Freemium as a route to AI category domination
For freemium and product-led growth (PLG) to work at scale, one thing has always been true:
you need millions of users.
If you only have thousands, the arithmetic doesn’t work. At that point, you should stop pretending and go sell it the old-fashioned way. What’s fascinating is that the AI era hasn’t changed this rule – it’s validated it. Spectacularly.
For a while, freemium looked like it had fallen out of favour. Not because the model was broken, but because fewer products had the kind of built-in network effects or universal utility required to support it.
Then AI arrived – and suddenly freemium wasn’t just viable again. It became the default route to growth and category domination.
ChatGPT: The ultimate freemium proof point
If there were ever any doubt that freemium works, ChatGPT erased it.
Despite years of claims that the model was “broken,” ChatGPT now sits at roughly 800 million weekly active users, converts around 2 – 5%, and entered 2026 at approximately $20 billion in ARR – roughly double what it was reporting mid-year.
The important detail isn’t the conversion rate. There’s nothing magical about it. In fact, it’s exactly what anyone who’s studied freemium economics would expect.
You don’t win freemium with better conversion rates. That’s brutally hard and usually unrealistic. You win with orders of magnitude more users – and the category presence that comes with them.
Scale first, revenue second
Do the maths. At $20 a month, 12–15 million paying users gets you a few billion dollars in consumer ARR. That’s already a great business. Everything else – API revenue, enterprise deals, $200/month Pro tiers – is the jam on top.
But none of it happens without a massive free funnel. Hundreds of millions of users trying the product, building habits, and defaulting to your tool. Even if most never convert, inertia keeps them loyal. They’re not in anyone else’s funnel – and that’s how the category crown stays firmly on your head.
This is why AI companies are dominating categories at record speed – and raising capital to match.
The evidence Is everywhere
Look around:
- Gemini crossed 650 million MAUs across free and paid tiers
- Perplexity reached 30 million MAUs and over $100M ARR with a freemium model
- Midjourney quietly built a $500M revenue business – around 20m users and 100 employees but no VC money, despite investors lining up
Freemium at AI scale doesn’t just grow revenue. It creates a market for you to monitise.
Once you become the free default for tens or hundreds of millions of users, competitors aren’t competing with your product anymore – they’re competing with your distribution. Just like Microsoft on the desktop, long before the cloud era.
At that point, freemium stops being a growth tactic. It becomes a moat.
The economics never changed – the sample size did
Here’s the part that’s easy to miss: AI didn’t change freemium economics at all.
RevenueCat data still shows average free-to-paid conversion hovering around 1.7% across subscription apps. That’s the same old number. The only thing AI changed was the scale.
Freemium has always required mega-scale to work. AI and its backers just made mega-scale achievable faster than ever.
And this explains why so many founders still get freemium wrong. They model for 10–15% conversion, miss by an order of magnitude, and conclude the strategy failed. In reality, freemium usually needs 10x–100x more users than people expect, takes longer to compound, and almost never converts more than a few percent.
But when your marginal costs are low, that’s fine. The maths still work.
SaaS history still applies
None of this is new. The AI era is simply replaying the same pattern at unprecedented speed.
- Dropbox needed ~700M free users to reach ~$2.5B ARR
- Wix has ~154M users for ~$1.8B ARR
- Calendly hit a $3B valuation largely on freemium
- Atlassian tripled signups after introducing freemium
- New Relic went freemium (again) and saw 10x signup growth
- Zoom and Slack built entire empires on free-first adoption
Even companies like Box – now over $1B in revenue – still trace their dominance back to early freemium scale, even if free users are now a smaller percentage of revenue.
The pattern is consistent: Use freemium to win the category. Monetise aggressively once you’ve won.
The immutable lesson for B2B founders
This leads to the core takeaway for AI and B2B founders alike:
Freemium dominates the category first.
Enterprise revenue consolidates and maximises it later.
Claude is a perfect example of what can be achieved if you get the numbers right. Claude hit 30 million MAUs and roughly $5B ARR. Despite having “only” tens of millions of users, 70–75% of its revenue already comes from enterprise and API. That’s shows how you can build billion-dollar businesses quickly without needing hundreds of millions of paying customers.
And, of course, if you get the proposition right high ACVs change everything. Investors love those numbers.
It also means the old “no sales team” PLG myth is mostly dead. ChatGPT already has millions of paid business users via enterprise plans. Even Midjourney is exploring enterprise partnerships.
At scale, PLG plus sales beats PLG alone.
What has been learned
The facts are remarkably simple:
Freemium works. ARPU matters. Scale matters more.
In AI, freemium isn’t just a growth strategy – it’s how you capture and lock a category before anyone else even has a chance.
AI didn’t invent freemium. It just proved – at historic scale – that the original premise was always right.
The calculation never changed. We were just using the wrong numbers.