You write code in VS Code. Maybe you've tried Cursor — that AI-powered fork everyone on your team keeps mentioning. You opened it once, poked around, went back to your old setup. No big deal.

Meanwhile, Cursor just crossed $2 billion in ARR (annual recurring revenue — total yearly subscription income). Two billion dollars. From a code editor. On March 2, 2026, Bloomberg confirmed that number doubled in just three months.

I needed to understand how a free download turns into that kind of money. So I reverse-engineered the funnel. 🔍

The free tier is not charity

Cursor's free plan — they call it "Hobby" — gives you 2,000 tab completions (inline code suggestions that finish your sentences as you type) and a handful of agent requests per month. Sounds generous. It's not. It's a precisely calibrated addiction machine.

Here's the four-week cycle I watched play out — on myself and three other devs I talked to:

Week 1: Tab completions save you maybe 20 minutes a day. Nice. You still remember how to type code manually. Your old workflow survives.

Week 2: You discover inline edits and the chat panel. Now you're describing what you want instead of writing it character by character. Your output doubles.

Week 3: You slam into the monthly cap. Mid-Tuesday. Three workdays left. The basic completions feel sluggish compared to the premium responses your fingers already memorized.

Week 4: You pay $20. Not because a pop-up nagged you. Because your own productivity drop forced your hand.

That's not a free trial. That's a freemium funnel — a pricing model where the free version deliberately creates dependency before restricting access. And Cursor's version converts at rates most SaaS (software-as-a-service — subscription software you access online) founders would frame on their wall.

The conversion math

According to TechCrunch's report from March 2, Cursor crossed 1 million daily active users. Data from their Series D announcement in November 2025 revealed 360,000 paying customers.

Run the numbers: roughly 36% of daily actives converted to paid seats. Industry-standard freemium conversion hovers at 2–5%. Cursor operates in a different universe.

At $20/month for Pro and $40/user/month for Business — with enterprise now making up 60% of revenue — the $2B ARR math checks out. More than half the Fortune 500 already have developers on Cursor.

Three mechanics that make it work

1. Muscle memory lock-in. After two weeks of AI-assisted tab completions, your fingers develop new patterns. Going back to vanilla VS Code feels like downgrading from a highway to a dirt road. The switching cost isn't financial — it's neurological. Your hands literally forget how to write boilerplate.

2. Progressive revelation. Cursor doesn't dump every feature on day one. You discover agent mode (where the AI makes multi-file changes on its own), codebase-wide context, and smart refactoring gradually. Each discovery is a dopamine hit that deepens the hook. By the time you understand what the tool can do, you can't imagine working without it.

3. The team virus. One developer ships 3x faster. The team notices. The free tier seeded one user. The Business plan at $40/seat harvests the whole team. CAC — customer acquisition cost, how much you spend to land one paying customer — approaches zero. One free user becomes 5–15 paid seats. 💰

What's NOT perfect

The credit-based pricing Cursor introduced in mid-2025 confuses people. Pick "Auto" mode and requests feel unlimited — choose a specific frontier model (the most capable, most expensive AI models like Claude or GPT-4) manually, and your credits drain in days. Power users burn through their $20 credit pool fast, then face throttled speeds or overage charges. The free tier's caps have only gotten tighter.

And there's a bigger question Fortune raised on March 21: Cursor doesn't own the AI models it runs on. Anthropic, OpenAI — any of them could build a competing editor tomorrow. The $29.3 billion valuation from their November 2025 Series D bets that the product layer matters more than the model layer. Nobody has stress-tested that bet yet.

What this means for your product

If you're building any tool where usage creates dependency, study this funnel:

  • Give enough to create a habit — not enough to satisfy permanently
  • Time-gate, don't feature-gate — users should hit limits mid-workflow, not before they start
  • Make the upgrade invisible — one click, same interface, just faster
  • Design for team spread — one free user should organically sell to their entire team

The old freemium playbook said cripple the free tier so it's barely usable. Cursor flipped it: make the free tier genuinely powerful, then make the paid tier feel like oxygen.

Two billion dollars says the second approach works. And the funnel that built it started with a free download and 2,000 tab completions. 🦝