Cursor's $100M ARR Fairy Tale and the AI Metric Scam

Everyone's celebrating Cursor hitting $100 million ARR in a year. Except there's one problem: half these AI "ARR" numbers are complete fiction.

The hottest metric in tech just became the most meaningless. While traditional SaaS companies spend years building actual annual recurring revenue, AI startups are slapping the ARR label on everything from enterprise pilots to revenue run rates extrapolated from their best month.

The Numbers Don't Add Up

Look at these supposedly impossible growth stories:

• Lovable: $0 to $17M ARR in three months, then $100M that summer
• ElevenLabs: $0 to $100M ARR in 20 months
• Midjourney: $0 to $200M ARR in under three years
• Cluely: Claimed they "doubled ARR to $7M in a week"

That last one should make you laugh. Doubling ARR in a week? That's not recurring revenue - that's a good sales month multiplied by 52.

Six VCs told Fortune that founders are routinely counting pilots, one-time deals, and unactivated contracts as recurring revenue. The dirty secret? Everyone knows.

VCs aren't stupid. They're just playing a different game.

The Elephant in the Room

AI products fundamentally don't work like SaaS subscriptions. Most are:

• Usage-based (pay per API call, not monthly seats)
• Pilot-heavy (enterprises testing before committing)
• Easily cancellable (no multi-year lock-ins)
• Cost-variable (inference costs fluctuate wildly)

Calling this "recurring revenue" is like calling a food truck a restaurant chain. Similar business model, completely different economics.

But here's the thing - it works. Inflated ARR gets you:

1. Investor meetings ("Did you see their growth rate?")

2. Talent acquisition ("We're the fastest-growing AI company")

3. Market buzz (Twitter loves a hockey stick)

4. Higher valuations (growth multiples on fake growth)

The incentives are perfectly aligned for everyone to lie.

What Developers Actually Need to Track

If you're building an AI startup, forget the vanity metrics. Focus on:

• Daily active usage (are people actually using your API?)
• Gross margin per user (after inference costs)
• Cohort retention (do customers stick around after month 1?)
• Seat expansion (are teams growing their usage?)

Your billing architecture becomes critical. Without clean metering and revenue recognition, you can't even measure real ARR properly. Most AI startups discover this the hard way during their Series A diligence.

The Reckoning Approaches

This metric inflation works until it doesn't. Just ask any startup from the 2021 vintage dealing with "growth at all costs" hangovers.

The smarter VCs are already adjusting. They're looking past headline ARR numbers and digging into:

• Contract concentration (how much comes from your top 3 customers?)
• Renewal rates (what percentage actually renews?)
• Usage trends (growing or declining post-signup?)
• Unit economics (profitable per customer?)

Meanwhile, founders keep posting screenshots of their "ARR" dashboards on Twitter, hoping nobody asks about churn rates.

The AI boom is real. The growth stories? Half fiction, half wishful thinking, and entirely unsustainable as a way to measure actual business health.

Stop optimizing for metric optics. Build something people actually want to pay for repeatedly. Revolutionary idea, I know.