The ARR Mirage: How AI Startups and VCs are Gaming Growth Metrics
Table of Contents
The ‘Huge Scam’ in AI Growth
Scott Stevenson, co-founder and CEO of the legal AI startup Spellbook, recently sparked a firestorm on X (formerly Twitter) by calling out what he described as a “huge scam” within the AI ecosystem. His target wasn’t a specific product or a faulty LLM, but rather the very numbers startups use to prove their viability: Annual Recurring Revenue (ARR).
Stevenson alleged that a growing number of AI companies are intentionally inflating their revenue figures in public announcements, often with the tacit support of major venture capital firms. The goal, according to Stevenson, is to secure favorable PR coverage and drive up valuations by misleading the public—and potentially other investors—about their actual market traction.
The sentiment has resonated deeply within the founder community. Jack Newton, CEO of the legal tech firm Clio, told TechCrunch that Stevenson’s post brought a necessary spotlight to the issue, echoing warnings previously issued by Y Combinator’s Garry Tan regarding the importance of using honest revenue metrics.
The CARR Substitution
To understand how these numbers are manipulated, one has to look at the distinction between ARR and what is known as Committed ARR, or CARR. Traditionally, ARR is a bedrock metric for SaaS and cloud companies, representing the predictable revenue from active customers currently under contract. Because it is a forward-looking projection rather than a historical accounting of cash-in-hand, it isn’t typically audited under Generally Accepted Accounting Principles (GAAP).
CARR, however, is a far more elastic metric. It includes not only current active contracts but also signed agreements for customers who haven’t been onboarded or gone live yet. In a healthy ecosystem, CARR is used as a growth indicator, but it is supposed to be adjusted for expected churn and “downsell”—the likelihood that a customer will reduce their spend over time.
The problem arises when startups simply label CARR as “ARR” in press releases or pitch decks. According to several investors and finance professionals who spoke on the condition of anonymity, this obfuscation is becoming pervasive. One investor noted that once a single company in a specific category begins reporting CARR as ARR, others feel compelled to do the same just to remain competitive in the eyes of the market.
The Gap Between Contracts and Cash
In some cases, the discrepancy is staggering. One venture capitalist revealed that they have seen companies where the reported CARR was 70% higher than the actual ARR, with a significant portion of those contracted funds likely never to materialize.
This risk is particularly acute in enterprise AI, where implementation cycles can be lengthy and complex. If a product fails to integrate or a trial period goes poorly, a signed contract can vanish before a single dollar of the projected revenue is collected. One former employee described a scenario where a startup counted a substantial, year-long free pilot as ARR. The company’s board, including a representative from a major fund, was reportedly aware that the revenue was being recognized before the customer had actually paid or even fully adopted the tech.
The ‘Rounding Error’ Justification
Not all misrepresentations are as extreme as counting free pilots. In other instances, the gap is smaller but still present—such as a company claiming $50 million in ARR when the internal books show $42 million. Sources suggest that some founders and investors view this as a “rounding error,” justifying the inflation by the sheer speed of AI growth. The logic is that the company will eventually “grow into” the inflated number before anyone notices the discrepancy.
Beyond the CARR confusion, there is the ongoing issue of “annualized run-rate revenue,” another variation of ARR that extrapolates a single high-performing month across an entire year. When combined with committed contracts, these tactics create a hall of mirrors that makes it nearly impossible for outside observers to discern which AI startups are actually generating cash and which are simply masterfully managing their spreadsheets.
