OpenAI's October revenue report, which fell below earlier estimates, prompted fresh questions about whether ARR remains the right metric for valuing AI companies. ARR was built for software firms selling predictable annual contracts, but most AI companies now operate hybrid models combining subscription fees with consumption charges tied to API calls, tokens, or compute usage.
The gap between headline ARR and actual business health has widened. A company reporting $10 million in annualized run rate could derive most revenue from contracted, steady customers whose usage expanded over 12 months. Another at the same $10 million figure might rest on a handful of customers whose consumption spiked in one quarter. The headline masks the difference.
Companies can report partner sales as either full transaction value or their own share, producing different numbers for identical businesses. Both conventions are legitimate but not comparable.
Kenza Zayani, an early-stage investor at RTP Global, notes that at pre-seed and seed stages, steep trajectory matters as a signal of market demand. Later, the story changes. Investors will increasingly ask what ARR actually counts, how much revenue is contracted versus consumption-based, and what retention looks like by cohort. Churn proves diagnostic where a growth figure does not, revealing demand the product misses, misaligned pricing, or shallow markets.
Investors expect founders to have those answers ready. Zayani writes: "The founders who have those answers ready will tell a much more convincing story than a single number ever could."