AI Startups Are Gaming Revenue and VCs Look Away
AI founders are reporting contracted ARR as actual revenue, inflating public figures by 3-5x - and their investors know it.
6 min read
A pattern has taken hold across the AI startup market: a company announces a revenue milestone, the press reports it, and investors use it to justify a valuation built on the assumption that momentum continues. The milestone is often real. The number attached to it usually isn't.
Scott Stevenson, co-founder and CEO of legal AI startup Spellbook, called it out in a viral tweet in April that drew more than 200 reshares from founders, investors, and reporters. "The reason many AI startups are crushing revenue records is because they are using a dishonest metric," he wrote. "The biggest funds in the world are supporting this and misleading journalists for PR coverage." Spellbook is on track to reach $100 million in actual ARR by year-end, which makes Stevenson's frustration both ethical and competitive.
TL;DR
- AI startups widely report Contracted ARR (CARR) as if it were earned Annual Recurring Revenue, inflating public figures by 3-5x
- Four main tactics: CARR misrepresentation, counting unpaid pilots, annualized run-rate extrapolation, and publishing numbers above internal books
- VCs know the gap exists and stay quiet - inflated metrics attract press, talent, and customers for portfolio companies
- Critics inside the industry are pushing back, but the incentive structure makes the practice self-reinforcing
The Metric That's Not What It Says
The gap lives between two numbers that most people outside venture treat as the same thing. Actual ARR is what customers currently pay, annualized - collected or legally owed revenue from active, paying accounts. CARR (Contracted ARR) counts the total value of signed agreements, including future installments from customers who haven't been billed yet, pilots that may be cancelled, and multi-year contracts where back-year pricing includes increases customers haven't agreed to in practice.
Stevenson confirmed cases where CARR runs 3-5x higher than actual ARR. One investor who spoke to TechCrunch anonymously found a high-profile enterprise startup claiming $100 million in ARR when only a fraction came from currently-paying customers; the remainder was undeployed contracts.
"Everyone has a company monetizing CARR as ARR," that VC said.
Revenue recognition rules exist for a reason - but they don't always govern what ends up in press releases.
Source: unsplash.com
Four Ways to Cook the Books
The tactics vary in sophistication but share the same underlying logic: find the biggest defensible number and call it ARR.
| Tactic | Typical Inflation | Mechanism |
|---|---|---|
| CARR reported as ARR | 3-5x | Signed contract value instead of earned or collected revenue |
| Counting unpaid pilots | Variable | Year-long free trials counted at full annual value |
| Run-rate extrapolation | 2-4x | One strong month multiplied by 12, presented as a stable annual baseline |
| Public vs. internal gap | ~15-20% | Marketing team publishes figures above what internal books show |
A former employee at one AI company described their team counting a year-long free pilot as ARR, with board awareness - including a partner from a large fund - that the customer could cancel before full payment arrived. Another case involved a startup claiming $50 million ARR in public statements while internal figures sat at $42 million, with the difference written off as a rounding error.
"To everyone who's inside, it just feels fake," said Alex Cohen, CEO of Hello Patient.
The run-rate extrapolation problem is specific to usage-based AI pricing models. A startup that billed heavily in one strong month may present that figure multiplied by 12 as ARR, converting volatile income into a stable-looking baseline. General Catalyst managing partner Hemant Taneja framed the underlying pressure clearly: AI startup growth expectations now require jumping from "1 to 20 to 100" million ARR rather than the traditional "1 to 3 to 9 to 27" progression. When the expected path is that steep, the temptation to construct a number that clears the bar is considerable.
Who Knows and Why Nobody Stops It
The practice persists because the incentive structure at every level points toward allowing it.
VCs benefit because inflated public metrics attract press coverage for portfolio companies, create interest from larger investors, and help startups recruit. "Investors can't call it out," the anonymous VC told TechCrunch. "Everyone has a company doing it."
In investor meetings, both sides usually see the real numbers - the inflated version is for everyone else.
Source: unsplash.com
Jack Newton, CEO of Clio - a company valued at $5 billion - put the dynamic plainly: "We see some investors looking the other way when their own companies are inflating numbers."
For startups, the calculation is direct. At revenue multiples of 20-30x, an inflated ARR figure inflates valuations by the same factor. Michael Marks of Celesta Capital noted that "the valuations have gotten higher, and so the incentives are stronger to do it."
The Q1 2026 record VC quarter created an environment where these distortions compound. When Cursor raised at a $50 billion valuation on reported $2 billion ARR and Lovable claimed $400 million ARR in under a year of operation, those numbers set a market for what headline revenue is supposed to look like. Other startups then face pressure to match figures they know may be constructed on different accounting assumptions.
"You're overinflating already crazy high multiples. It's super bad hygiene, and it's going to come back and bite you." - Ross McNairn, CEO, Wordsmith
The Counter-Argument
Not everyone in the industry sees this as straightforward fraud.
Several VCs argue that CARR is a legitimate leading indicator for software businesses with long sales cycles. A signed $5 million contract represents a real, contractual obligation even if no cash has changed hands. GAAP revenue recognition rules delay booking revenue until it's earned, but that doesn't mean a contracted pipeline has no meaningful value to someone assessing a company's path.
The most defensible version of the argument holds that sophisticated investors understand the difference, rely on actual ARR internally, and that public figures are closer to marketing materials than financial disclosures. No securities laws govern private company press releases, so technically nothing in a headline CARR figure is false.
A16z general partner Jennifer Li drew the line differently. She warned founders this spring to stop believing viral claims about startups hitting $100 million ARR in months, noting that building durable businesses with customers who stick around is what attracts serious investors - ones who distinguish between real ARR and inflated run rates.
What the Market Is Missing
The risk isn't that sophisticated investors get fooled. They don't. The risk is structural.
When inflated ARR figures become the public benchmark, honest startups face pressure to match them or appear weak. Reporters write the headline number. Job candidates evaluate offers against it. Enterprise customers use it to decide whether a vendor is a real business. The fiction becomes a selection mechanism rewarding companies best at manufacturing credibility over those best at collecting revenue.
Ross McNairn, Wordsmith's CEO, rejected the practice directly: "I think it's short-sighted. You're overinflating already crazy high multiples. Super bad hygiene, and it's going to come back and bite you."
OpenAI's reported struggles to meet revenue targets in its IPO process showed what happens when markets priced on inflated expectations meet the concrete question of whether customers actually pay. For the broader class of AI startups currently claiming triple-digit million ARR figures on aggressive CARR accounting, that collision is a question of timing, not probability.
Sources:
