The metrics that made SaaS investable are being weaponized to make AI companies look like SaaS companies, and everyone in the room knows it's theater.
The Summary
- AI startups are inflating ARR numbers by stretching traditional SaaS revenue definitions, and their VCs are complicit.
- The practice distorts market signals, making it harder to tell which AI companies have real traction versus inflated metrics.
- This isn't ignorance—it's strategy. Founders and investors both benefit from the narrative inflation before reality catches up.
The Signal
AI startups are playing fast and loose with ARR, the annual recurring revenue metric that became gospel in SaaS. The problem is that AI companies don't operate like SaaS companies. They have usage-based pricing, API calls that spike and crater, pilot contracts that evaporate, and enterprise deals structured more like services than subscriptions. But ARR sounds clean. It sounds predictable. It sounds like the kind of thing that justifies a $500 million valuation.
So founders report it anyway. And VCs, who know better, repeat it in their portfolio updates and fundraising decks. The game is quiet but widespread. Take a three-month pilot contract, annualize it, call it ARR. Count API credits purchased upfront as recurring revenue even if the customer never comes back. Blend one-time professional services fees into the recurring number. The result is a metric that technically uses the right acronym but means something completely different.
"The practice distorts market signals, making it harder to tell which AI companies have real traction versus inflated metrics."
This isn't just sloppy accounting. It's a coordinated narrative strategy. VCs need their AI bets to look like the next Snowflake or Datadog to raise their next fund. Founders need those inflated numbers to recruit, to close customers who want to bet on a winner, and to set up their Series B at a markup. The incentive structure makes honesty expensive and exaggeration profitable, at least in the short term.
The downstream consequences hit everyone else. Later-stage investors inherit cap tables built on sand. Employees join with equity packages priced on fantasy. Competitors feel pressure to inflate their own numbers just to stay in the conversation. And the market as a whole loses the ability to price risk accurately when no one can trust the scoreboard.
The Implication
If you're evaluating an AI company, ignore the ARR number in the deck. Ask for net dollar retention, gross margins, and customer cohort analysis. Ask what percentage of revenue comes from contracts longer than 12 months. Ask how much of last quarter's ARR is still active this quarter. The companies with real momentum won't flinch. The ones built on narrative will pivot to talking about "market opportunity" and "strategic partnerships."
For founders, this is a short-term game with long-term costs. Inflated metrics buy you 18 months of hype and a hard reckoning when the numbers stop growing. Build for actual recurring revenue, not the appearance of it. The market will figure it out eventually. It always does.