Skip to content
0MESubscribe
Startups18 Nov 20262 min read

Why AI startups face a reckoning as investors look beyond standard ARR

Investors are digging past simple revenue run rates as AI startups mix subscriptions with usage fees, making headline numbers far less reliable.

Listen to this storyAbout 3 min, read by your device
Picture made with AI

When news emerged in October that OpenAI’s annualized revenueRevenueAll the money a company takes in from sales, before costs. Open in glossary missed previous projections, AI-related stocks dropped and scrutiny around AI company valuations increased. The shift has pushed venture investors to ask whether annual recurring revenueAnnual recurring revenueThe predictable revenue a company expects to repeat every year from active customer subscriptions. Open in glossary, or ARRARRThe money a subscription business expects to bring in over a year from its current customers. Open in glossary, still works as a primary metric for evaluating artificial intelligence businesses.

The problem with usage-based revenue

ARR gained widespread use in the late 2000s when software firms switched from one-off software licenses to recurring annual subscriptions. Customers signed fixed contracts, making forward revenue easy to calculate. But modern AI firms often make money differently. Rather than relying purely on predictable contracts, many operate hybrid models combining base subscriptions with fees for consumed computing power, APIAPIA way for one program to ask another program for data or a service. Open in glossary calls, processed tokens, or completed outcomes.

As a result, what many startups present as ARR is actually an annualized revenue run rateRevenue run rateAn estimate of annual sales calculated by multiplying short-term recent earnings over a full year. Open in glossary based on recent short-term sales. This calculation combines guaranteed contract income with variable customer usage that could disappear next month. A firm relying on a recent spike in usage from a few clients can show the exact same headline run rate as a business with steady, expanding contracts across a wide customer base.

What investors evaluate during due diligence

While top-line growth figures may satisfy early-stage investors looking for basic signs of demand, growth-stage investors are demanding deeper operational metrics.

  • Revenue predictability: Assessing the balance between locked-in contract value and variable, consumption-based fees.
  • Customer concentration: Checking whether a company relies dangerously on just a few accounts.
  • Margin durability: Verifying if gross margins can hold steady as computing usage expands.
  • Retention and churnChurnThe share of customers who stop paying in a given period. Open in glossary: Tracking cohort retention to see if customers keep paying over time or drop off once the initial hype fades.

Investors do not expect ARR to disappear completely. However, as more AI startups raise growth rounds, founders must break down what their headline numbers actually include.

How does this story land?

One tap per reader.

Quick poll

Is traditional ARR an accurate metric for evaluating AI startups?

One tap. No sign-up.

Ask this story

Still have a question?

Answers use only what is in this story. They are written by AI and can be wrong.

Sources

  1. Crunchbase News, What To Measure When ARR Isn’t Enough
Up next in StartupsAtomico urges Europe to back its breakout tech firms and keep ownership local2 min read