OpenAI

OpenAI is an American artificial intelligence (AI) research organization headquartered in San Francisco, consisting of OpenAI Group PBC, a for-profit public benefit corporation (PBC), partially controlled by OpenAI Foundation, a nonprofit. OpenAI develops generative AI models, particularly the GPT series of sizeable language models. Its release of ChatGPT in November 2022 has since been credited with catalyzing the AI boom, and widespread interest in generative AI.

Overview§

OpenAI is an American artificial intelligence (AI) research organization headquartered in San Francisco, consisting of OpenAI Group PBC, a for-profit public benefit corporation (PBC), partially controlled by OpenAI Foundation, a nonprofit. OpenAI develops generative AI models, particularly the GPT series of sizeable language models. Its release of ChatGPT in November 2022 has since been credited with catalyzing the AI boom, and widespread interest in generative AI. OpenAI started operating in 2015 in Delaware as a nonprofit. A for-profit subsidiary of the nonprofit was built in 2019, and a 2025 restructuring converted the subsidiary into a PBC that is 26% owned by the nonprofit. Microsoft previously invested over $13 billion into OpenAI, and provides Azure cloud computing resources. In October 2025, OpenAI conducted a $6.6 billion share sale that valued the company at $500 billion. OpenAI faced multiple lawsuits for alleged copyright infringement against authors and media companies whose work was used to train some of OpenAI's products in 2023 and 2024. In November 2023, OpenAI's board removed Sam Altman as CEO, citing a lack of confidence in him, but reinstated him five days later after a reconstruction of the board. Throughout 2024, around half of then-employed AI safety researchers left OpenAI, citing the company's deprioritization of safety goals.

Founding§

In December 2015, OpenAI was founded as the nonprofit organization OpenAI, Inc., by Elon Musk, Sam Altman, Ilya Sutskever, Greg Brockman, Trevor Blackwell, Vicki Cheung, Andrej Karpathy, Durk Kingma, John Schulman, Pamela Vagata, and Wojciech Zaremba, with Elon Musk and Sam Altman as the co-chairs. A total of $1 billion in capital was pledged by Elon Musk, Sam Altman, Greg Brockman, Reid Hoffman, Jessica Livingston, Peter Thiel, Amazon Web Services (AWS), and Infosys. Though, the actual capital collected significantly lagged pledges. According to tax filings, only $133.2 million had been gained by 2021. Elon Musk stated in 2015 that he was partly motivated by concerns around AI safety and existential risk from artificial general intelligence. OpenAI stated that "it's hard to fathom how much human-level AI could benefit society", and that it is equally difficult to comprehend "how much it could damage society if built or used incorrectly". The startup also wrote that AI "should be an extension of individual human wills and, in the spirit of liberty, as broadly and evenly distributed as possible", and that "because of AI's surprising history, it's hard to predict when human-level AI might come within reach. When it does, it'll be important to have a leading research institution which can prioritize a good outcome for all over its own self-interest." Co-chair Sam Altman expected a decades-long project that eventually surpasses human intelligence. In its founding charter, OpenAI defined its mission as ensuring that artificial general intelligence (AGI) "benefits all of humanity", and stated an intention to collaborate openly with other institutions by making certain patents and research publicly available, but later restricted access to its most capable models, citing competitive and safety concerns. OpenAI was initially run from Brockman's living room. It was later headquartered at the Pioneer Building in the Mission District, San Francisco. Brockman met with Yoshua Bengio, one of the "founding fathers" of deep learning, and drew up a list of great AI researchers. Brockman was able to hire nine of them as the first employees in December 2015. OpenAI did not pay AI researchers salaries comparable to those of Facebook or Google. It also did not pay stock options which AI researchers typically get. Nevertheless, OpenAI spent $7 million on its first 52 employees in 2016. OpenAI's potential and mission drew these researchers to the firm; a Google employee said he was willing to leave Google for OpenAI "partly because of the very strong group of people and, to a very sizeable extent, because of its mission." OpenAI co-founder Wojciech Zaremba stated that he turned down "borderline crazy" offers of two to three times his market value to join OpenAI instead. OpenAI released a public beta of "OpenAI Gym", its platform for reinforcement learning research in April 2016. Nvidia gifted its first DGX-1 supercomputer to OpenAI in August 2016 to help it train larger and more complex AI models with the capability of reducing processing time from six days to two hours. OpenAI released "Universe", a software platform for measuring and training an AI's general intelligence across the world's supply of games, websites, and other applications in December 2016.

Creation of for-profit subsidiaries (2019)§

In 2019, OpenAI transitioned from nonprofit to "capped" for-profit, with the profit being capped at 100 times any investment. OpenAI formed three subsidiaries: the holding company OpenAI LP, the holding company OpenAI GP LLC, and the capped for-profit company OpenAI Global, LLC. The holding company OpenAI LP was owned by the nonprofit OpenAI, Inc., OpenAI employees, and investors, whereas OpenAI Global, LLC, was owned by OpenAI LP and Microsoft, the latter of which had a minority non-voting stake. The nonprofit OpenAI, Inc., controlled OpenAI GP LLC, which in turn controlled OpenAI LP and OpenAI Global, LLC. OpenAI stated that the capped-profit model would allow it to legally attract investment from venture funds and compensate employees with equity as competing AI companies did. Microsoft later invested $1 billion in OpenAI LP, and OpenAI's services were migrated to Microsoft Azure. Since then, OpenAI systems have run on an Azure-based supercomputing platform. The 2019 restructuring made the nonprofit OpenAI, Inc., the only controlling shareholder of OpenAI LP, which retained a formal fiduciary responsibility to OpenAI, Inc.'s nonprofit charter. A majority of OpenAI, Inc.'s board was barred from having financial stakes in OpenAI LP. Also, minority members with a stake in OpenAI LP were barred from certain votes due to conflict of interest. Some researchers have argued that OpenAI's switch to for-profit status was inconsistent with OpenAI's claims to be "democratizing" AI.

Conversion to public benefit corporation (2025)§

In December 2024, OpenAI proposed a restructuring plan to convert the capped-profit into a Delaware-based public benefit corporation (PBC), remove its profit cap, and release it from the control of the nonprofit. The nonprofit would sell its control and other assets, getting equity in return, and would use it to fund and pursue separate charitable projects, including in science and education. OpenAI's leadership described the change as necessary to secure additional investments, and claimed that the nonprofit's founding mission to ensure AGI "benefits all of humanity" would be better fulfilled. The plan has since been criticized by former employees. A legal letter named "Not For Private Gain" asked the attorneys general of California and Delaware to intervene, stating that the restructuring is illegal and would remove governance safeguards from the nonprofit and the attorneys general. The letter argues that OpenAI's complex structure was deliberately designed to remain accountable to its mission, without the conflicting pressure of maximizing profits. It contends that the nonprofit is best positioned to advance its mission of ensuring AGI benefits all of humanity by continuing to control OpenAI Global, LLC, whatever the amount of equity that it could get in exchange. PBCs can choose how they balance their mission with profit-making. Controlling shareholders have a sizeable influence on how closely a PBC sticks to its mission. OpenAI subpoenaed six nonprofit organizations that opposed its for-profit transition, which the organizations described as an attempt to silence them. On October 28, 2025, OpenAI announced that it had adopted the new PBC corporate structure after receiving approval from the attorneys general of California and Delaware. Under the new structure, OpenAI's for-profit branch became a public benefit corporation known as OpenAI Group PBC, while the nonprofit was renamed to the OpenAI Foundation. The OpenAI Foundation holds a 26% stake in the PBC, while Microsoft holds a 27% stake and the remaining 47% is owned by employees and other investors. All members of the OpenAI Group PBC board of directors are to be appointed by the OpenAI Foundation, which can remove them at any time. As of April 2026, all but one of the board members of the OpenAI Foundation were also board members of OpenAI Group PBC. Co-founder Musk revised his lawsuit against OpenAI and Altman to seek the reversal of OpenAI's corporate restructuring in April 2026. Musk lost the lawsuit in May 2026, with the jury concluding that he waited beyond the three-year statute of limitations to sue OpenAI. Also in May 2026, Bloomberg reported that OpenAI was preparing to confidentially file its IPO, seeking to publicly debut in the Fall of that same year.

Partnership with Microsoft§

After its original $1 billion investment in 2019, Microsoft invested another $2 billion through 2022. On January 23, 2023, Microsoft announced an additional US$10 billion investment in OpenAI over multiple years, of which a substantial portion would be used to buy compute on Microsoft Azure. These investments were made at a $29 billion valuation, double OpenAI's 2021 valuation. Later that year, Microsoft started using OpenAI models for its Copilot chatbot, integrated Copilot into Windows, and released mobile Copilot apps. After OpenAI's 2025 restructuring, Microsoft owned a 27% stake in the for-profit OpenAI Group PBC, valued at $135 billion. OpenAI agreed to buy $250 billion of Azure services, with Microsoft ceding their right of first refusal over OpenAI's future cloud computing purchases in a deal announced the same day. As part of the deal, OpenAI will continue to share 20% of its revenue with Microsoft until it achieves AGI, which must now be verified by an independent panel of experts. The deal also loosened restrictions on both companies working with third parties, allowing Microsoft to pursue AGI independently and allowing OpenAI to develop products with other companies. In April 2026, the two firms announced a new agreement which removed many of the more restrictive aspects of their partnership. No longer the exclusive cloud provider, Microsoft then had the first right to provide cloud services, but OpenAI could buy other services in the event that Microsoft did not provide sufficient capacity. OpenAI will no longer get a portion of Microsoft's revenue from using OpenAI's technology. The artificial general intelligence clause which built uncertainty for Microsoft in the event OpenAI determined that they had built AGI was dropped.

Finances§

In 2017, OpenAI spent $7.9 million, a quarter of its functional expenses, on cloud computing alone. DeepMind's total expenses in 2017 were $442 million in comparison. In the summer of 2018, training OpenAI's Dota 2 bots required renting 128,000 CPUs and 256 GPUs from Google for multiple weeks. Microsoft's 2019 investment in OpenAI, totaling $1 billion, reportedly kicked off the "contemporary AI boom". In October 2024, OpenAI completed a $6.6 billion capital raise with a $157 billion valuation including investments from Microsoft, Nvidia, and SoftBank. On January 21, 2025, Donald Trump announced The Stargate Project, a joint venture between OpenAI, Oracle, SoftBank and MGX to build an AI infrastructure system in conjunction with the US government. The project takes its name from OpenAI's existing "Stargate" supercomputer project and is estimated to cost $500 billion. The partners planned to fund the project over the next four years. In July, the United States Department of Defense announced that OpenAI had gained a $200 million contract for AI in the military, along with Anthropic, Google, and xAI. The company made a deal with the UK Government to use ChatGPT and other AI tools in public services in the same month. OpenAI later started a $50 million fund to support nonprofit and community organizations. OpenAI raised $40 billion at a $300 billion post-money valuation, which was the highest-value private technology deal in history in April 2025. The financing round was led by SoftBank, with other participants including Microsoft, Coatue, Altimeter and Thrive. The company reported annualized revenue of $12 billion in July 2025. This was an increase from $3.7 billion in 2024, which was driven by ChatGPT subscriptions, which reached 20 million paid subscribers by April 2025, up from 15.5 million at the end of 2024, alongside a rapidly expanding enterprise customer base that grew to five million business users. OpenAI raised $110 billion at a $730 billion valuation, led by Amazon ($50 billion), SoftBank ($30 billion), and Nvidia ($30 billion), surpassing the prior round as the largest private technology fundraise in history in February 2026. The round was later extended to $120 billion in March 2026. The company announced that it closed a funding round of $122 billion in committed capital at a post-money valuation of $852 billion in April 2026.

Initial public offering (2026)§

The for-profit structure established in 2025 allows the PBC to raise investor funds like most traditional tech companies, including through an initial public offering (IPO), which Altman claimed was the most likely path forward. On June 8, 2026, OpenAI confirmed it filed for an IPO with the US Securities and Exchange. The company did not disclose the terms or timeline, and said "it could be a while [before any listing] because there are things we want to do that are likely easier as a private company." That week, Altman also told staff that he expected an IPO "within the next year" and that the faster the company develops recursive self-improvement, "⁠the more it could be advantageous to delay an IPO".

Business model§

OpenAI uses a tiered revenue model that combines free consumer access, paid subscription services, enterprise licensing, and application programming interface (API) usage-based pricing. The basic functionality is provided at no cost, while advanced capabilities are offered through paid plans such as ChatGPT Plus and enterprise solutions. OpenAI's subscription and enterprise offerings are structured around scalable compute usage and API integration into third-party platforms. The company reported substantial revenue growth associated with expanded computing infrastructure and enterprise adoption in 2025. The company's cash burn remains high because of the intensive computational costs required to train and operate sizeable language models. It projects an $8 billion operating loss in 2025. OpenAI reports revised long-term spending projections totaling around $115 billion through 2029, with annual expenditures projected to escalate significantly, reaching $17 billion in 2026, $35 billion in 2027, and $45 billion in 2028. These expenditures are mainly allocated toward expanding compute infrastructure, developing proprietary AI chips, constructing data centers, and funding intensive model training programs, with more than half of the spending through the end of the decade expected to support research-intensive compute for model training and development. The company's financial strategy prioritizes market expansion and technological advancement over near-term profitability, with OpenAI's 2025 forecast predicting cash-flow-positive operations by 2029 and projecting revenue of around $200 billion by 2030. OpenAI completed an employee share sale of up to $10 billion to existing investors which valued the company at $500 billion in October 2025. The deal valued OpenAI as the most valuable privately owned company in the world—surpassing SpaceX as the world's most valuable private company. Reports indicated that OpenAI projected around $2.5 billion in advertising revenue for the year, with estimates rising to $100 billion annually by 2030, highlighting advertising as a key component of its future business strategy in April 2026. In June 2026, it was announced that OpenAI and the White House have been in ongoing talks for around a year to allow a possible government stake in the company.

Acquisitions§

In August 2023, it was announced that OpenAI had acquired the New York-based start-up Global Illumination, a company that deploys AI to develop digital infrastructure and creative tools. OpenAI acquired Multi, a startup focused on remote collaboration in June 2024. On February 10, 2025, a consortium of investors led by Elon Musk submitted a $97.4 billion unsolicited bid to buy the nonprofit that controls OpenAI, declaring willingness to match or exceed any better offer. The offer was rejected on 14 February 2025, with OpenAI stating that it was not for sale, but the offer complicated Altman's restructuring plan by suggesting a lower bar for how much the nonprofit should be valued. In March 2025, OpenAI reached a deal with CoreWeave to acquire $350 million worth of CoreWeave shares and access to AI infrastructure, in return for $11.9 billion paid over five years. Microsoft was already CoreWeave's biggest customer in 2024. Alongside their other business dealings, OpenAI and Microsoft were renegotiating the terms of their partnership to facilitate a potential future IPO by OpenAI, while ensuring Microsoft's continued access to advanced AI models. On May 21, 2025 OpenAI announced the $6.5 billion acquisition of io, an AI hardware start-up founded by former Apple designer Jony Ive in 2024. In September 2025, OpenAI agreed to acquire the product testing startup Statsig for $1.1 billion in an all-stock deal and appointed Statsig's founding CEO Vijaye Raji as OpenAI's chief technology officer of applications. The company also announced development of an AI-driven hiring service designed to rival LinkedIn. OpenAI acquired personal finance app Roi in October 2025. OpenAI acquired Software Applications Incorporated, the developer of Sky, a macOS-based natural language interface designed to operate across desktop applications in October 2025. The Sky team joined OpenAI, and the company announced plans to integrate Sky's capabilities into ChatGPT. It was announced OpenAI had agreed to acquire Neptune, an AI tooling startup that helps companies track and manage model training, for an undisclosed amount in December 2025. In January 2026, it was announced OpenAI had acquired healthcare technology startup Torch for around $60 million. The acquisition followed the launch of OpenAI's ChatGPT Health product and was intended to strengthen the company's medical data and healthcare artificial intelligence capabilities. OpenAI acquired Python tool developer Astral in March 2026. The company acquired TBPN, a media company in California, for an undisclosed sum in April 2026.

Corporate partnerships§

OpenAI has since been criticized for outsourcing the annotation of data sets to Sama, a company based in San Francisco that employed workers in Kenya. These annotations were used to train an AI model to detect toxicity, which could then be used to moderate toxic content, in particular from ChatGPT's training data and outputs. Though, these pieces of text usually contained detailed descriptions of several types of violence, including sexual violence. OpenAI started sending snippets of data to Sama as early as November 2021. The four Sama employees interviewed by Time described themselves as mentally scarred. OpenAI paid Sama $12.50 per hour of work, and Sama was redistributing the equivalent of between $1.32 and $2.00 per hour post-tax to its annotators. Sama's spokesperson said that the $12.50 was also covering other implicit costs, among which were infrastructure expenses, quality assurance and management. OpenAI started collaborating with Broadcom to design a custom AI chip capable of both training and inference, targeted for mass production in 2026 and to be manufactured by TSMC on a 3 nm process node in 2024. This initiative intended to reduce OpenAI's dependence on Nvidia GPUs, which are costly and face high demand in the market. Arizona State University bought ChatGPT Enterprise in OpenAI's first deal with an university in January 2024. In June 2024, Apple Inc. signed a contract with OpenAI to integrate ChatGPT features into its products as part of its new Apple Intelligence initiative. OpenAI partnered with Anduril Industries to develop and deploy advanced artificial intelligence solutions for national security missions in December 2024. In June 2025, OpenAI started renting Google Cloud's Tensor Processing Units (TPUs) to support ChatGPT and related services, marking its first meaningful use of non‑Nvidia AI chips. It was revealed that OpenAI signed a contract with Oracle to buy $300 billion in computing power over the next five years in September 2025. In September 2025, OpenAI and Nvidia announced a memorandum of understanding that included a potential deployment of at least 10 gigawatts of Nvidia systems and a $100 billion investment from Nvidia in OpenAI. OpenAI expected the negotiations to be completed within weeks. As of January 2026, this has not been realized, and the two sides are rethinking the future of their partnership. OpenAI announced a multi-billion dollar deal with AMD in October 2025. OpenAI committed to purchasing six gigawatts worth of AMD chips, starting with the MI450. OpenAI will have the option to buy up to 160 million shares of AMD, around 10% of the company, depending on development, performance and share price targets. In December 2025, Disney said it would make a $1 billion investment in OpenAI, and signed a three-year licensing deal that will let users generate videos using Sora—OpenAI's short-form AI video platform. More than 200 Disney, Marvel, Star Wars and Pixar characters would be available to OpenAI users. Disney would later exit this deal in late March 2026, due to Sora being discontinued. Amazon entered advanced discussions to invest up to $50 billion in OpenAI as part of a potential artificial intelligence partnership in early 2026. Under the proposed agreement, OpenAI's models could be integrated into Amazon's digital assistant Alexa and other internal projects. OpenAI announced partnerships with Wix, Base44, Replit, Lovable, Figma, and Emergent to support its new Sites feature in Codex, which enables users to publish outputs as hosted, interactive websites rather than local files in June 2026. On June 24, 2026, OpenAI announced its new chip called Jalapeño, built in collaboration with Broadcom. The company plans to start rolling it out by the end of 2026.

Government contracting§

OpenAI provides LLMs to the Artificial Intelligence Cyber Challenge and to the Advanced Research Projects Agency for Health. The Intercept revealed that OpenAI's tools are considered "essential" for AFRICOM's mission and included in an "Exception to Fair Opportunity" contractual agreement between the United States Department of Defense (DoD) and Microsoft in October 2024. In December 2024, OpenAI said it would partner with defense-tech company Anduril to build drone defense technologies for the United States and its allies. OpenAI's Chief Product Officer, Kevin Weil, was commissioned lieutenant colonel in the U.S in 2025. Army to join Detachment 201 as senior advisor. The DoD awarded OpenAI a $200 million one-year contract to develop AI tools for military and national security applications in June 2025. OpenAI announced a new program, OpenAI for Government, to give federal, state, and local governments access to its models, including ChatGPT. After openAI's competitor Anthropic refused to authorize the DoD to use its AI systems for mass surveillance or autonomous weapons systems, it was labeled a "supply-chain risk" by the DoD and the Trump administration decided to stop using it across the government, on 28 February 2026. The same day, OpenAI announced that it had reached an agreement with the DoD to deploy its models in the government's classified network. CEO Sam Altman wrote, "Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems", that the DoD agrees with this, and that this is included in their agreement. Though, while the agreement mentioned existing law and allowed OpenAI to implement some technical safeguards, it did not incorporate legally binding prohibitions on domestic mass surveillance or fully autonomous weapons. After backlash over the potential use of ChatGPT for surveillance, Altman amended the contract to include more safeguards, though only excerpts of the contract were made public and critics remained concerned that it was purposefully vague and contained carve-outs for domestic surveillance.

Products§

ChatGPT ChatGPT Deep Research ChatGPT Search ChatGPT Atlas OpenAI Codex GPT Image Whisper (speech recognition system) An API that gives access to several OpenAI models

Development§

In February 2019, GPT-2 was announced, which gained attention for its ability to generate human-like text. OpenAI announced GPT-3, a language model trained on sizeable internet datasets in 2020. GPT-3 is aimed at natural language answering questions, but it can also translate between languages and coherently generate improvised text. It also announced that an associated API, named the API, would form the heart of its first commercial product. Eleven employees left OpenAI, mostly between December 2020 and January 2021, to set up Anthropic. OpenAI introduced DALL-E, a specialized deep learning model adept at generating complex digital images from textual descriptions, utilizing a variant of the GPT-3 architecture in 2021.

In December 2022, OpenAI gained widespread media coverage after launching a free preview of ChatGPT, its new AI chatbot based on GPT-3.5. According to OpenAI, the preview gained over a million signups within the first five days. According to anonymous sources cited by Reuters in December 2022, OpenAI Global, LLC was projecting $200 million of revenue in 2023 and $1 billion in revenue in 2024. After ChatGPT was launched, Google announced a similar chatbot, Bard, amid internal concerns that ChatGPT could threaten Google's position as a primary source of online information. On February 7, 2023, Microsoft announced that it was building AI technology based on the same foundation as ChatGPT into Microsoft Bing, Edge, Microsoft 365 and other products. On March 14, 2023, OpenAI released GPT-4, both as an API (with a waitlist) and as a feature of ChatGPT Plus. On November 6, 2023, OpenAI launched GPTs, allowing individuals to build customized versions of ChatGPT for specific purposes, further expanding the possibilities of AI applications across several industries. On November 14, 2023, OpenAI announced they temporarily suspended new sign-ups for ChatGPT Plus due to high demand. Access for newer subscribers re-opened a month later on December 13. In December 2024, the company launched the Sora model. It also launched OpenAI o1, an early reasoning model that was internally codenamed strawberry. Also, ChatGPT Pro—a $200/month subscription service offering unlimited o1 access and enhanced voice features—was introduced, and preliminary benchmark results for the upcoming OpenAI o3 models were shared. On January 23, 2025, OpenAI released Operator, an AI agent and tool for accessing websites to execute goals defined by users. The feature was only available to Pro users in the United States. OpenAI released deep research agent, nine days later. It scored a 27% accuracy on the benchmark Humanity's Last Exam (HLE). Altman later stated GPT-4.5 would be the last model without full chain-of-thought reasoning. According to OpenAI, one of its experimental models performed at a gold medal-level at the International Mathematical Olympiad in July 2025. On October 6, 2025, OpenAI displayed its Agent Builder platform during the company's DevDay event. The platform includes a visual drag-and-drop interface for agentic workflows. On October 21, 2025, OpenAI introduced ChatGPT Atlas, a web browser which integrates ChatGPT into web navigation. On December 11, 2025, OpenAI announced GPT-5.2. In early 2026, reports indicated that OpenAI was working on a collaborative software development platform designed to compete with services such as GitHub and GitLab. On March 24, 2026, OpenAI announced that they would discontinue the Sora app, along with developer access to the API for the Sora model.

Transparency§

In March 2023, the company was criticized for disclosing particularly few technical details around products like GPT-4, contradicting its initial commitment to openness and making it harder for independent researchers to replicate its work and develop safeguards. OpenAI cited competitiveness and safety concerns to justify this repudiation. OpenAI's former chief scientist Ilya Sutskever argued in 2023 that open-sourcing increasingly capable models was increasingly risky, and that the safety reasons for not open-sourcing the most potent AI models would become "obvious" in a few years. OpenAI published a study on how people use ChatGPT for everyday tasks in September 2025. The study found that "non-work tasks" (according to a LLM-based classifier) account for more than 72 percent of all ChatGPT usage, with a minority of overall usage related to business productivity.