VCs Question AI Startups' Revenue Claims as ARR Loses Meaning

Venture capitalists are growing openly skeptical of revenue figures reported by AI startups, saying the industry's favorite metric—annual recurring revenue—has become so elastic it often bears little resemblance to predictable income.

"The amount of VC-backed AI companies lying about their ARR publicly is absolutely unsettling," Greg Isenberg, CEO of Late Checkout, wrote in a viral post that voiced what many investors have been saying privately.

Shruti Gandhi, a general partner at Array Ventures, calls the phenomenon "ARR inflation." When she confronts founders about inflated numbers, their response is consistent: everyone else is doing it. "There's no accountability right now," Gandhi told Business Insider, which first reported the story.

Why it matters

The erosion of ARR as a reliable metric creates real risk for investors making multimillion-dollar bets on early-stage companies. When revenue claims blend contracted subscriptions with extrapolated token usage, hardware sales, and strong individual months, valuations become untethered from financial reality. This dynamic mirrors the late-2021 bubble mentality, according to some VCs, and could lead to painful corrections when growth fails to materialize.

The shift from software to tokens

ARR gained prominence during the software-as-a-service era, when companies like Salesforce sold predictable annual contracts. The metric provided a straightforward way to estimate future revenue and justify valuations.

AI companies operate differently. Many charge based on token consumption rather than fixed seats, creating volatile revenue streams. A customer might use a product heavily one month, then scale back or switch to a cheaper model. The startup's own costs fluctuate because it pays cloud providers or model companies each time someone uses its service.

"When you have ARR, you have a signed contract, and you know what you've signed up for for the year," said Matt Murphy, a partner at Menlo Ventures. That certainty has largely disappeared in AI.

Run rate emerges as alternative

Many AI startups now report "run rate" instead—a figure that extrapolates one month's revenue across a full year without claiming it will recur. Pocket, which makes an AI recording device, recently announced it crossed the $100 million run rate threshold. Founder Akshay Narisetti said he avoids ARR specifically because token-based pricing makes revenue too unpredictable.

But run rate introduces its own problems. A single exceptional month doesn't guarantee sustained performance. Narisetti said Pocket avoids this pitfall by never annualizing spike months.

The major AI labs use both metrics. OpenAI describes subscription sales as ARR while calling its advertising business an annualized revenue run rate. Anthropic generally uses run-rate revenue for overall sales.

Rare exception proves the rule

Linear, a project management tool that operates as traditional SaaS, recently reported crossing $100 million in ARR. Cofounder and CEO Karri Saarinen said he stands by the metric because Linear sells multiyear contracts and has operated since 2019, providing enough history to forecast accurately.

For newer AI startups, he said, "it's gotten really murky."

Alexander Niehenke, a partner at Scale Venture Partners, sees the loose revenue reporting as a sign of market froth. "It feels like we're at the tail end of 2021 again, given some of the behavior I'm seeing in the venture ecosystem," he said. Founders understand that VCs respond to ARR figures, and "the smart and brilliant entrepreneurs use it like catnip on us."

The issue gained rare public acknowledgment earlier this year when Cluely CEO Roy Lee admitted to misstating his startup's ARR to a TechCrunch reporter. Such admissions remain exceptional. Unlike public companies, startups face minimal regulatory scrutiny, and revenue claims are rarely independently verified.

Business Insider reported these details.