Latest / The Series A Show with Fexingo: Early-Stage Funding, Pitch Decks, and Startup Milestones / The ARR Game How AI Startups Inflate Revenue Numbers
Transcript
- Lucas: So there's a piece in TechCrunch from last Friday that has been making the rounds in my feeds — and it's about something a lot of us in the startup world have been feeling but maybe haven't wanted to say out loud. The headline says it: 'How VCs and founders use inflated ARR to crown AI startups'. Luna: Yeah, I saw that. And 'inflated' is doing some heavy lifting there. What are we talking about — like, counting revenue that hasn't really materialized yet? Lucas: Exactly. The article walks through a few specific tactics. One is the pre-purchase trick — a startup's biggest customer buys three years of software upfront, and the startup puts all of that into month-one ARR instead of spreading it out. Another is bundling non-recurring services — like consulting or one-time implementation fees — into the recurring number. And then there's the classic: extending contract terms beyond the actual usage to smooth out churn. Luna: Right, so you get a headline like 'Startup X hits ten million in ARR in just twelve months' — but really it's more like three million in genuine recurring revenue, and the rest is smoke and mirrors. Lucas: Exactly. And the article makes the point that this is particularly acute in AI right now because the hype cycle is so intense and the funding window is so competitive. Founders feel pressure to show hockey-stick growth to justify the valuations VCs are throwing around. And VCs, for their part, are often complicit — they need the narrative to sell the next round or to mark up their own portfolios. Luna: It's a weird equilibrium. Everyone knows the numbers are fuzzy, but nobody wants to be the one who calls it out because then your own deal might get scrutinized. Lucas: That's the tension exactly. And I think it's worth noting that this isn't entirely new — we saw similar dynamics during the SaaS boom of the late 2010s, and even back in the late 90s with 'eyeballs' metrics. But the AI wave has a couple of unique features. One is the sheer speed of capital deployment. Another is that many AI startups sell into other AI startups, so the whole ecosystem can end up inflating each other's numbers in a kind of circular fashion. Luna: That's a great point. If Startup A buys from Startup B and Startup B buys from Startup A, both books look stronger, but nothing real has happened. Lucas: Right. And the article highlights a specific example — a company that counted a multi-year, non-cancellable contract with a large enterprise as ARR, even though the product was still in beta and the customer had no intention of rolling it out to more than a handful of users. It was essentially a 'land and expand' strategy where the land was counted immediately, but the expand never came. Luna: So the question becomes: how do you as an investor or a founder separate the signal from the noise? Is there a reliable way to audit ARR? Lucas: The article suggests a few things. One is to look at gross ARR versus net new ARR. Gross ARR includes everything — upsells, renewals, the full contract value. Net new ARR only counts revenue from new customers that wasn't there before. If a company's gross ARR is growing but net new ARR is flat, that's a red flag — it means existing customers are being squeezed for bigger contracts, but the customer base isn't actually expanding. Luna: That's a really useful distinction. I've also seen some firms start asking for cohort-level retention data — not just the blended churn number, but how specific cohorts of customers behave over time. That can expose if the early customers are sticky but the later ones are dropping off. Lucas: Yeah, cohort analysis is powerful. And there's another metric mentioned in the piece: dollar-based net retention. If that number is over 120 percent, you should be skeptical — because it's really hard to get existing customers to expand that much without some sort of accounting gymnastics. Especially for early-stage startups that haven't even fully defined their product-market fit yet. Luna: Let's tie this to what's happening in the market right now. We're recording on May 23, 2026. The ARK Genomic Revolution ETF, ARKG, is up over 9 percent in the last five days. That's a huge move for a thematic fund. And a lot of the holdings in ARKG are early-stage biotech and ai enabled drug discovery startups — exactly the kind of companies that might feel pressure to inflate ARR to attract more capital. Lucas: That's a really interesting connection. So if you're an investor looking at a company in that space and they're claiming, say, five million in ARR with a 140 percent net retention, you should probably ask: 'Show me the cohort data. Break down the net new ARR. And tell me how many of those contracts include non-recurring services.' Luna: And I'd add: ask about the definition of 'recurring'. Some companies define it as any contract that auto-renews, even if the customer can cancel at any time. That's not really recurring revenue — that's just a subscription with an opt-out. Lucas: Right. And the article points out that the SEC has started to take notice. There have been a few comment letters sent to AI startups that went public via SPAC — asking them to clarify how they define ARR in their filings. That could be a signal that regulators are starting to tighten the screws. Luna: So what's the takeaway for a founder listening? I mean, if you're early stage and you're seeing competitors touting huge ARR numbers, the temptation to inflate your own must be enormous. Lucas: I think the honest answer is: resist it. Because the air comes out eventually. If you raise a round based on inflated ARR, you're setting yourself up for a down round later when the truth emerges. And more importantly, you're building a culture of fudging numbers from day one — that's really hard to unwind. Better to be the company that says 'We have two million in genuine ARR and we're growing 30 percent quarter over quarter' than the one that says 'We have ten million in adjusted ARR' and nobody knows what that means. Luna: And for investors, the lesson is probably: do your own work. Don't take the ARR number at face value. Ask for the backup. And if a founder gets defensive when you ask for cohort data, that's a red flag in itself. Lucas: Yeah, I've seen that happen. A founder once told me, 'You're the only investor who's ever asked me for cohort retention.' And I thought — that's a problem. If you've raised from ten other firms and none of them asked, what else aren't they looking at? Luna: It goes back to something you said earlier — the equilibrium where everyone knows the numbers are fuzzy but nobody wants to pop the bubble. But eventually, the bubble pops. And the companies that survive are the ones that built real, sustainable businesses with honest metrics. Lucas: Exactly. And I think that's the real story here. The AI boom is real — the technology is transformative. But the hype cycle around startup finance is a human phenomenon, not a technological one. And humans have a tendency to take a good thing and stretch it until it breaks. Luna: So the next time you see a headline like 'AI Startup Hits 50 Million ARR in 18 Months', take it with a grain of salt. And maybe check the fine print. Lucas: Or just ask 'What's your net new ARR?' It's a simple question, but it cuts through a lot of noise. Luna: Alright, that's our episode. I'm Luna. Lucas: And I'm Lucas. We'll be back tomorrow with another deep dive into the world of early-stage funding.