 million is not what we paid. Nor would we want the alternative. If replacement costs could be blended upward, they could be blended downward too, and a locomotive replaceable today for $2 million would be sending credits through our depreciation schedule. The historical cost convention saves us from that absurdity. The consequence, accepted and unavoidable, is that in any environment where replacement costs have risen since purchase, depreciation understates the true maintenance burden. The older and larger the asset base, the wider that gap.

Now consider interest. Maintenance capital expenditure does not fund itself. A business that cannot cover its asset replacement entirely from cash must borrow. Most businesses, most of the time, are in exactly this position. Debt for capital projects tends to be raised in large amounts at once, because that is the nature of the projects. A new plant, a server build-out, a fleet renewal: these require committed capital upfront, not a gradual draw. The result is that many businesses carry a more or less permanent debt load, and that debt carries interest.

If you own a business that must continually replace assets to keep operating, and if funding that replacement requires borrowing, then the interest on that borrowing is as much a cost of the base business as the depreciation on the assets themselves. You cannot have the productive capacity of the assets, exclude the depreciation because it is non-cash, and also exclude the interest because it is a financing charge. That would be like saying a rental property earns a fine return before we account for repairs and the mortgage.

The interest line has the same historical cost convention issue as depreciation, albeit with the notable option to refinance. Debt is priced at issuance, so what appears on the income statement today tells you what it cost to borrow then, not what it would cost to borrow now. Whether rates are higher or lower in the future is unknowable. Just as depreciation tells you what an asset cost to buy rather than what it will cost to replace, interest tells you what capital cost to raise rather than what it costs to carry the business forward. A capital allocator who excludes both and calls the remainder earnings is reading yesterday's prices and calling them today's economics.

Some will argue that cheap debt offsets the depreciation shortfall: if you borrowed at or near the inflation rate, the real cost of that debt approaches zero, and perhaps that compensates for replacement costs rising in nominal terms. This attempt is generally unpersuasive. Inflation does not move both borrowing rates and the sticker price of specific capital goods in lockstep. The price of a steel mill, a locomotive, or a semiconductor fab is set by its own supply and demand, and there is no mechanism that ties it reliably to the rate on a ten-year note. Claiming the benefit of cheap debt while dismissing the depreciation shortfall requires an undue preference of convenient ignorance over any possibility of intelligent capital allocation.

I will acknowledge one narrow class of exception. There are a small number of businesses with cash reserves so large, and cash generation so reliable, that they finance heavy capital expenditure entirely from internal resources and carry no meaningful debt tied to that spending. For those businesses, excluding interest does not distort the picture in the way I am describing, because there is no interest. But this describes perhaps two or three companies in the world at any given time. It does not describe the typical industrial business, the typical utility, or the typical large technology company whose capital requirements have grown alongside its ambitions.

Curiously, it is precisely these exceptional businesses that face a different temptation. A company already blessed with low maintenance burdens and enormous cash generation has every incentive to make things look even better by extending the useful lives assigned to its assets, which reduces the annual depreciation charge and lifts reported earnings without a single dollar of additional revenue. Google has done exactly this with its data center equipment on multiple occasions over the past several years, extending the assumed useful lives of servers and data center hardware each time. The accounting is technically permissible; management sets useful life estimates, and estimates can be revised. But in a business where the practical life of computing hardware is collapsing at a compounding rate, according to Moore's law, extending the depreciation life linearly in the opposite direction produces a gap that widens precipitously in that final year. The honest question to ask of any such revision is what purpose it serves other than making earnings look better immediately. For a business already generating cash at a rate that most competitors would find difficult to imagine, the answer is hard to find.

When evaluating a business, we want to know what it earns after it has paid to keep itself in the same condition it started the year. That means after depreciation, which approximates maintenance capital expenditure. It means after interest, which approximates the cost of the capital required to fund that maintenance when internal cash is insufficient. What is left is something closer to the real number: the earnings that belong to the owner after the business has taken care of itself.