Annual Recurring Revenue
Annual Recurring Revenue, or ARR for short, is the revenue a company expects from ongoing subscriptions over the course of a year. The metric is considered the most important yardstick for how fast software and AI companies are growing.
Many software companies don’t sell their product just once, but instead charge a fixed fee per month or year. Such ongoing contracts are called subscriptions. Annual Recurring Revenue, abbreviated ARR, is the amount a company would take in from all currently active subscriptions over the course of a year. So it’s not a figure from the past, but a projection for the next twelve months. One-time revenues explicitly do not count. Anyone who pays a company 50,000 euros once for a consulting project increases its ARR by zero.
Why investors stare at ARR
ARR reveals something a normal annual revenue figure doesn’t: how reliable the money is. A company with 100 million in ARR starts the new year with 100 million in revenue that’s already nearly certain. A company that earned the same sum through individual projects has to win every contract anew. Predictable income is worth considerably more to investors than fluctuating income.
That’s why investors often value young software companies as a multiple of ARR. A startup with 20 million in ARR might be valued at 400 million in a funding round, i.e. twenty times its ARR. Whether that multiple turns out high or low depends above all on the pace of growth. A company that triples its ARR every year gets a different multiple than one that grows by ten percent.
But this is exactly where a danger lies too. Because ARR receives so much attention, there’s a strong incentive to dress it up. Some companies count trial customers, others count cancellable monthly contracts or non-binding letters of intent. ARR is not a legally defined figure, and nobody audits it the way they audit a balance sheet.
How the number is arrived at
The calculation itself is simple. You take all ongoing subscription contracts and project them onto a full year. If a customer pays 500 euros a month, they contribute 6,000 euros to ARR. With 2,000 such customers, that adds up to 12 million in ARR. The closely related metric MRR, Monthly Recurring Revenue, does the same thing on a monthly basis and is usually just multiplied by twelve.
Things only get interesting once you look at the movement. New customers bring additional ARR, and existing customers can expand their contracts. At the same time, others cancel or downgrade their package. This loss is called churn. Net growth is the difference between the two, and it determines whether a company is really growing.
A common mistake is confusing ARR with profit. ARR only describes revenue, not costs. AI companies in particular often spend more on data centers and staff than they take in. A high ARR and a deeply red annual balance sheet are therefore not mutually exclusive.
ARR in headlines about AI startups
In business news about AI companies, ARR is almost always the figure cited. Headlines like “OpenAI reaches 10 billion dollars in ARR” or “Anthropic quadruples its ARR” follow exactly this pattern. The metric works well for headlines because it condenses a company’s pace into a single number.
You’ll also come across the term with perfectly ordinary services. Spotify, Netflix, Microsoft 365, and cloud storage all work on the subscription principle. Every paid subscription, including yours, is a tiny part of these companies' ARR.
When you read such a report, two follow-up questions are worth asking. First: how fast did the figure grow, and compared to what point in time? Second: was the ARR extrapolated from a single good month? Among AI companies that are only two years old, this is quite common and makes the number considerably less reliable.