GAAP

Every large company must regularly publish how much money it took in and spent. So that not every firm does this according to its own taste, there are fixed rules for it. In the US, these rules are called GAAP, short for “Generally Accepted Accounting Principles.” They determine, for example, in which year a given piece of income may be counted. Anyone listed on a US stock exchange must prepare their reports according to these rules. You can think of GAAP as a kind of building code for financial reports: it doesn’t dictate what gets built, but how it must be documented.

Why investors insist on GAAP figures

Without shared rules, balance sheets would hardly be comparable. One company could book an entire ten-year contract as revenue in the first year, while another spreads it over ten years. Both would have done the same business, but on paper they would look completely different. GAAP prescribes one of the two variants, and that makes it possible to lay numbers side by side.

The second reason is trust. Anyone who buys shares relies on the published figures. In the US, the securities regulator SEC checks whether companies stick to the rules. Violations can trigger penalties, lawsuits, and a share price collapse. That’s why GAAP figures are considered the binding version of the truth about a company.

It’s important to distinguish this from a related set of rules: outside the US, IFRS usually applies. These are international standards that many countries have adopted, including the EU. IFRS and GAAP lead to slightly different profits for the same company. A comparison between a US corporation and a European one is therefore never entirely clean.

What the rules specifically determine

The core of GAAP is timing. Profit is not simply the money in the account. A piece of income is only booked as revenue once the service has actually been delivered. Anyone who sells an annual subscription in December may only count one-twelfth of it as revenue in December. The rest is carried as a liability and recognized gradually over the course of the following year.

It works similarly for expenses. If a company buys servers for a data center for 100 million dollars, that doesn’t show up as the cost of a single month. The amount is spread over the estimated useful life, say five years. This spreading is called depreciation. It ensures that large investments don’t completely distort a single quarter’s profit.

The role of stock-based compensation is frequently misunderstood. Many tech companies pay employees partly in their own shares instead of cash. Under GAAP, this counts as a real expense and must reduce profit, even though no cash goes out. This exact line item is the reason why many AI and software companies report losses under GAAP.

Non-GAAP in tech companies' quarterly reports

Anyone reading quarterly results from Nvidia, Microsoft, or an AI startup often sees two profit figures side by side. One is the GAAP profit, the other is called “Non-GAAP” or “adjusted.” For the adjusted figure, the company itself has stripped out items it considers atypical. Most of the time these are stock-based compensation and costs related to acquisitions.

The Non-GAAP figure is almost always the prettier one. It isn’t automatically dishonest, since one-off special effects can genuinely distort the view of the ongoing business. But each company decides for itself what it adjusts for. If the gap between the two figures is very large, it’s worth taking a closer look at the footnotes.

In news reports, phrasing like “loss under GAAP, profit on an adjusted basis” therefore turns up. What’s meant is: under the official rules there’s a minus, under the company’s own calculation there’s a plus. For the stock market’s valuation, both figures often matter, but for regulators only the GAAP figure counts.