Skip to content
The Next Marketers
November 18, 2026
Bitcoin-3.20%
Ethereum-2.85%
Solana-2.17%
The Next Marketers
Artificial intelligence

Metrics to Track Beyond Annual Recurring Revenue

Investors are increasingly questioning whether Annual Recurring Revenue remains an adequate metric for evaluating AI companies, as many modern AI businesses rely on hybrid subscription and consumption models.

Metrics to Track Beyond Annual Recurring Revenue

Back in October, news that OpenAI’s annualized revenue was trailing earlier projections caused a sharp drop in AI-related stocks and triggered renewed evaluation of AI company valuations. This development also revived an ongoing debate among investors regarding whether Annual Recurring Revenue (ARR) remains an appropriate metric for assessing AI businesses.

These limitations have been recognized for quite some time. Originally designed for software firms that relied on predictable 12-month subscriptions, ARR only captures part of how modern AI enterprises generate income. During the initial surge of generative AI, rapid expansion overshadowed these concerns. However, the recent OpenAI situation brought the issue back to the forefront, demonstrating how heavily public markets now weigh the figures of a single private firm.

Startup founders should take note of this shift immediately. As a new wave of AI companies reaches the growth-capital phase, the inquiries they encounter during due diligence are evolving.

What ARR was built for

ARR emerged as an industry standard during the mid-to-late 2000s, coinciding with the transition of software companies from upfront license fees to ongoing recurring subscriptions. With customers locked into annual agreements, revenue became predictable, allowing firms to calculate the annual figure through simple arithmetic rather than complex forecasting.

The initial wave of post-ChatGPT companies adopted this metric without much scrutiny, even though their monetization strategies differed significantly. Today, most operate on hybrid models combining baseline subscriptions with usage-based charges tied to compute usage, token processing, or API calls. An increasing number of providers charge directly based on specific outcomes. Consequently, an annualized metric often blends guaranteed customer commitments with variable revenue that fluctuates depending on the following month’s consumption.

Figures presented as ARR are frequently revenue run rates, calculated by multiplying recent earnings to project an annual total. While this serves as a decent indicator of current momentum, it offers less certainty regarding the year ahead. Consequently, investors typically begin by asking what the reported figure actually includes.

Firms that utilize channel partners might report either the gross value of those transactions or just their retained share. While both accounting methods are valid, they yield vastly different numbers for identical business models and cannot always be compared directly.

Consider two distinct companies that both report a $10 million annualized run rate. For the first enterprise, the majority of that revenue stems from guaranteed contracts and clients whose usage has grown steadily over the course of a year. For the second enterprise, the figure relies heavily on a small group of buyers whose consumption surged abruptly during the previous quarter.

These operations are fundamentally different. The first has proven that users consistently return and increase their spending over time, yet that critical distinction is entirely hidden within a standard headline figure.

Looking beyond the headline

When a single metric fails to differentiate between such varied operational models, alternative indicators become necessary. The appropriate metrics depend heavily on company maturity. At the pre-seed and seed phases, the metric still holds value; rapid growth can signal genuine market demand, paying customers, and an early ability to execute a go-to-market strategy, as investors do not typically expect exact annual forecasts at such an early stage.

As startups mature, greater depth is required. Investors examine the underlying revenue components, including client concentration, the ratio of contracted revenue versus variable consumption, revenue predictability, and whether gross margins remain stable as transaction volume scales.

Customer retention and expansion serve as the core focus. These indicators reveal how indispensable a product has become and whether it will maintain utility over the coming years. Unlike basic growth metrics, churn rates offer valuable diagnostic insights. High churn can point to unmet market demands, misaligned pricing models, temporary needs rather than truly recurring problems, or a total market size that is smaller than initially anticipated.

What investors will expect next

Because ARR has long served as a primary gauge of momentum for high-growth technology companies, it is unlikely to disappear any time soon.

Instead, the change lies in the level of scrutiny applied to that single figure. Moving forward, investors are increasingly likely to question what a reported metric actually measures, how much of it represents committed revenue, and what cohort-based retention looks like over time. Founders who come prepared with those comprehensive answers will present a far more compelling narrative than any single summary statistic could provide.


Kenza Zayani is an early-stage investor at RTP Global, based in London. She is drawn to ambitious founders tackling the most challenging problems, with a particular interest in AI from infrastructure to applications. Zayani has built experience in venture capital across stages, and at RTP Global, she brings her multistage experience and data-driven approach to company building with founders.

Illustration: Dom Guzman

Frequently Asked Questions

Why is ARR facing scrutiny for AI companies?

ARR was originally designed for traditional software firms with predictable annual subscriptions. Most AI companies use hybrid models featuring consumption charges—such as API calls or token usage—making a standard annual run rate less reliable for forecasting future revenue.

What is the difference between contracted revenue and consumption revenue?

Contracted revenue involves guaranteed commitments from customers over a set period, whereas consumption revenue fluctuates based on how much of the product a client actually uses from month to month.

Why are retention and churn important metrics for AI startups?

Retention and churn reveal how essential a product is to its users. While growth numbers show initial traction, churn rates help diagnose pricing mismatches, unmet customer needs, or inaccurate assessments of market size.

Do investors still look at ARR for early-stage AI startups?

Yes, at the pre-seed and seed stages, a steep trajectory or run rate can demonstrate early market demand and a team’s ability to execute, even if it cannot accurately predict the entire year ahead.

What should AI founders prepare for during investor diligence?

Founders should be ready to explain what their reported revenue metrics actually count, the exact proportion of committed versus variable revenue, and cohort-based customer retention data.


Related stories

Comments 0 responses

Join the discussion

Comments are moderated and appear after review.