The story in full
On 30 November 2022, a small San Francisco lab put a chatbot online and called it a research preview. Five days later it had more than a million users, and two months later analysts estimated a hundred million. The lab had started seven years earlier as a non-profit, promising to share its work with the world. Today it is one of the most valuable private companies on Earth. This is the history of OpenAI.
A lab with no product
OpenAI was announced in December 2015. It was not a company in the normal sense but a non-profit research lab, with a stated goal of building artificial intelligence that would benefit humanity as a whole, "unconstrained by a need to generate financial return". Its backers pledged $1 billion between them, among them Sam Altman, then president of Y Combinator, the engineer Greg Brockman, the researcher Ilya Sutskever and Elon Musk.
Big tech firms were hiring the best AI minds with huge pay packets. OpenAI offered a mission instead, and promised to publish its research openly. There was no product and no plan to make money.
Games, robots and a hand
The early years were mostly experiments. In 2016 OpenAI released Gym, a free toolkit for training AI by trial and error. Its bots, OpenAI Five, learned the strategy game Dota 2 by playing themselves, and in April 2019 beat the reigning world champion team. The same year it showed a robot hand that could solve a Rubik's Cube. The demos were impressive and expensive, and none of them earned anything. In 2018 Musk left OpenAI's board, and the lab was learning that the AI it wanted to build would need far more computing power than a charity could pay for.
The money problem
In March 2019 OpenAI changed shape for the first time. It created a "capped-profit" company controlled by the non-profit: investors could make a return, but only up to a limit, which for the first investors was a hundred times their money. Altman became chief executive that year, and four months later Microsoft invested $1 billion and became OpenAI's cloud computing partner. Critics said the open, non-profit lab was turning into something else; OpenAI said it was the only way to pay for the mission.
Too dangerous to release?
The breakthrough came from language. In 2017 Google researchers published a new AI design called the Transformer, and OpenAI used it in 2018 to build its first GPT model. In February 2019 GPT-2 wrote such convincing text that OpenAI held back the full version at first, worried it could be misused for spam and fake news. In 2020 came GPT-3, with 175 billion parameters. OpenAI sold access to it to other businesses. For the first time, the lab had a product.
The research preview
ChatGPT launched as a free research preview on 30 November 2022, a version of GPT-3.5 trained to hold a conversation using feedback from human testers. Altman posted that it passed a million users in five days. By January, UBS analysts estimated a hundred million monthly users; TikTok had taken about nine months to get there. In January 2023 Microsoft announced a new multi-year, multi-billion-dollar investment, and in March OpenAI launched GPT-4.
Five days in November
On Friday 17 November 2023, the non-profit board fired Altman, saying he had not been "consistently candid in his communications with the board". Brockman quit the same night. What followed took five days: Microsoft said it would hire Altman and Brockman, nearly all of OpenAI's staff signed a letter threatening to leave unless the board resigned, and the board appointed a second interim chief executive. On the night of 21 November, OpenAI announced Altman would return with a new board chaired by Bret Taylor. The crisis exposed the odd tension at OpenAI's heart: a non-profit board in charge of one of the world's most valuable start-ups.
Changing shape again
After nearly a year of talks with the attorneys general of California and Delaware, OpenAI restructured in October 2025. The non-profit became the OpenAI Foundation, and the business became OpenAI Group, a public benefit corporation. At the time the Foundation owned about 26 per cent and Microsoft about 27 per cent. Co-founder Elon Musk sued, saying OpenAI had abandoned its original mission; OpenAI denied it. In May 2026 a jury found he had brought his claims too late, and the judge dismissed them without ruling on their merits. Musk has said he will appeal.
Where OpenAI is now
OpenAI says ChatGPT has about 1.2 billion weekly users. Bloomberg reported its yearly revenue run rate at about $50 billion at the end of September 2026. In March 2026 it raised money at a valuation of $852 billion, and it was raising again at a reported $1.25 trillion. It still spends huge sums: a company presentation reported by the Financial Times put its computing and infrastructure bill at about $856 billion. Figures are as reported up to 5 November 2026 and may change.
Key dates
- December 2015: OpenAI is announced as a non-profit lab.
- March 2019: the "capped-profit" company is created.
- 2020: GPT-3 becomes OpenAI's first product.
- 30 November 2022: ChatGPT launches as a research preview.
- 17–21 November 2023: Sam Altman is fired and reinstated.
- October 2025: OpenAI restructures as a foundation and a public benefit corporation.
Three lessons
- Your structure is part of your strategy. OpenAI's non-profit beginning won it trust and talent, then made it hard to raise money, then nearly cost it its chief executive.
- Ship it, even if it's rough. ChatGPT was called a research preview; the world treated it like a finished product, and that head start still counts.
- Momentum isn't the same as money. A billion users and a trillion-dollar price tag sit alongside enormous costs.
Accurate as of 5 November 2026. Figures such as valuations and revenue move over time. This is an independent editorial history, not affiliated with or endorsed by any company mentioned. The full source list is in the YouTube description and pinned comment.


