
Run-Rate Revenue
Run-rate revenue projects a short time period out to a full year: taking one month's revenue and multiplying it by twelve gives you the annual revenue a company would achieve if it kept up the same pace. Especially among fast-growing AI companies, this projection often stands in for real annual figures.
A company brings in 50 million dollars in one month. Multiply that by twelve, and you get 600 million dollars. This figure is called run-rate revenue. It doesn’t say what the company actually earned over the past year. It says: this is how much would add up over a year if things kept going exactly as they did in that one month. Run-rate revenue is thus a projection, not a measured accounting figure.
Why young AI companies tout this number
For a corporation like Siemens, the run rate is uninteresting. There, revenue barely changes from month to month, and the real annual financial statement tells the whole story. The projection only becomes interesting where numbers change extremely fast. That’s exactly the case with AI companies.
An example: A company starts near zero in January and reaches 100 million dollars in monthly revenue by December. The real annual revenue might then be 400 million, because the early months were so weak. But the run rate in December stands at 1.2 billion. Both figures are correct, and both describe the same year. One looks backward, the other looks forward.
That’s why founders and investors reach for the run rate. It shows the current pace, not the diluted past. When OpenAI or Anthropic cite revenue figures in the news, they almost always mean a run rate. That’s sometimes only mentioned in passing — but it’s crucial for putting the number in context.
The calculation and its weak points
The formula is simple: one month’s revenue times twelve, or one quarter’s revenue times four. Some companies even take just the last week and multiply by 52. The shorter the period, the more heavily any random fluctuation weighs on the figure. A single major customer paying in the measured month can inflate the number significantly.
A second problem is seasonal effects. A game maker earns a lot in December and little in February. The December run rate would thus be far too optimistic, the February run rate too pessimistic. The projection always assumes that every month looks the same.
It’s also important to note: run-rate revenue says nothing about profit. An AI provider can have a billion-dollar run rate and still be racking up losses, because data centers and graphics cards cost more than what customers pay. Closely related is the term ARR, short for Annual Recurring Revenue. This refers only to recurring income, such as ongoing subscriptions. ARR is thus the more reliable variant, since one-off special earnings are excluded.
How to spot the figure in headlines
Phrases like “on an annualized basis,” “annualized,” or “annualized revenue” are the signal. If instead it says “revenue for fiscal year 2024,” that’s a real, audited figure. Both are often used side by side in articles without the difference being explained.
The run rate is most useful as a benchmark for comparison over time. If a company jumps from 1 to 4 billion in run rate within six months, that says something about its pace of growth. Company valuations are also frequently expressed as a multiple of the run rate, such as “30 times ARR.”
Outside the world of finance, similar projections show up constantly. The fuel gauge in a car calculates consumption per 100 kilometers from just a few minutes of driving. A sports commentator says a striker is “on pace for 30 goals this season.” The principle is identical — and so is the caveat: the number only holds as long as nothing changes.