USV Partner Claims ARR is a False Growth Metric for AI Startups
USV General Partner Mike Mignano reiterated in an interview with Sourcery that "ARR is a false metric," pointing out that it is now easier than ever for users to pay for AI products through subscriptions, but payment does not equate to genuine liking or continued use. He explained that many subscriptions are annual, and users often must pay upfront to fully experience the product; companies can calculate ARR by multiplying yesterday's subscription rate by 365, but this does not reflect revenue durability or quality. The real focus should be on user retention, such as whether users return daily or weekly, or if they churn after a month. In the interview, Rebecca Kaden noted that the trial rate in the AI consumer market is extremely high while actual engagement is very low; the speed of product dissemination, development, and commercialization has significantly increased, but the speed at which users form stable habits has not kept pace, leading to revenue curves that can be generated within months, but whether they are still used after a year requires real-time validation. Mignano previously stated that "0 to $100 million ARR in X months is a vanity metric," as underlying AI technology creates a "magical market fit" that encourages users to try paying in the short term, but does not prove long-term value; many products merely resell inference capabilities, lacking network effects or proprietary data moats. USV has just completed its largest-ever fundraise of $900 million, with partners questioning the reliability of growth data while betting earlier and on a larger scale due to the potential that winners may become too expensive before retention is fully validated; capital is pricing future growth faster than users' actual habit formation. This mismatch creates new tensions in the primary market: products can easily be validated as "someone is willing to pay," but it is difficult to validate as "someone is willing to keep using it;" funding flows to applications that can quickly demonstrate revenue curves, benefiting from early narratives and financing, but are pressured by the reliance on real usage and long-term value from network effects. Ali Ansari, founder of micro1, mentioned in a related podcast that the company just announced $700 million ARR and emphasized "start doing things, get people to pay, and learn from it," without waiting for co-founders. Source: Public Information
ABAB AI Insight
Mignano, from co-founder of Anchor (later acquired by Spotify) to USV partner, has long focused on consumer product retention and real usage; his view that "ARR is a false metric" continues the criticism of vanity growth numbers and is reiterated in the new fund cycle. On the capital path, USV is positioning itself earlier with its $900 million fund in AI applications and infrastructure while emphasizing the importance of user retention over revenue extrapolation; the motivation is to avoid overpricing short-term subscriptions in an environment where models converge and migration costs are low, while capturing true winners with network effects or data moats. Similar cases include the emphasis on DAU/MAU during the social era and the reliance on ARR during the SaaS era failing under the AI subscription model; currently, we are in a phase where AI applications can easily gain paid trials, but retention and repurchase validation lag behind. Essentially, this is a shift in pricing power and information asymmetry: capital can price signals of "someone is paying" in advance, while real value depends on retention after 90 days, renewals, and the emergence of substitutes; this creates a new valuation risk structure between short-term revenue curves and long-term user stickiness. ABAB News · Cognitive Law
- Payments can happen within days, habits take a year to prove.
- ARR is a snapshot of revenue, retention is a time series of value.
- Capital prices the future in advance, while users may not have decided whether to open it next month.