OpenAI

  Executive Briefing

OpenAI is an American artificial-intelligence research and deployment company headquartered in San Francisco, founded in December 2015 with the stated mission of ensuring that artificial general intelligence (AGI) "benefits all of humanity."¹ It is the most recognizable name in the modern AI boom, having brought large language models from research curiosity to mainstream technology with the November 2022 launch of ChatGPT — the fastest-growing consumer application in history at the time.² The lab's GPT ("generative pre-trained transformer") family, its o-series reasoning models, and multimodal systems such as DALL·E and Sora have defined the competitive frontier that rivals measure themselves against.

The organization is led by co-founder and CEO Sam Altman, alongside president and co-founder Greg Brockman, chief scientist Jakub Pachocki, and COO Brad Lightcap. OpenAI began as a non-profit research lab, restructured in 2019 into a "capped-profit" entity to raise the capital required for large-scale training, and entered a deep commercial and compute partnership with Microsoft that has since exceeded $13 billion in committed investment.³ Its unusual governance — a non-profit board overseeing a for-profit subsidiary — became globally visible during the November 2023 crisis in which Altman was briefly removed and then reinstated within days.

Today OpenAI sits at the center of the AI industry's commercial, scientific, and policy debates: a frontier lab racing toward more capable systems, a consumer-products company with hundreds of millions of weekly users, and a lightning rod for arguments about AI safety, openness, and concentration of power. Its alumni have seeded a remarkable share of the field's other leading labs, including Anthropic, Safe Superintelligence, and Thinking Machines Lab.

  At a Glance

  Origins & Founding

OpenAI was announced on 11 December 2015 as a non-profit AI research company, with an initial funding pledge of up to $1 billion from a group of technology figures including Sam Altman, Elon Musk, Greg Brockman, Reid Hoffman, Peter Thiel, and Amazon Web Services, among others.¹ Brockman (formerly CTO of Stripe) and Altman (then president of Y Combinator) recruited a founding research team that included Ilya Sutskever, a key author of the deep-learning breakthroughs of the early 2010s, lured away from Google Brain to serve as chief scientist, along with Wojciech Zaremba and John Schulman.

The founding thesis was explicitly framed around safety and openness: the founders argued that AGI would be among the most consequential technologies in history and that it should be developed transparently and distributed broadly, rather than concentrated inside a single profit-driven company.¹ Musk and Altman served as initial co-chairs. Within a few years, the tension between this open, non-profit charter and the escalating cost of frontier-scale compute would force the organization to remake its own structure.

  History & Timeline

    2015–2018: Non-profit research lab

OpenAI's early output centered on reinforcement learning and tooling — the OpenAI Gym and Universe platforms, and reinforcement-learning agents that reached professional level at the video game Dota 2 (OpenAI Five). Elon Musk departed the board in February 2018, citing potential conflicts of interest with Tesla's own AI work; he remained an early donor.⁵ The same period saw the lab pivot decisively toward large-scale language modeling and the transformer architecture.

    2019: Capped-profit restructuring and Microsoft

In March 2019 OpenAI created OpenAI LP, a "capped-profit" subsidiary under the non-profit parent, allowing it to raise investment and grant equity while capping investor returns.³ Altman left Y Combinator to become full-time CEO. In July 2019, Microsoft invested $1 billion and became OpenAI's exclusive cloud provider via Azure — the start of a defining partnership.³ The lab also published GPT-2, initially withholding the full model over misuse concerns.

    2020–2022: GPT-3, the API, and ChatGPT

GPT-3 (2020) demonstrated striking few-shot capabilities and launched commercially through the OpenAI API.⁶ DALL·E (2021) extended the lab into text-to-image generation, Codex powered GitHub Copilot, and Whisper delivered open speech recognition. On 30 November 2022, OpenAI released ChatGPT, a conversational interface over a fine-tuned GPT-3.5 model, which reached an estimated 100 million users within two months.²

    2023–present: Frontier scaling, governance crisis, and reasoning models

GPT-4 launched in March 2023, followed by the GPT-4o ("omni") multimodal model in 2024, the o1/o3 reasoning models that introduced inference-time deliberation, and GPT-5 in 2025. This era was punctuated by the November 2023 board crisis (below), a wave of high-profile departures, the Stargate compute initiative, and a continued restructuring toward a public-benefit-corporation model designed to support further large-scale fundraising.⁴

  Mission, Philosophy & Research Agenda

OpenAI's chartered mission is to ensure that artificial general intelligence benefits all of humanity, defined as highly autonomous systems that outperform humans at most economically valuable work.¹ Its published Charter commits the organization to broadly distributed benefit, long-term safety, technical leadership, and a "cooperative orientation" — including a notable pledge to assist rather than compete with a value-aligned, safety-conscious project that comes close to building AGI first.¹

In practice the lab's research agenda combines aggressive capability scaling (ever-larger pre-training, then reasoning via inference-time compute) with alignment work such as reinforcement learning from human feedback (RLHF), scalable oversight, and interpretability. This dual posture — racing to the frontier while arguing that doing so safely requires being at the frontier — is the lab's central and most contested philosophical bet.

  Research & Publications

OpenAI's research has repeatedly reset the field's expectations:

• "Attention"-era scaling — the GPT line operationalized the thesis that transformer language models improve predictably with scale, formalized in the lab's work on scaling laws (2020).
• GPT-3 paper, "Language Models are Few-Shot Learners" (2020) — demonstrated in-context learning at scale.⁶
• InstructGPT / RLHF (2022) — established human-feedback fine-tuning as the standard alignment recipe for chat models.
• GPT-4 Technical Report (2023) — frontier multimodal reasoning, with limited disclosure of architecture and training data.
• Reasoning models (2024–) — the o-series introduced large-scale reinforcement learning on chains of thought, shifting performance gains toward inference-time compute.

The lab also maintains open-source releases including Whisper, CLIP, Triton, and various tokenizers and evals, though its frontier models are closed-weight.

  Models & Products

OpenAI ships through a consumer app (ChatGPT), a developer API, and enterprise offerings, layered over successive model generations:

• GPT-5 — flagship frontier model (2025).
• GPT-4o — natively multimodal model handling text, vision, and audio.
• o3 and o1 — reasoning models optimized for math, code, and science.
• GPT-4 and GPT-3.5-turbo — the models that powered the original ChatGPT and API boom.
• DALL·E 3 — text-to-image generation.
• Sora — text-to-video generation.
• Whisper — open-source speech recognition.

➡️ See the individual model cards for specifications, pricing, and benchmark results.

  People

OpenAI's leadership today centers on Sam Altman (CEO), Greg Brockman (President), Jakub Pachocki (Chief Scientist, succeeding Ilya Sutskever in 2024), and Brad Lightcap (COO). The non-profit board oversees the for-profit subsidiary and has been substantially reconstituted since 2023.

The lab is also one of the field's most prolific talent incubators, and its alumni diaspora maps much of the contemporary AI landscape:

• Dario and Daniela Amodei, Jared Kaplan, Tom Brown, and others left in 2021 to found Anthropic.
• Ilya Sutskever departed in 2024 to co-found Safe Superintelligence (SSI).
• Mira Murati (former CTO) left in 2024 to found Thinking Machines Lab.
• Andrej Karpathy departed to found the education startup Eureka Labs.
• John Schulman and Jan Leike moved to Anthropic, the latter citing disagreements over safety prioritization.

  Funding, Ownership & Business

OpenAI's distinctive structure places a non-profit parent over a capped-profit (and, increasingly, public-benefit-corporation) operating entity built to absorb the enormous capital frontier AI requires.³ Microsoft is the defining investor and infrastructure partner, with cumulative commitments reported above $13 billion and a revenue/IP-sharing arrangement alongside exclusive Azure hosting.³ Subsequent rounds led by investors including Thrive Capital and SoftBank pushed the company's reported valuation into the hundreds of billions of dollars across 2024–2025.⁴

Revenue is driven by ChatGPT subscriptions (Plus, Team, Enterprise), API consumption, and enterprise deals, against very large training and inference costs. The business model — selling access to closed frontier models while subsidizing a free consumer tier — funds the compute that the research agenda depends on.

  Partnerships & Ecosystem

The Microsoft relationship is foundational, spanning Azure compute, distribution through Copilot products, and capital. Beyond it, OpenAI's API underpins thousands of startups and enterprise deployments, and the company has pursued large infrastructure alliances, most prominently the Stargate initiative — a multi-year, multi-hundred-billion-dollar data-center buildout announced in 2025 alongside partners including Oracle and SoftBank.⁴ The ecosystem also includes hardware and consumer-device explorations and a growing set of enterprise and government engagements.

  Compute & Infrastructure

Frontier training and inference make compute OpenAI's central operational constraint. Historically delivered through Microsoft Azure, the lab's compute strategy has expanded toward dedicated, purpose-built capacity via the Stargate program and diversified supply arrangements, reflecting the industry-wide scramble for GPUs, power, and data-center capacity.⁴ Training scale and exact hardware configurations for recent frontier models are not fully disclosed.

  Notable Events & Controversies

• November 2023 board crisis — On 17 November 2023 the non-profit board abruptly removed Sam Altman as CEO; after employee and investor backlash, including a near-total threat of staff resignations, he was reinstated within days and the board was reconstituted.⁷ The episode became the defining case study in AI-lab governance.
• Safety-team departures (2024) — The dissolution of the "Superalignment" team and the exits of Ilya Sutskever and Jan Leike drew public criticism over the lab's safety prioritization.
• Litigation — OpenAI faces copyright suits, most prominently from The New York Times (filed December 2023), and a lawsuit from co-founder Elon Musk alleging departure from the founding non-profit mission.
• Restructuring debate — The ongoing shift toward a public-benefit-corporation structure reignited scrutiny of how the original non-profit charter is honored.

  Competitive Position

OpenAI is widely regarded as a frontier leader, competing most directly with Google DeepMind, Anthropic, Meta AI, and xAI, with strong open-weight pressure from labs such as DeepSeek and Mistral. Its principal advantages are brand and consumer reach (ChatGPT), the Microsoft partnership, and a deep research bench; its principal risks are capital intensity, governance scrutiny, talent attrition to rivals it helped seed, and the commoditization pressure of fast-improving open models.

  Outlook & Roadmap

OpenAI's stated direction is continued progress toward AGI: more capable multimodal and reasoning models, deeper agentic capabilities, and expansion of compute via Stargate-scale infrastructure.⁴ Strategically, the company is balancing further large-scale fundraising and structural change against its founding safety commitments. Specific unreleased model timelines remain speculative and should be treated as such pending official announcements.

---

  References

1. OpenAI — About / Charter
2. ChatGPT reaches 100M users — Reuters
3. Microsoft and OpenAI partnership
4. History of OpenAI — Wikipedia
5. Elon Musk and OpenAI
6. Language Models are Few-Shot Learners (GPT-3)
7. Removal and reinstatement of Sam Altman — Wikipedia

  References

1. OpenAI's mission, founding announcement, and Charter — openai.com/about. ↩ ↩² ↩³ ↩⁴ ↩⁵

2. ChatGPT was reported as the fastest-growing consumer app to that date — Reuters, Feb 2023. ↩ ↩²

3. Capped-profit restructuring and the Microsoft partnership — Wikipedia: OpenAI. ↩ ↩² ↩³ ↩⁴ ↩⁵

4. Valuation, Stargate, and restructuring details evolve rapidly; figures are reported estimates — History of OpenAI. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

5. Elon Musk's 2018 board departure — Wikipedia: OpenAI. ↩

6. GPT-3 — Brown et al., 2020. ↩ ↩²

7. The November 2023 leadership crisis — Wikipedia. ↩