OpenAI was founded in December 2015 as a non-profit research laboratory, backed by a group of Silicon Valley heavyweights including Sam Altman (then president of Y Combinator), Elon Musk, Reid Hoffman, Peter Thiel, and Jessica Livingston, with $1 billion in pledged funding. The stated mission was to ensure artificial general intelligence benefits all of humanity — and to counterbalance what the founders saw as the risk of AI being monopolized by a single corporation, most likely Google, which had acquired DeepMind the year before. Ilya Sutskever, one of the most respected deep learning researchers in the world and a former student of Geoffrey Hinton, was recruited from Google as co-founder and chief scientist. The early team also included Greg Brockman (CTO, formerly of Stripe), Wojciech Zaremba, and John Schulman, who would become central to the RLHF breakthroughs that made ChatGPT possible.
The non-profit structure didn't last. By 2018, Musk had departed the board (citing conflicts with Tesla's AI work, though personal tensions with Altman were widely reported), and OpenAI was struggling with a fundamental problem: competing for talent and compute against Google and Facebook required money that donations couldn't sustain. In 2019, OpenAI created a "capped-profit" subsidiary — a novel legal structure where investor returns were capped at 100x their investment, with excess profits flowing back to the non-profit. Microsoft promptly invested $1 billion, beginning a partnership that would grow to over $13 billion in committed capital and reshape the competitive landscape of both AI and cloud computing. The capped-profit structure has been the subject of ongoing controversy, particularly after a 2024-2025 effort to convert to a full for-profit corporation — a move that drew lawsuits from Elon Musk and scrutiny from California's attorney general.
OpenAI's technical story is the story of scaling. GPT-1 (2018) was a proof of concept for generative pre-training. GPT-2 (2019) generated genuine concern about misuse — OpenAI initially withheld the full model, drawing both praise for caution and criticism for hype. GPT-3 (2020), with 175 billion parameters, demonstrated that scale could produce emergent capabilities nobody had explicitly trained for, and the API made it accessible to developers worldwide. But the real inflection point was ChatGPT, launched on November 30, 2022. It was essentially GPT-3.5 with RLHF fine-tuning wrapped in a simple chat interface, and it reached 100 million users in two months — the fastest-growing consumer application in history at the time. GPT-4 followed in March 2023, establishing OpenAI as the undisputed frontier lab. The o1 and o3 reasoning model series, launched in late 2024 and early 2025, introduced "chain-of-thought" reasoning that could tackle complex math and science problems, while DALL-E (images) and Sora (video) expanded OpenAI's reach beyond text.
On November 17, 2023, OpenAI's board of directors fired Sam Altman as CEO, citing a loss of confidence in his leadership. What followed was five days of chaos that laid bare every tension in the organization. Ilya Sutskever and three outside board members had orchestrated the move, reportedly over concerns about the pace of commercialization and whether safety was being adequately prioritized. Microsoft, which had billions invested, was blindsided. Within days, over 700 of OpenAI's 770 employees signed a letter threatening to leave for Microsoft unless Altman was reinstated. He was — with a reconstituted board that included Bret Taylor (chair) and Larry Summers. Sutskever, who initially backed the firing then signed the employee letter, entered a period of silence before departing in May 2024 to found Safe Superintelligence Inc. (SSI). The crisis effectively settled a power struggle in favor of Altman and the commercial direction, though questions about OpenAI's governance have never fully gone away.
OpenAI operates a multi-layered business: a consumer subscription (ChatGPT Plus/Pro/Team/Enterprise), an API platform for developers, and increasingly, enterprise deals brokered alongside Microsoft through Azure. Revenue reportedly crossed $5 billion annualized by late 2025, driven heavily by ChatGPT subscriptions and enterprise API usage. The Microsoft relationship is both OpenAI's greatest asset and its most complex dependency — Azure hosts OpenAI's training and inference infrastructure, Microsoft bundles OpenAI models into its Copilot products, and the two companies share a revenue-split arrangement that has evolved through multiple renegotiations. OpenAI's competitive moat comes from brand recognition, developer ecosystem, and the sheer scale of its user base, but the gap with Anthropic, Google, and open-weights models has narrowed considerably since the GPT-4 era. The company's bet is that it can stay ahead through a combination of continued scaling, multimodal expansion, and the transition toward agentic AI systems that can take actions in the real world.