OpenAI
| OpenAI Group PBC | |
|---|---|
 i
| |
| Type | Public benefit corporation (controlled by the OpenAI Foundation nonprofit) |
| Industry | Artificial intelligence |
| Founded | December 11, 2015 (San Francisco, California) |
| Founders | Sam Altman · Elon Musk · Ilya Sutskever · Greg Brockman · Wojciech Zaremba · John Schulman · and others |
| Headquarters | Mission District, San Francisco |
| CEO | Sam Altman |
| President | Greg Brockman |
| Products | ChatGPT · GPT · DALL-E · Sora · Whisper · Codex |
| Valuation | About $852 billion (March 2026) |
| Website | openai.com |
OpenAI is an American artificial intelligence company headquartered in San Francisco, California. Its official legal name is OpenAI Group PBC, and it operates as a public benefit corporation controlled by a nonprofit parent called the OpenAI Foundation.
OpenAI is most famous for creating ChatGPT, a conversational chatbot that became one of the fastest-growing applications in history after its release in November 2022. The company also developed the GPT family of large language models, the DALL-E image generator, and the Sora video generator. Its work helped kick off the worldwide boom in generative AI that began in the early 2020s.
OpenAI was founded in 2015 as a nonprofit research lab with the mission of building artificial general intelligence (AGI) that benefits all of humanity. Over the next decade it grew rapidly, changed its corporate structure twice, attracted hundreds of billions of dollars in investment, and became one of the most valuable private companies in the world.
1 History✎
OpenAI's history can be broken into roughly four phases: its founding as a nonprofit, its early research years, the launch of ChatGPT and the AI boom that followed, and its restructuring into a for-profit benefit corporation.
1.1 Founding (2015)✎
OpenAI was launched in December 2015 in San Francisco by a group of technology leaders and AI researchers. The most well-known co-founders were Sam Altman, who had previously led the startup accelerator Y Combinator; Elon Musk, CEO of Tesla and SpaceX; Ilya Sutskever, a top AI researcher recruited from Google; and Greg Brockman, former chief technology officer of Stripe. Other co-founders included Wojciech Zaremba and John Schulman.
The company was set up as a nonprofit so that researchers could focus on safe and beneficial AI without pressure from shareholders. Backers including Musk, Peter Thiel, Amazon Web Services, and Infosys pledged a total of about $1 billion in funding. Elon Musk later left the board in 2018.
1.2 Early research (2016–2019)✎
In its first few years, OpenAI focused on basic research in reinforcement learning and deep learning. It released open-source tools such as OpenAI Gym, a software toolkit used to train AI agents, and built systems that learned to play complex video games like Dota 2 at a professional level.
In 2018, OpenAI introduced the first version of its Generative Pre-trained Transformer, known as GPT. GPT was a neural network trained on large amounts of text from the internet so it could predict what words come next in a sentence. Later versions of GPT became the foundation for nearly all of OpenAI's later products.
By 2019, OpenAI's leaders had decided that building AGI would cost far more money than a nonprofit could raise. The company restructured into a hybrid "capped-profit" model: a for-profit subsidiary, controlled by the original nonprofit, with investor returns limited to 100 times the original investment. Around the same time, Microsoft announced a major partnership and committed an initial $1 billion investment.
1.3 ChatGPT and the AI boom (2020–2024)✎
OpenAI released GPT-3 in 2020, a much larger and more capable language model that could write essays, answer questions, and even produce simple computer code. In January 2021, the company unveiled DALL-E, a system that could create original images from text descriptions.
The biggest turning point came on November 30, 2022, when OpenAI released ChatGPT, a free chatbot built on a fine-tuned version of GPT-3.5. ChatGPT reached one million users in five days and 100 million users within two months, making it one of the fastest-growing consumer products in history. Its success is widely credited with starting the modern AI boom.
OpenAI followed up with GPT-4 in March 2023, which could handle both text and images, and with GPT-4o ("omni") in 2024, which added real-time audio. Subscription products such as ChatGPT Plus and ChatGPT Enterprise generated rapidly growing revenue. Microsoft expanded its partnership with a multi-billion-dollar investment in early 2023 and integrated OpenAI's technology into products such as Bing, Edge, Microsoft 365, and GitHub Copilot.
In November 2023, OpenAI's board briefly removed Sam Altman as CEO, citing concerns about communication and trust. After roughly 500 employees threatened to resign, Altman was reinstated within five days alongside a new board of directors.
1.4 Restructuring and rapid expansion (2025–2026)✎
In October 2025, OpenAI completed a major corporate restructuring. The for-profit subsidiary was converted into a public benefit corporation called OpenAI Group PBC, with the OpenAI Foundation nonprofit retaining roughly a 26% controlling stake. Critics noted that the word "safely" was removed from the official mission statement during this period.
OpenAI's commercial expansion accelerated. In 2025 the company released GPT-5 and agreed to acquire IO Products, an AI hardware startup founded by former Apple design chief Jony Ive, for about $6.5 billion. It also acquired the testing startup Statsig, the macOS interface company Software Applications Incorporated (maker of Sky), and Python tooling company Astral. In April 2026, the company released GPT-5.5, a more advanced version of its flagship model.
In March 2026, OpenAI closed the largest single funding round in private company history, raising about $122 billion at an $852 billion post-money valuation. Major investors included SoftBank, Amazon, and Nvidia.
2 Products✎
OpenAI's products are built on top of its large language models and other AI systems. Most are offered both as consumer apps and as developer tools accessible through an API.
2.1 GPT and ChatGPT✎
The Generative Pre-trained Transformer (GPT) is the family of large language models at the heart of OpenAI. Each new version is trained on more data and more computing power than the last:
- GPT-1 (2018) — the first model in the series, a research prototype.
- GPT-2 (2019) — much larger; initially held back over concerns about misuse.
- GPT-3 (2020) — a major leap in fluency; powered the first version of ChatGPT.
- GPT-4 (2023) — handled text and images together.
- GPT-4o (2024) — added real-time speech and faster responses.
- GPT-5 (2025) — flagship model with stronger reasoning.
- GPT-5.5 (2026) — the current flagship as of mid-2026.
ChatGPT is OpenAI's chatbot interface for these models. Users can ask it questions, get writing help, work through math problems, brainstorm ideas, or have a conversation. As of late 2025, ChatGPT had about 800 million weekly users — roughly one in ten adults worldwide. It is offered as a free product as well as through paid plans such as ChatGPT Plus, ChatGPT Pro, ChatGPT Team, and ChatGPT Enterprise.
2.2 DALL-E✎
DALL-E is OpenAI's image-generation model. Given a written description, it produces an original picture in many possible styles, from photorealistic to cartoonish. The first version was unveiled in January 2021, followed by DALL-E 2 in 2022 and DALL-E 3 in 2023. DALL-E is now available inside ChatGPT and through the OpenAI API.
2.3 Sora✎
Sora is a text-to-video model that creates short video clips from written prompts. It was first previewed in 2024 and gradually rolled out to ChatGPT users in 2025.
2.4 Whisper✎
Whisper is an automatic speech recognition system released as open source in 2022. It transcribes spoken audio into written text and can translate speech in dozens of languages.
2.5 Codex✎
Codex is OpenAI's coding assistant. It is trained to read and write computer code in many programming languages and helps power tools like GitHub Copilot. In 2025 and 2026 OpenAI relaunched Codex as a standalone product for developers.
2.6 Atlas, Operator, and other products✎
In October 2025, OpenAI launched ChatGPT Atlas, a web browser for macOS that replaces the traditional address bar with a natural-language interface powered by ChatGPT. The company has also released Operator, an agent that can complete tasks inside web browsers, and sector-specific products such as ChatGPT Health and ChatGPT Edu.
3 Corporate structure✎
OpenAI has an unusual two-part structure designed to balance its original nonprofit mission with the enormous cost of building advanced AI.
The OpenAI Foundation is the nonprofit parent. It holds a controlling stake of about 26% in the for-profit business and is responsible for guiding the company's mission.
OpenAI Group PBC is the operating company. It runs all of OpenAI's products and research and is the entity that raises money from investors. Because it is a public benefit corporation, its directors are legally allowed to consider the public benefit of their decisions, not just shareholder profits.
OpenAI's largest commercial partner is Microsoft, which has invested over $13 billion and provides much of the cloud computing power used to train and run OpenAI's models through Azure. Other major investors include SoftBank, Amazon, and Nvidia.
4 Reception and controversies✎
OpenAI is widely credited with launching the modern era of generative AI and with putting powerful AI tools into the hands of ordinary people. ChatGPT in particular has been praised for helping with writing, learning, and programming, and has been adopted by businesses, schools, and governments worldwide.
At the same time, the company has faced significant criticism. Concerns include:
- Safety and alignment. Several leading AI safety researchers, including co-founder Ilya Sutskever, left OpenAI in 2024 and 2025. Critics worry that the company has reduced its focus on long-term safety as it grows more commercial.
- Drift from the original mission. OpenAI began as a nonprofit dedicated to safe AI for all humanity. After two major restructurings, some observers argue it now resembles an ordinary for-profit technology company.
- Copyright lawsuits. Several authors, news organizations, and artists have sued OpenAI, claiming that its models were trained on their work without permission.
- Labor practices. Investigations have reported that OpenAI used low-paid workers in Kenya, through the firm Sama, to label disturbing training data.
- Energy and water use. Training and running large AI models requires enormous amounts of electricity and cooling water, raising environmental concerns.
A long-running legal dispute with co-founder Elon Musk over the company's nonprofit roots reached trial in 2026.