How Sam Altman turned a nonprofit research lab into the most consequential AI company in history

The origin

OpenAI was founded in San Francisco in December 2015 as a nonprofit artificial intelligence research laboratory. The founding team included Sam Altman, Elon Musk, Greg Brockman, Ilya Sutskever, Wojciech Zaremba, and John Schulman — a group that combined Silicon Valley's most successful investors and operators with some of the world's leading AI researchers.

The founding premise was explicitly about safety. The founders believed that artificial general intelligence — AI systems that could match or exceed human performance across all cognitive tasks — was coming, and that it was better to have safety-focused researchers at the frontier than to cede that ground to organizations with purely commercial incentives. The nonprofit structure was designed to ensure that OpenAI's work would benefit humanity broadly rather than shareholders specifically.

Sam Altman was not a founding member. He was chairman of the board from the beginning, but he was still running Y Combinator, the startup accelerator he had led since 2014. He became CEO of OpenAI in 2019, the same year the organization created a "capped profit" subsidiary to attract the investment capital that frontier AI research required. The nonprofit retained control; the for-profit subsidiary could raise money and pay employees with equity.

The structure was unusual, the mission was ambitious, and the early results were impressive but not yet consequential. GPT-2, released in 2019, was capable enough that OpenAI initially withheld the full model out of concern about misuse. GPT-3, released in 2020, was genuinely remarkable — a language model that could write coherent prose, answer questions, and complete code with a fluency that surprised even its creators.

But GPT-3 was an API. Developers could build with it. Ordinary people could not use it.

The challenge

The gap between what OpenAI's models could do and what ordinary people could access was the central challenge of the company's first seven years. The research was world-class. The products were not yet there.

Altman understood that the path to OpenAI's mission — ensuring that AI benefited humanity broadly — ran through products that people actually used. A research lab that published papers and provided APIs to developers was not going to shape how AI developed in society. A product that hundreds of millions of people used every day would.

The challenge was that building consumer products was not what OpenAI's culture was optimized for. The company had been built to attract the world's best AI researchers, and the world's best AI researchers were not primarily motivated by product metrics. The tension between research excellence and product velocity was real, and it played out in the company's internal debates throughout 2021 and 2022.

The decision to build ChatGPT — a simple conversational interface on top of a fine-tuned version of GPT-3.5 — was, in retrospect, obvious. At the time, it was not. The model was not the most capable version OpenAI had. The interface was not technically sophisticated. The product was, in the most literal sense, a text box.

The breakthrough

ChatGPT launched on November 30, 2022. It reached one million users in five days — faster than any product in history. It reached 100 million users in two months, a record that stood until Meta's Threads app launched in 2023.

The speed of adoption was not primarily about the technology. GPT-3.5 was not dramatically more capable than GPT-3, which had been available to developers for two years. The breakthrough was the interface. A text box that anyone could type into, that responded in natural language, that could help with writing, coding, research, and conversation — this was something that required no technical knowledge to use and delivered immediate, visible value.

The ChatGPT moment changed the technology industry's roadmap overnight. Google, which had been developing large language models for years and had produced some of the foundational research that made GPT possible, declared a "code red" and accelerated its AI product development. Microsoft, which had invested $1 billion in OpenAI in 2019, announced a $10 billion follow-on investment and began integrating OpenAI's models into every product in its portfolio. Meta, Amazon, and every major technology company announced AI initiatives within weeks.

The companies that had been cautious about releasing AI products — concerned about accuracy, bias, and misuse — found themselves in a different competitive environment. The question was no longer whether to ship AI products. It was how fast.

GPT-4, released in March 2023, was substantially more capable than GPT-3.5 and became the foundation for a wave of AI applications across every industry. The API business that had been OpenAI's primary revenue source grew rapidly as developers built on GPT-4. The consumer ChatGPT product grew to 100 million weekly active users by early 2023.

The boardroom crisis

In November 2023, OpenAI's board of directors fired Sam Altman, citing concerns about his candor with the board. The firing triggered a crisis that lasted four days and became one of the most dramatic corporate governance episodes in Silicon Valley history.

Within 24 hours of Altman's firing, nearly all of OpenAI's 770 employees signed a letter threatening to resign and join Altman at Microsoft if he was not reinstated. Microsoft CEO Satya Nadella announced that Altman would lead a new AI research team at Microsoft. The board, facing the prospect of losing essentially the entire company, reversed course. Altman was reinstated as CEO five days after being fired.

The episode revealed something important about OpenAI's structure: the nonprofit board had legal authority over the company, but the company's value resided in its people and its relationships — with employees, with Microsoft, with the developers and enterprises that had built on its APIs. The board's authority was real but brittle. When it conflicted with the interests of the people who actually built the products, the board lost.

The impact

OpenAI's valuation reached $157 billion in a 2024 funding round — making it one of the most valuable private companies in history. The company's revenue grew from essentially zero in 2022 to an estimated $3.4 billion annualized run rate by late 2023, driven by ChatGPT subscriptions and API usage.

The broader impact is harder to measure but more significant. ChatGPT's launch accelerated the deployment of AI across every industry by years. The companies that had been planning cautious, multi-year AI strategies found themselves in a world where their customers were already using AI tools and expecting AI-powered products. The timeline compressed.

For small businesses, the impact has been immediate and practical. AI writing tools, coding assistants, customer service automation, and data analysis tools — all built on the foundation that OpenAI's research and products created — are available at costs that would have been unimaginable five years ago.

The legacy

OpenAI's story is still being written. The company's structure — a nonprofit controlling a for-profit subsidiary — has created ongoing governance tensions. The safety mission that motivated the founding has been in constant tension with the commercial pressures of competing in the fastest-moving technology market in history.

But the product legacy is clear. ChatGPT proved that AI could be a consumer product, not just a developer tool. That proof changed what every technology company built, what every investor funded, and what every business leader had to think about.

ChatGPT was not the best possible AI product. It was the first one that anyone could use. That difference — between perfect and accessible — is what changed everything.

The lesson for smaller businesses is not about AI specifically. It is about the relationship between capability and accessibility. OpenAI had GPT-3 for two years before ChatGPT. The technology was there. The product that made it accessible was not. The gap between those two things — between what your product can do and what your customers can actually use — is where most business value is lost.