OpEx

A company spends money in two very different ways. It can buy something that then belongs to it for years, such as a building or a machine. Or it can pay on an ongoing basis for things it consumes month after month: electricity, salaries, rent, insurance. This second kind is called OpEx, short for Operating Expenditure, or simply operating expenses. The decisive difference is not the size of the amount but the time course: OpEx keeps recurring as long as operations continue. If a company stops an activity, the associated OpEx disappears almost immediately.

Why investors watch ongoing costs

OpEx is deducted directly from revenue and thus immediately reduces a fiscal year’s profit. Large acquisitions, on the other hand, are depreciated on the balance sheet over many years. A company that buys its own server farm therefore burdens its current profit less heavily than one that rents the same computing power. That is why the split between ongoing and one-time costs says a great deal about how a financial report ultimately looks.

For investors, a high share of OpEx is also a sign of flexibility. Whoever owns nothing can quickly stop paying if demand collapses. Whoever has built a data center for billions, on the other hand, is stuck with it whether it is fully utilized or not. This flexibility comes at a price, however: rented capacity is, calculated over many years, almost always more expensive than purchased capacity.

A common misconception is that OpEx are the “small” expenses. That is not true. At an AI provider, the monthly bills for cloud computing power – that is, for servers owned by another company – can reach hundreds of millions. What is small about OpEx is only the individual line item, not the total sum.

What falls into the OpEx bucket

Operating expenses include everything consumed within a fiscal year. Typical examples are wages and salaries, office rent, marketing budgets, travel expenses, software licenses, and energy bills. Accountants check a simple question for this: Does the expense create value that will still exist next year? If the answer is no, it is OpEx.

The counter-category is called CapEx, short for Capital Expenditure, meaning capital investment spending. This refers to purchases that remain with the company long-term: land, factories, vehicles, purchased graphics cards. These amounts do not appear as costs all at once. Instead, they are spread over the estimated useful life, for servers often over five to six years.

The boundary is partly a matter of interpretation, and that is exactly what makes it interesting. If a company extends the assumed useful life of its servers from four to six years, annual depreciation falls and reported profit rises. Nothing changes in the actual cash flow. Such decisions are found in the fine print of quarterly reports and are read closely by analysts.

OpEx in cloud contracts and AI quarterly results

The term is most clearly encountered with cloud providers. Their entire business model consists of turning other people’s CapEx into their own OpEx. A start-up doesn’t need to buy servers but pays per hour used. This lowers the barrier to entry enormously, which is why many AI companies can start without owning their own data center.

In business news, OpEx regularly comes up around quarterly earnings. Sentences like “the company is cutting its OpEx” usually mean layoffs, reduced marketing, or terminated leases. Rising OpEx alongside flat revenue, on the other hand, is considered a warning sign of shrinking margins.

At AI companies, this is currently being watched especially closely, because both types of costs are ballooning at the same time. Building one’s own data centers is CapEx; the electricity to run them and the ongoing responses generated by the models are OpEx. Anyone who wants to understand whether an AI provider is making money must therefore read both figures side by side.