AI Startups are plagued by Fake ARR
If you’ve been following the AI startup scene lately, you might have noticed a wave of headlines boasting jaw-dropping revenue numbers. But are these numbers the real deal? Or are we witnessing the rise of AI’s own version of “fake it till you make it”? Let’s break down what’s really going on with AI startups and their so-called Annual Recurring Revenue (ARR).
ARR, or Annual Recurring Revenue, was designed as a way to measure predictable, contract-based income—think multiyear software subscriptions. But as the AI gold rush continues, some startups are muddying the waters, swapping in “Contracted” or “Committed” ARR (CARR) and passing it off as actual ARR.
Why does this matter? Because CARR includes revenue from contracts that are signed but not yet live or even delivered. It’s a much squishier number—one that can be 70% higher than true ARR according to some VCs—because a big chunk of those “committed” dollars may never materialize. Implementation delays, failed pilots, or quick customer churn can turn those promises into vapor.
Investors and insiders admit that reporting CARR as ARR is so common that, once one company in a category does it, others feel forced to keep up. Some even count free pilot programs as ARR, knowing full well customers might bail before ever paying. The pressure to show hyper-growth is intense, and the fast pace of the AI sector means even big revenue “gaps” are sometimes brushed off as rounding errors.
It doesn’t stop at CARR. Some founders promote “annualized run-rate revenue”—a metric that simply takes one good month, week, or even day of revenue and multiplies it for a year. This is especially misleading in AI, where usage-based pricing means one month’s spike may not translate long-term. The result is ARR figures that look amazing on paper but may be impossible to sustain.
I've personally seen this time and time again as I review pitches from AI startups and then proceed to dig in a bit further with the founders, only to uncover the truth behind the numbers, and ultimately get disappointed.
For example, a recent pitch presented to me showed $2.4MM in ARR, and boasted this as 5X YoY ( 5 times growth year over year).
While that sounds impressive, it is the AI startup pitch equivelemt of click-bait.
The type of startup involved, which is in the online marketing and analytics space, primarily uses a usage-based pricing system, even though they may have some contracts with annual minimums and other terms.
The 2.4MM in "ARR" was really accounting for the last month's total revenue of $200k, which was an unusually good month for them (i.e., $200kx12 = $2.4MM).
Unfortunately, this is typical and the new norm, rather than an aberrant exception.
Not all founders want to play this game. Some insist on reporting only true ARR and being crystal clear with investors and the public. They know that overinflated numbers might help in the short term, but could create bigger problems down the road—especially if public markets (and future acquirers) come calling.
**Bottom line:** The next time you see an AI startup touting eye-popping ARR, take a closer look at what’s really being counted. Amid the hype, there’s a lot of creative accounting—but also a growing call for honesty and clear metrics in the AI gold rush.