OpenAI: Products, Company History & AI Timeline

What OpenAI sells

OpenAI builds and sells general-purpose AI models, mainly through ChatGPT, a consumer and business chat product, and an API that developers pay to call directly. Sam Altman remains chief executive, a role he was removed from and restored to within a single week in November 2023. Greg Brockman is president and Jakub Pachocki is chief scientist. The company also ships Codex, a coding agent, and has run video generation under the Sora brand, though it shut the standalone Sora app down in March 2026 after a short run. [2][3]

OpenAI's own July 2026 disclosure put its products at more than a billion active users and more than two million businesses, a figure its CFO folded into a post about revenue and margins rather than a users-milestone announcement of its own. Independent reporting treats OpenAI and Anthropic as the two AI companies large enough to be valued as standalone businesses rather than as a division of something else, which is the frame most coverage of OpenAI's finances now uses. [4][5]

The lab that set the pace

OpenAI's relevance to AI history runs through a specific bet: that making language models bigger, on more text, kept making them better in ways that were useful rather than just measurable on a benchmark. GPT-3, described in a 2020 paper as a few-shot learner, demonstrated the bet at a scale nobody else had shipped. ChatGPT took the same idea and put a chat interface in front of it. It was framed internally as a research preview, and it is the event most of the rest of the timeline follows from: it set the template of shipping a product to the public rather than a paper to researchers, and every major lab has been reacting to that choice since. [6]

Atlas interpretation: OpenAI's second contribution is less visible but arguably as consequential: showing that spending more compute at inference time, not just training time, is a separate axis you can push on. o1 traded a slower answer for a more reliable one by having the model reason before responding, and every frontier lab now ships a model family built around the same trade. [7]

The corporate machine behind the lab

OpenAI was announced in December 2015 as a nonprofit with a billion dollars pledged and, deliberately, nothing to sell. That structure produced the November 2023 board crisis and, ultimately, could not survive the amount of capital the company needed to keep training frontier models. In October 2025 OpenAI completed a recapitalization into OpenAI Group PBC, a public benefit corporation, with the original nonprofit foundation holding roughly 26 percent and Microsoft roughly 27 percent, after review by two state attorneys general. The nonprofit foundation still appoints the PBC's board and can remove its members. [8][9]

Microsoft has been the constant in that history: an initial billion-dollar investment in 2019, a further commitment reported at ten billion dollars in 2023, and a continuing equity stake after the 2025 restructuring. OpenAI has since diversified its compute suppliers rather than relying on Microsoft's Azure alone, committing to the Stargate data center program and to a Nvidia partnership to deploy ten gigawatts of Nvidia systems, on top of an AMD chip deal structured around equity. [10][11]

OpenAI on the timeline

OpenAI's turning points on the timeline track a company that keeps producing events a general history of AI would need to mention: the 2015 founding, GPT-3, ChatGPT, GPT-4, the November 2023 board crisis, o1, and GPT-5, which shipped as a router across a model family rather than a single model and got restored to the old picker within a week after users objected. The 2025 recapitalization into a public benefit corporation is also marked as a turning point, because it resolved the governance question the board crisis raised without answering it. [9][12]

Two 2025 events show the competitive pressure OpenAI has been under and applying. In December 2025, Sam Altman declared a code red over Google's Gemini, the same phrase Google's own CEO had used three years earlier when ChatGPT caught his company flat-footed. And in September 2025, Nvidia committed up to $100 billion to OpenAI, a compute-for-equity arrangement that ties OpenAI's largest supplier's returns to OpenAI's own growth. [13][14]

In June 2026, OpenAI's rollout of GPT-5.6 was held back by the same Commerce Department directive that suspended foreign-national access to Anthropic's Fable 5 and Mythos 5, the first time access to a frontier model itself, rather than the chips that train it, became a controlled export. In July 2026, OpenAI reported crossing a billion active users across its products, a figure its CFO disclosed inside a post about revenue and margins rather than as a milestone announcement in its own right. [15][4]