The storyHow they got there
OpenAI was founded in December 2015 by Sam Altman, Elon Musk, Ilya Sutskever, and Greg Brockman — not as a startup chasing profits, but as a nonprofit research lab. The stated mission: advance AI in a way that benefits all of humanity. They raised $1 billion in commitments from Peter Thiel, Reid Hoffman, Amazon Web Services, and YC Research.
The nonprofit structure was deliberate. It let them recruit world-class AI researchers who turned down 2–3x market-rate offers from Google and Meta because the mission mattered. Co-founder Wojciech Zaremba famously turned down offers worth multiples of his market value just to be part of it. The research director was Ilya Sutskever, one of the world's leading deep learning experts trained under Geoffrey Hinton. These weren't B-players.
For the first few years, OpenAI ran pure research — publishing papers, releasing open-source tools like OpenAI Gym, and winning games like Dota 2 with their reinforcement learning agents. Nobody was paying for any of this. They were building credibility and trust in the research community — the distribution foundation that would matter enormously later.
In 2019, OpenAI made the structural move that made everything possible. They converted from a pure nonprofit into a "capped-profit" hybrid model — investors could earn up to 100x their money, but no more, with the nonprofit retaining control. No one had tried this legal structure before. The pivot immediately unlocked a $1 billion investment from Microsoft, which came with exclusive API rights on Azure and deep product integration rights across Microsoft's ecosystem. It was a distribution deal disguised as a capital deal.
The model releases that followed — GPT-2, GPT-3, DALL-E, Codex — each built credibility and created a developer ecosystem. When OpenAI made GPT-3 available via API in 2020, thousands of apps were built on top of it overnight. Each app became a distribution channel. GitHub Copilot, powered by Codex, embedded OpenAI into the daily workflows of millions of developers — without OpenAI spending a dollar on those users.
Then came November 30, 2022. ChatGPT launched. Free access, no waitlist, just an email and phone number to sign up. Within 5 days: 1 million users. Within 60 days: 100 million. Instagram took 2.5 years to reach 100 million users. TikTok took 9 months. ChatGPT did it in 60 days — with zero paid acquisition, no celebrity partnerships, no existing platform to piggyback. People just couldn't stop sharing the outputs.
ChatGPT Plus launched at $20/month in February 2023, after tens of millions of users were already hooked on the free tier. The free product was the funnel; the upgrade was the business. By 2025, more than 50 million people were paying monthly. ARR crossed $10 billion. Weekly active users hit 700 million. Valuation reached $730 billion — the largest private company in history.
The core lesson: OpenAI spent 7 years building research credibility before they had a single product anyone could use. When ChatGPT launched, trust was already banked across the research community, the developer ecosystem, and the press. The product went viral because the world already believed in the team. The launch was just the withdrawal.
Channel MixWhere the growth actually came from
Most case studies hand-wave channels. Here's the rough allocation — not in dollars spent, but in users acquired — across the routes that actually mattered.