The metric every AI founder brags about might be measuring contracts they haven't earned yet.
The Summary
- Scott Stevenson, CEO of legal AI startup Spellbook, called out AI startups for reporting CARR (contracted annual recurring revenue) as ARR (annual recurring revenue), inflating revenue numbers by counting future payments that may never materialize
- ARR should reflect actual recurring revenue; CARR includes contracted future revenue that depends on product delivery, implementation milestones, or customer satisfaction over multi-year deals
- VCs and journalists are either missing the distinction or choosing to ignore it, letting startups claim growth they haven't actually achieved
The Signal
ARR is supposed to be simple. You invoice $1 million in January, you have $12 million in ARR, assuming that monthly revenue continues. It's a projection, but it's grounded in cash that already moved. CARR takes that assumption and stretches it across multi-year contracts where the revenue hasn't been earned yet. The difference matters because in AI, especially enterprise AI, those future payments are contingent on things that routinely fall apart.
A healthcare AI deal might promise $10 million over three years, but only if the model integrates with existing systems, passes regulatory review, and actually delivers the promised accuracy improvements. An energy optimization contract might span five years but depend on hitting specific savings thresholds each quarter. Report that as ARR and you're claiming revenue you might never see.
"When companies go to press they are actually reporting CARR and calling it ARR, in order to have the biggest number possible."
Why this is happening now:
- Enterprise AI deals are getting longer and more complex, creating legitimate reasons to track contracted future revenue
- The funding environment rewards growth narratives, and ARR growth is the single most important metric for AI startup valuations
- Most journalists and some VCs don't understand the difference between the metrics, making it easy to slip CARR into ARR slots without correction
Stevenson points out that CARR and ARR often appear as separate metrics in pitch decks shown to sophisticated investors. But when those same companies talk to press or publish growth announcements, the CARR number quietly becomes "ARR" in the headline. The biggest funds know this is happening. They're not stopping it because inflated portfolio company metrics make their own fundraising easier.
The practice started innocently enough. AI companies with genuinely long sales cycles needed a way to communicate deal value beyond monthly subscription math. But once the door opened, the metric got weaponized. Now you have companies with $2 million in actual annual revenue claiming $15 million ARR based on contracts that won't pay out for years, if they pay out at all.
The Implication
If you're evaluating AI companies, ask for revenue recognition details. How much of their reported ARR is cash in the bank versus contracted future payments. What are the contingencies on those contracts. How many similar deals have they actually delivered on before.
For founders, the short-term win of an inflated metric creates a long-term problem. You set expectations you can't meet, you attract investors who don't understand your actual business model, and you train the market to distrust your numbers. The companies that will still be standing in three years are the ones reporting revenue they've actually earned.