#2 Revenue that Grows Itself: How AI-Native Startups Turn Every Customer into a Bigger One
In our last note, we talked about how the next wave of software isn’t coming from Big SaaS - it’s being built by teams solving very specific problems with tiny, flexible tools.
This time, we’re taking on the next myth:
“Sure, these AI startups grow fast… but the revenue won’t stick.”
We hear this often - and we get it. Fast revenue sometimes does mean flimsy revenue. But not here.
The numbers (and retention curves) tell a different story.
Each row is a signup cohort. Each column is a month after that signup, and each cell shows how much more those same users are spending over time.
By month five? Most cohorts are spending more than 2x what they started with. The standout December cohort? Over 4.5× expansion by month one.
Compare that to public SaaS, where the average Net Dollar Retention (NDR) is closer to 108% - this is a different league.
And if you’re wondering just how fast this can happen…
Cursor’s growth from $1M to $100M ARR outpaces almost every name in this category - faster than Deel, Wiz, and even OpenAI.
🟢 Yes, this revenue sticks around - and grows
Take a look at the cohort heatmap below. It’s from Fyxer, one of the fastest-scaling AI-native apps out of London.
Each row is a signup cohort. Each column is a month after that signup, and each cell shows how much more those same users are spending over time.
By month five? Most cohorts are spending more than 2x what they started with. The standout December cohort? Over 4.5× expansion by month one.
Compare that to public SaaS, where the average Net Dollar Retention (NDR) is closer to 108% - this is a different league.
🚀 A few examples we’re tracking closely
| 🏢 Company | 💸 ARR & Pace | 📈 What unlocks the growth? |
| Fyxer | $9M ARR in 8 months, still growing >30% MoM | Lives inside exec inboxes. Every new seat and every extra task nudges the bill up. Result? ~160% NDR. |
| ElevenLabs | $90M ARR in under 2 years. Latest round: $3.2B valuation | Per-character pricing + a viral dev API. Voice becomes a utility embedded everywhere. |
| Cursor / Anysphere | $100M ARR in ~12 months. Currently sprinting past $200M | Devs use it daily. Prompt libraries and community extensions deepen the wedge. |
Cursor’s growth from $1M to $100M ARR outpaces almost every name in this category - faster than Deel, Wiz, and even OpenAI.
🔄 Why the dollars compounds
It’s not just timing - it’s structure. These businesses are built for flywheel growth:- Usage-based pricing: Every prompt, seat, or API call nudges revenue upward
- Personalisation moats: The product learns from your data and habits
- Frictionless land-and-expand: First use happens fast, full adoption follows naturally
- Community-driven scale: Templates, voice packs, shared workflows help sell the product for you
🤔 The tension we’re watching
| Potential snag | What could break the flywheel? |
| Security & compliance lag | One breach could spook buyers. SOC-2 in-a-box may become essential. |
| Maintenance invisibility | Prompt-generated tools are fast - but who patches them when workflows break? |
| Model pricing volatility | If inference costs spike, usage-based pricing might wobble. Edge-native startups may win here. |
🧭 Where Kinetic leans in
We're most excited by founders building for:- High-NDR expansion from day one. A great customer isn’t just retained - they grow the account every month.
- AI-native UX that sells itself. Fast onboarding, instant feedback loops, and usage-based value that compounds.
- Tiny teams with outlier velocity. We’re seeing <$1M of burn unlock eight-figure revenue when the wedge is right.