“For Freemium and PLG to work at scale, you probably need millions of folks to use your product.
If it’s 1000’s … then go sell it.”
Freemium is back. In fact, it’s never been bigger. The Age of AI is build on freemium. ChatGPT is the shining example.
The AI leaders have proven the freemium math works — at truly massive scale:
- ChatGPT has 800 million weekly active users, with roughly 12-15 million paying subscribers at $20/month — and hit $20 billion ARR exiting 2025. That’s a conversion rate of about 2-5%.
- Google Gemini crossed 650 million monthly active users by Q3 2025, with a free tier and paid plans at $20-$250/month.
- Claude (Anthropic) reached 30 million monthly active users with $5 billion ARR by mid-2025, driven heavily by enterprise and API revenue.
- Perplexity hit 30 million monthly active users and $100-150 million ARR — all on a freemium model at $20/month Pro.
- Midjourney reached $500 million in revenue in 2025 with 20+ million registered users — and importantly, with zero external funding and a team of ~100 employees.
- Higgsfield got to $100M in less than a year, all freemium, for AI video
These AI giants have validated the freemium model at a scale we’ve never seen before. But they’ve also proven the math I’ve been talking about for years.
The Classic B2B and SaaS Examples Still Hold True
- Atlassian has recently added Freemium editions for the first time to several of its core products — and saw a tripling of sign-ups!
- New Relic has gone back to adding Freemium, and saw 10x the sign-ups than before.
- Zoom and Slack have built empires on top of Free and Freemium.
- Calendly hit a $3B valuation and 60m in ARR almost all on top of freemium.
- GoDaddy has crossed $4B in ARR!
- Wix is now worth $15 Billion!
But having built a partly freemium service that hundreds of millions use … I’ve learned one thing: almost everyone gets the math wrong in Freemium.
Freemium generally requires about 10x-100x more users to work as a model than most founders expect
And it takes longer to take off than almost all of us expect
The AI Leaders Prove the Freemium Math — On Steroids
Here’s what’s stunning about the AI freemium explosion: the conversion rates are exactly what I predicted years ago.
ChatGPT’s numbers tell the story perfectly:
- 800 million weekly active users
- ~12-15 million paid subscribers (ChatGPT Plus + Pro + Enterprise)
- That’s roughly a 2% conversion rate from free to paid
- Revenue: $20 billion ARR exiting 2025
Do the math: At $20/month (roughly $240/year average), 12-15 million paying customers gets you to $3-4 billion just on consumer subscriptions. Add in API revenue, enterprise deals, and ChatGPT Pro at $200/month, and you’re in the $15-20B range.
But here’s the key insight: To hit those numbers, OpenAI needed hundreds of millions of free users. Not millions. Hundreds of millions.
RevenueCat’s Data Confirms It
RevenueCat manages a stunning 34% of all mobile subscription payment apps. What do they see? The same. A 1.7% conversion from free-to-paid on average.
The AI era hasn’t changed the fundamental economics. It’s just given us much bigger examples.
Let’s Do The Math (Again)
- Assume you can get $10/mo per paid user (a good basic price for Freemium for SMBs).
- To build a $100m business (i.e., get ready to go IPO), you’d need 1,000,000 paid seats to hit that.
- Assume a 2% conversion rate, for simplicity’s sake. You’d need 50,000,000 active users over time. Not pretend users. Not users who registered and never came back. Not even users that use you once a year. 50,000,000 active, passionate, engaged users using your business app, on their own, electively, with some regularity. Enough not to churn.
The AI leaders prove this math:
The standouts here? Midjourney and Claude both have higher conversion rates because they serve more professional/power users. Midjourney doesn’t even have a free tier anymore — it’s all paid plans starting at $10/month. Claude’s revenue is 70-75% enterprise/API, not consumer freemium.
The Classic Examples Confirm It
In VSB or SMB business apps, you can get to one million paying customers — but you really, really need something broad-based to get there. Intuit, Microsoft, Adobe, PayPal. But there are not too many with 1,000,000 paid business customers, or 50m truly active users.
- Hubspot sells to SMBs and is approaching $2.5B in ARR. But it got there with ~200,000 paying customers, not millions. 60% of customers now try a free product first — freemium as funnel, not freemium as business model.
- Dropbox is at $2.5B ARR now, but it needed 700,000,000 free users to get there. The math there ties.
- Wix has 154,000,000 users generating $1.8B in ARR. So perhaps you can get to $100m ARR with 15m-20m users, with higher conversions than the above math.
- Box has grown past $1B in revenue, but it “tilted” from a mostly freemium product to an enterprise focus, with freemium today being about 9% of their revenue.
The AI Freemium Lessons for B2B Founders
#1. The 2% conversion rate is real — and it’s the ceiling for most products.
ChatGPT, with arguably the most viral product launch in history, converts at 2-5%. Perplexity, with a highly engaged power user base, converts at 2-3%. If you’re planning for 10%+ conversion on a freemium consumer product, you’re likely being too optimistic.
#2. Revenue per user matters more than you think.
Midjourney’s $500M revenue on 20M users (~$25/user/year average) vs. ChatGPT’s $20B on 800M (~$25/user/year average) — the per-user economics are remarkably similar. The question is: can you get enough users?
#3. Enterprise and API revenue can change everything.
Claude (Anthropic) has 70-75% of revenue from enterprise and API, not consumer subscriptions. That’s why they can hit $5B ARR with “only” 30M monthly users. The ACV is dramatically higher.
#4. Freemium still works best as a funnel, not a standalone business model.
Look at HubSpot: 60% of customers try free first, but they end up paying $10K-$100K+ per year on average. Freemium acquired them; enterprise sales closed them.
#5. The no-sales-team model is rarer than ever.
Even ChatGPT has 3+ million paying business users through enterprise sales. Even Midjourney, famous for no sales team, is now doing enterprise partnerships. At scale, you need both.
So Is Freemium >Alone< Hopeless to Build a $100,000,000+ Business?
Absolutely not. First, SMBs are flocking to AI + B2B more than ever. And even if Freemium alone isn’t enough, it can be a critical part of getting there. What it can do is create it, build it, and seed it. This is the critical insight, and one to be flexible around.
The AI era has proven that freemium at truly massive scale works spectacularly well:
- ChatGPT: $0 → $20B ARR in 3 years
- Claude: $0 → $5B ARR in 2 years
- Midjourney: $0 → $500M in 3 years (with zero funding!)
But these are exceptions that prove the rules:
- You need hundreds of millions of users for pure freemium to work
- Or you need enterprise sales layered on top of freemium
- Or you need higher ACVs through professional/power user targeting
Another great lesson from the AI boom: Even ChatGPT, with 800 million weekly users and the most viral product launch ever, still needs to go enterprise. OpenAI has 3+ million paying business users, and that number is growing faster than consumer subscriptions.
Freemium Alone Still Has a Ceiling, But It’s Higher For AI Apps
Pure, automated freemium — you put up a website, and the money automatically rolls in, once people hit the choke point, no salespeople — it’s wonderful. It works. But to get to an IPO, alone? It can happen — but it’s still less common than tilting upmarket to get to higher ACVs.
More likely, you’ll need to build on top of it, and add more.
So just don’t fear adding a sales team, and driving up your ACV and value over time, if you start Freemium. Many of the best have done just that — including the AI leaders.
The AI freemium giants have proven the math works at scale. But they’ve also proven you need truly massive scale to make it work.
