The OpenAI IPO: A Company Racing to Build the Last Invention Mankind Will Ever Need

PART 1 OF 3

SPECIAL REPORT: THE OPENAI IPO

OpenAI has filed confidentially for an IPO that could be the most consequential in the history of American technology. Here are the twenty-five things every investor and every American must understand before the bell rings.

PART ONE: HOW A CHARITY BECAME THE MOST VALUABLE STARTUP IN HISTORY

The Beginning: A Nonprofit in a Converted Victorian

On a gray December evening in 2015, in the back rooms of a converted Victorian mansion in San Francisco's Mission District, twelve people gathered around a table with a single shared conviction: that the most powerful technology in the history of human civilization was being built inside the research labs of the world's largest corporations — and that if it remained there, without independent oversight or mission-driven development, the consequences could be catastrophic. They had pooled together one billion dollars in initial commitments. They had no product. They had no revenue. They had an idea, and a mission statement that sounded more like philosophy than a business plan: to ensure that artificial general intelligence benefits all of humanity.

The company they founded was OpenAI. The investors and founders in that room included Elon Musk, then the most famous entrepreneur on earth, who pledged $1 billion and served as co-chairman; Sam Altman, the then-thirty-year-old president of Y Combinator, who would eventually become CEO; Ilya Sutskever, a Canadian AI researcher who had been the principal student of Geoffrey Hinton — the man often called the godfather of modern deep learning — and who would become the company's first chief scientist; and Greg Brockman, the former chief technology officer of Stripe, who would lead the company's technical infrastructure. They were joined by Trevor Blackwell, Vicki Cheung, Andrej Karpathy, Durk Kingma, John Schulman, Pamela Vagata, and Wojciech Zaremba — eleven founders in total. They structured it as a nonprofit, took no salaries for the first year, and opened their research to the world.

Ten and a half years later, OpenAI is many things it once promised it would never be: a for-profit company, albeit organized as a Public Benefit Corporation; a business with an estimated $25 billion in annualized revenue growing at a pace that defies conventional financial modeling; an employer offering average stock-based compensation of $1.5 million per person — the highest of any technology startup in history; a company that has accepted investments from the Saudi Arabian government's sovereign wealth fund, from SoftBank, from Amazon, from Nvidia, and from the government of the United Arab Emirates. It is, by the private market's current reckoning, worth $852 billion — a figure that exceeds the gross domestic product of Saudi Arabia, the Netherlands, and Switzerland combined.

And on May 22, 2026, it filed confidentially with the Securities and Exchange Commission for an initial public offering.

The ChatGPT Moment: When Everything Changed

The single event that transformed OpenAI from a well-funded research laboratory into a global phenomenon occurred on November 30, 2022, when the company released ChatGPT to the public with no fanfare, no coordinated press campaign, and no retail advertising budget. Within five days, the product had one million users. Within two months, it had one hundred million — the fastest consumer product adoption in recorded history, outpacing Instagram, TikTok, and every other platform that had ever existed. Sam Altman described the response as 'not what we expected.' That was the understatement of the decade.

ChatGPT did something that decades of artificial intelligence hype had failed to do: it made the technology viscerally real to ordinary people. You could hold a conversation with it. You could ask it to write a letter, explain a concept, draft a legal contract, debug a piece of code, or compose a poem — and it would do so with a fluency and coherence that felt, for the first time in computing history, genuinely human. The moment it launched, the race was on. Google, whose own researchers had invented the underlying transformer architecture that made ChatGPT possible, went to 'code red.' Microsoft announced a $10 billion follow-on investment in OpenAI. A thousand startups pivoted to AI overnight. And a company that had operated in research obscurity for seven years found itself at the center of what every credible analyst now agrees is the most consequential technological transition since the invention of the internet.

The Most Dramatic Boardroom Crisis in Silicon Valley History

On November 17, 2023, in what became the most dramatic corporate governance crisis in the history of American technology, OpenAI's board of directors fired Sam Altman as chief executive — notifying him minutes before the announcement in a brief Google Meet call while he was at the Las Vegas Grand Prix. The board's terse public statement said only that Altman had not been 'consistently candid' with the board, and that the board 'no longer had confidence in his ability to continue leading OpenAI.'

What followed was five days of organizational chaos that would have destroyed any conventional company. OpenAI's president Greg Brockman resigned within hours. More than 700 of the company's approximately 770 employees — nearly the entire workforce — signed an open letter threatening to resign and follow Altman to Microsoft unless the board reversed course. Microsoft, which had invested $13 billion in OpenAI and stood to lose hundreds of billions in strategic value, announced it would hire Altman and Brockman immediately. Ilya Sutskever, the chief scientist who had initially voted to oust Altman, posted publicly that he 'deeply regretted' his decision. Within 96 hours, the board that had fired Altman had dissolved itself. Altman was reinstated. A new board was constituted with Bret Taylor — former Salesforce co-CEO and the chairman of Twitter's board at the time of Musk's acquisition — as its chair. The crisis was over. OpenAI survived. It would never be quite the same organization again.

The firing, subsequent investigation, and reinstatement revealed a company in the grip of profound internal tension: between its original nonprofit mission and its commercial reality; between the safety-first researchers who worried that the pace of development was outrunning human understanding; and between a CEO who was building one of the fastest-growing companies in history and a board that had been constituted for a different era. The investigation that followed found no evidence of financial fraud or sexual misconduct, as early rumors had speculated — only a pattern of 'communications failures' between Altman and the board. What it revealed, fundamentally, was that OpenAI had outgrown the governance structure it had built for itself in a Victorian mansion in 2015.

The Restructuring: From Nonprofit to Public Benefit Corporation

The structural transformation of OpenAI was years in the making and finally completed on October 28, 2025. The original nonprofit — now renamed the OpenAI Foundation — converted its capped-profit subsidiary into a fully-fledged Public Benefit Corporation called OpenAI Group PBC, allowing the company to issue conventional equity, pay market-rate compensation, and pursue a traditional initial public offering. The Foundation holds approximately 26 percent of the new PBC's equity — a stake worth approximately $220 billion at the $852 billion valuation — but crucially, that equity stake understates the Foundation's actual power. Through a special class of shares (Class N Common Stock), the Foundation retains the sole authority to appoint and remove every member of the OpenAI Group PBC board of directors. No investor — not Microsoft, which holds roughly 27 percent of the equity, and not SoftBank or any other financial backer — has a direct board appointment right. The Foundation controls governance; the investors own economics.

The restructuring was not without controversy. Public Citizen, a consumer advocacy organization, argued that 'subordinating the nonprofit to the for-profit is impermissible' under California law. The California Attorney General's office, which oversees nonprofit corporations in the state, negotiated specific governance conditions with OpenAI as a condition of approving the conversion. State attorneys general in Texas and Delaware, where OpenAI has operations and is incorporated respectively, also scrutinized the transaction. And Elon Musk, who had left OpenAI's board in 2018 and spent the next several years building a competing AI company, filed a federal lawsuit in 2024 alleging that the conversion violated the terms under which he had originally made his founding pledge. A federal jury in Oakland, California dismissed all of Musk's claims on May 18, 2026 — three days before OpenAI filed its S-1 — finding that Musk had waited beyond the statute of limitations to bring his case. Musk has announced he will appeal.

 PART TWO: THE TOP 25 THINGS YOU MUST KNOW

1. The Filing: A Confidential S-1 at $852 Billion

On May 22, 2026, OpenAI filed a confidential draft registration statement on Form S-1 with the U.S. Securities and Exchange Commission — the formal legal step that initiates the process of becoming a publicly traded company. The filing was confidential, as permitted for any company under SEC rules revised in 2017 — which extended the confidential draft submission process well beyond its original JOBS Act limitation to Emerging Growth Companies, allowing any issuer to submit a draft S-1 privately provided the registration statement is publicly filed at least 15 days before the roadshow begins. The full prospectus will not become public until OpenAI formally launches its roadshow, expected in late summer or early autumn of 2026.

The valuation at which OpenAI will seek to go public has not been formally stated in the filing. But its most recent private market financing — a $122 billion round that closed on March 31, 2026, the largest private technology fundraise in history — valued the company at $852 billion on a post-money basis. Bankers familiar with the transaction expect the IPO to target a valuation between $852 billion and $1 trillion, with a capital raise in the range of $40 billion to $60 billion. Goldman Sachs and Morgan Stanley are the lead underwriters. A Nasdaq listing is targeted, with September 2026 as the current planning assumption, though OpenAI has stated publicly that it 'has not decided on timing yet.'

2. The Scale: 900 Million Weekly Users, 50 Million Paying Customers

The core commercial reality of OpenAI today is this: ChatGPT has 900 million weekly active users — a figure exceeded, among the world's nations, only by the populations of India and China. Of those, more than 50 million are paid subscribers. Another 9 million are paying business users. The company's flagship product, which did not exist four years ago, has become one of the most widely used software applications in the history of computing — a fact with profound implications for both the bull case and the regulatory risk section of any prospectus.

Consumer revenue — primarily ChatGPT Plus at $20 per month, ChatGPT Team, and the enterprise tiers — accounts for approximately 60 percent of total revenue, with API access and enterprise contracts representing the remaining 40 percent. That ratio is shifting rapidly in enterprise's favor. Large enterprise customers are the fastest-growing segment of OpenAI's business.

3. The Revenue Trajectory: $0 to $25 Billion in Three Years

OpenAI's revenue growth is without precedent in the history of enterprise software. At the end of 2022 — the year ChatGPT launched — OpenAI's annualized revenue was approximately $28 million, primarily from API access sold to developers. One year later, at the end of 2023, that figure had risen to $2 billion. By year-end 2024, it reached approximately $6 billion. By year-end 2025, it had crossed $20 billion. By February 2026, it had reached $25 billion on an annualized run-rate basis — a growth trajectory of approximately 80x in under three years.

CEO Sam Altman described the company's early 2026 performance as 'crazy 80x growth' — a phrase that sounds like promotional hyperbole until you review the actual numbers. The compound annual growth rate over this period exceeds any comparable technology company at similar scale. Salesforce took 10 years to reach $10 billion in revenue. Google took 5 years. Amazon Web Services took 9 years. OpenAI has done it in roughly two and a half.

4. The Losses: $14 Billion Expected in 2026 — and Not Profitable Until 2030

The other side of OpenAI's extraordinary growth story is an extraordinary cash burn. The company is projected to lose approximately $14 billion in 2026 alone — a monthly cash consumption of approximately $1.17 billion. Internal financial projections, reviewed by multiple news organizations, do not show OpenAI reaching profitability until approximately 2030. In Q1 2026, the company lost $1.22 for every $1.00 of revenue generated — a loss ratio that, in a conventional business, would be disqualifying, and in OpenAI's case reflects the extraordinary cost of training and deploying frontier AI models at global scale.

The core reason for the losses is compute. Training a frontier large language model requires thousands of specialized chips — primarily Nvidia H100 and H200 GPUs — running continuously for months. Inference, meaning actually responding to the nearly one billion weekly user sessions, requires a parallel infrastructure of comparable scale. OpenAI is building both simultaneously, while also investing in next-generation model development and in Stargate, the $500 billion infrastructure project described below. Whether the losses narrow on the current trajectory, and whether 2030 profitability is achievable, will be among the most scrutinized questions of the IPO roadshow.

5. The Largest Private Fundraise in History: $122 Billion at $852 Billion

Before it goes public, OpenAI completed what is already the largest private technology financing in history. The $122 billion round that closed on March 31, 2026 was led by Amazon, which committed $50 billion; SoftBank, which committed $30 billion; and Nvidia, which committed $30 billion — with additional participation from a consortium of strategic and financial investors. The round extended a prior $110 billion raise that had closed in February 2026.

The strategic significance of the round goes beyond the capital. Amazon's $50 billion commitment secures preferred access to OpenAI's models through Amazon Web Services, replicating the model Google and Amazon have both pursued with Anthropic. Nvidia's participation is as much a market endorsement as an investment — the world's dominant AI chip maker publicly betting on OpenAI as the dominant AI model provider. And SoftBank's position reflects CEO Masayoshi Son's conviction, stated publicly and repeatedly, that OpenAI represents the central infrastructure investment of the AI era. Son sits alongside Altman as a co-founder and co-funder of the Stargate infrastructure project.

6. Stargate: The $500 Billion Infrastructure Bet That Defines OpenAI's Future

On January 21, 2025, standing alongside President Donald Trump in the East Room of the White House, Sam Altman announced the Stargate Project: a joint venture between OpenAI, SoftBank, Oracle, and investment firm MGX to invest $500 billion in AI computing infrastructure across the United States by 2029. It is the largest announced private infrastructure investment in American history — exceeding the cost of the interstate highway system, adjusted for inflation.

Stargate's scale is staggering. Data center sites are already operational or under construction in Texas, New Mexico, Ohio, and multiple other states. By the time OpenAI filed its S-1, the project had already secured more than three gigawatts of computing capacity — meeting a target originally set for 2029, four years ahead of schedule. International sites in the UAE are expected to open in 2026, with the UK, Norway, South Korea, and Argentina in development. OpenAI is also developing its own custom AI chip — codenamed 'Titan' — in partnership with Broadcom, to be manufactured on TSMC's three-nanometer process, with mass production targeted for the second half of 2026. The strategic logic: whoever controls the compute infrastructure controls the AI economy.

7. Who Owns OpenAI: The Cap Table That Will Reshape Fortunes

The ownership structure of OpenAI is unlike any company that has previously entered the public markets. Microsoft, which has invested approximately $13 billion across multiple rounds beginning in 2019, holds approximately 26.79 percent of the company — a stake worth approximately $228 billion at the $852 billion valuation. That investment has produced a paper return of approximately 17.6 times. Microsoft's stake is the single largest individual shareholder position, slightly ahead of the OpenAI Foundation's 25.80 percent — a stake that the Foundation uses to exercise governance control.

SoftBank holds approximately 11.66 percent following its $41 billion cumulative investment, a position currently worth approximately $99 billion. Amazon's and Nvidia's stakes from the March 2026 round will be disclosed in the S-1 and are expected to be material. Early investors including venture capital firms and angels — among them Reid Hoffman, the co-founder of LinkedIn, and Peter Thiel, the co-founder of PayPal — are estimated to have achieved returns of approximately 140 times their initial investment.

The most remarkable entry in the cap table belongs to Sam Altman, the company's CEO: he owns no equity in OpenAI. His annual salary is $76,001. He does not hold shares, restricted stock units, or any other disclosed direct equity interest in the company he has run for seven years and built into the most valuable private technology company in the history of American enterprise. His pending equity grant — expected to be formalized as part of the PBC conversion and disclosed in the S-1 — will be among the most closely read sections of the prospectus.

8. Sam Altman's Equity Grant: The Most Anticipated Disclosure in the Prospectus

The question of Sam Altman's compensation has been one of the longest-running mysteries in Silicon Valley. For years, he has acknowledged publicly that he holds no direct equity in OpenAI, noting that he intended to donate the bulk of any eventual windfall to charity in any case. His compensation, at a $76,001 annual salary, is nominally the lowest of any CEO of a company approaching trillion-dollar valuation in recorded history.

The S-1 will be the first opportunity for public investors to read, in legally binding SEC disclosure language, the full terms of any equity package Altman receives. Bankers familiar with the process expect the prospectus to describe a new grant — structured as profit participation units or a conventional equity award under the PBC framework — that compensates Altman for his leadership role and aligns his financial interests with those of public shareholders. The magnitude of that grant, and the vesting conditions attached to it, will be among the first details analysts and media will extract from the public filing.

9. The Governance Structure: Foundation Control, PBC Structure, and Investor Rights

OpenAI is organized as a Public Benefit Corporation with a nonprofit foundation — the OpenAI Foundation — sitting atop the corporate structure and exercising control over the board of directors of OpenAI Group PBC. The Foundation holds approximately 26 percent of the PBC's equity and has the contractual right to appoint all directors and remove them at any time. The Foundation's board is currently chaired by Bret Taylor, the former co-CEO of Salesforce, and also includes Adam D'Angelo, the CEO of Quora; Dr. Sue Desmond-Hellmann, former CEO of the Bill and Melinda Gates Foundation; General Paul M. Nakasone, former director of the National Security Agency; Adebayo Ogunlesi; Nicole Seligman; and Dr. Zico Kolter. Sam Altman serves as the sole executive on the Foundation board.

For public investors, this structure creates governance realities with no precedent in the history of large-capitalization public markets. Common shareholders in OpenAI's IPO will hold equity in a company whose board is appointed by a nonprofit foundation — not by them, and not by any conventional shareholder vote. The conditions under which the Foundation could be modified, the circumstances under which common shareholders could assert control, and the legal standing of the PBC's public benefit mission relative to shareholder economic interests are all questions that the S-1 will be required to address — and that no securities lawyer has previously had to structure for a company of this size.

10. The Products: ChatGPT, GPT-5.5, o3, Sora — and the Custom Chip

OpenAI's product portfolio has expanded dramatically from the ChatGPT chatbot that launched the company into public consciousness. The flagship product today is GPT-5.5, the latest iteration of the generative pre-trained transformer model that was released on April 23, 2026. Alongside it sits the o3 reasoning model — a system capable of multi-step, chain-of-thought reasoning that achieves PhD-level performance on mathematics and coding benchmarks, and which represents OpenAI's answer to the 'thinking' problem that has historically limited large language model applications in high-stakes professional domains.

Sora, the video generation model released in December 2024, has attracted substantial attention for its ability to generate coherent, high-fidelity video from text prompts — a capability with significant implications for media, entertainment, advertising, and education. DALL-E, the image generation model, continues to serve a large base of individual and enterprise creative users. The API layer underlying all of these products supports the developer ecosystem that accounts for roughly 40 percent of OpenAI's revenue. And in development: the 'Titan' custom AI chip, being built with Broadcom and designed for inference workloads, is expected to reach mass production in the second half of 2026 — a critical step toward OpenAI's stated goal of reducing its dependence on Nvidia's hardware.

11. The Enterprise Business: 9 Million Business Users and Growing

OpenAI's enterprise segment — comprising ChatGPT for Teams, ChatGPT Enterprise, and API access for business applications — is the fastest-growing and most strategically important segment of its business. As of February 2026, the company served more than 9 million paying business users across tens of thousands of companies. Enterprise revenue now represents approximately 40 percent of total revenue and is on track to reach parity with consumer revenue by the end of 2026.

The enterprise product suite has evolved from a simple API into a comprehensive platform: custom model fine-tuning, private data integration, compliance and security controls, and increasingly, agentic capabilities — the ability for AI systems to autonomously execute multi-step tasks on behalf of enterprise users without requiring constant human supervision. This agentic pivot is the central theme of OpenAI's next phase of growth, and the key to the argument that the company's total addressable market expands from tens of billions to trillions as autonomous AI agents begin to replace or augment white-collar work.

12. The Microsoft Relationship: Partner, Investor, Customer, and Competitor

The relationship between OpenAI and Microsoft is the most complex and consequential in the history of American technology. Microsoft's $13 billion in cumulative investment — which began with a $1 billion check in 2019, expanded to $10 billion in January 2023, and has continued with additional follow-on commitments — gave the software giant a 26.79 percent stake worth approximately $228 billion at OpenAI's current valuation. That investment was structured to give Microsoft preferred access to OpenAI's models, which powered the company's Copilot products across Office 365, Azure, Bing, and GitHub.

The relationship was renegotiated in April 2026 as a condition of OpenAI's PBC restructuring. Under the revised terms, Microsoft surrendered its exclusive licensing rights — allowing OpenAI to deliver its models through any cloud provider — in exchange for an extended revenue-sharing arrangement and a perpetual license to OpenAI's technology through a specified date. The restructuring reduced Microsoft's governance influence while preserving its economic interest and its commercial relationship. The new arrangement also cleared a structural obstacle to OpenAI's IPO: public market investors were unlikely to purchase shares in a company whose most valuable product was exclusively licensed to a single strategic investor that simultaneously competed with OpenAI in the AI application market through Copilot.

13. The Financial Architecture: Consumer, API, and Enterprise Economics

OpenAI generates revenue through three principal channels. The first is consumer subscriptions: ChatGPT Plus at $20 per month, ChatGPT Team at $25 per user per month, and ChatGPT Pro at $200 per month for power users — a relatively recent addition targeting professionals who run intensive research, coding, and analysis workflows. The second channel is API access, through which developers and businesses access OpenAI's models programmatically and pay per token consumed. The third is the enterprise product suite, sold through direct contracts to large organizations.

The economics of these channels differ significantly. Consumer subscriptions generate predictable, recurring revenue but carry relatively high customer acquisition costs and some churn. API revenue scales with developer and startup ecosystem growth and is highly responsive to model capability improvements that expand use cases. Enterprise contracts are typically larger, stickier, and longer-term, but require substantial sales infrastructure and longer closing cycles. The disclosure of the precise revenue split, margin structure by channel, and customer cohort data in the S-1 will allow analysts, for the first time, to build detailed financial models of OpenAI's business.

14. The Compute Cost Problem — and the Path to Gross Margin Improvement

The economics of AI model deployment at scale are, in their current state, brutally challenging. The cost of inference — running AI models to generate responses to user queries — has been falling rapidly as hardware efficiency improves and as Nvidia's successive GPU generations deliver more performance per dollar. But the absolute scale of OpenAI's deployment means that even rapidly declining per-unit costs must be compared against rapidly growing volumes. At 900 million weekly users and 15 billion API tokens processed per minute, the raw compute bill is extraordinary.

Anthropic, OpenAI's closest competitor, has disclosed a trajectory from 38 percent gross margins at the inference level in 2023 to approximately 70 percent in early 2026, with projections of 77 percent by 2028. OpenAI's margins are believed to be on a comparable trajectory, though the precise figures will be disclosed for the first time in the S-1. The Titan custom chip initiative is central to the long-term margin story: proprietary inference hardware would eliminate Nvidia's approximately 70-80 percent gross margin on chips from OpenAI's cost structure, potentially transforming the financial profile of the business at scale.

15. The Legal Minefield: Copyright, Antitrust, and Musk's Lawsuit

OpenAI faces a constellation of legal challenges that will require extensive disclosure in the S-1 risk factors section. The largest in potential financial exposure is the consolidated copyright litigation, titled In re: OpenAI, Inc. Copyright Infringement Litigation, which combines 16 copyright lawsuits brought by publishers, authors, and media organizations — including the New York Times, which has alleged that its articles constitute one of the largest sources of training data used to build ChatGPT. In January 2026, a federal judge ordered OpenAI to produce 20 million ChatGPT conversation logs in connection with the Times litigation — a discovery ruling that dramatically expands the plaintiff's access to evidence. OpenAI argues that training on publicly available text constitutes fair use under copyright law; the outcomes of these cases remain uncertain, with potential adverse results including mandatory licensing fees, damages awards, or court-ordered restrictions on training practices.

The Department of Justice has opened an antitrust inquiry into the relationships between major technology companies and frontier AI laboratories, including the tripartite arrangement in which Microsoft is simultaneously OpenAI's largest shareholder, its primary cloud infrastructure provider, and a competitor in AI applications. EU AI Act compliance triggers mandatory safety testing, transparency requirements, and incident reporting for providers of general-purpose AI systems with systemic risk, with fines of up to three percent of global annual revenue for violations. And Elon Musk's lawsuit, though dismissed on statute of limitations grounds by the jury verdict of May 18, 2026, is now subject to an appeal that Musk has promised vigorously — and which introduces the risk of further litigation expense and strategic distraction.

16. The Elon Musk Dimension: The War That Started OpenAI

The story of OpenAI cannot be told without understanding Elon Musk — not because he still works there, but because the war between Musk and the company he helped found has become one of the defining narratives of the AI era. Musk was present at the creation, providing early funding and serving as co-chairman from 2015 to 2018. He left the board in 2018, citing the need to avoid conflicts of interest as Tesla developed its own AI capabilities for autonomous vehicles — though subsequent reporting has suggested the departure was accelerated by disagreements over control and strategic direction. In 2023, Musk founded xAI, a direct competitor to OpenAI, and began publicly attacking OpenAI's transition from nonprofit to for-profit, calling it a betrayal of its founding mission.

The lawsuit that followed — filed in 2024 and dismissed in 2026 — alleged that Altman had violated the original agreement under which Musk made his founding pledge. Altman testified at trial that Musk had sought control over OpenAI that the other founders were unwilling to grant, and that when Musk left, he 'left the nonprofit for dead.' The jury's dismissal of Musk's claims, on statute of limitations grounds rather than on the merits, has not resolved the underlying narrative conflict. Musk continues to operate xAI — now folded into SpaceX at a combined valuation of $1.25 trillion — as a direct competitor to OpenAI. The personal and competitive dimensions of that rivalry will be present throughout OpenAI's life as a public company.

17. The Talent Crisis: Nine Founders, Three Remain

OpenAI began with eleven founders and has, at various points, struggled to retain the people who built what it has become. The departures tell a story worth understanding. Andrej Karpathy, the pioneering deep learning researcher who co-founded the company and briefly returned after a stint at Tesla, has departed a second time. John Schulman, the co-founder responsible for pioneering Reinforcement Learning from Human Feedback — the technique that made ChatGPT usable by ordinary people — left in August 2024 to join Anthropic, citing a desire to focus more directly on AI alignment research. Ilya Sutskever, the chief scientist and perhaps the single most revered AI researcher at the company, resigned in May 2024 and founded Safe Superintelligence, which has raised $6 billion at a $32 billion valuation despite having no commercial product. Mira Murati, the company's highly regarded chief technology officer who had served as acting CEO during the five-day boardroom crisis, resigned in September 2024.

Of the eleven original founders, only three remain active in leadership roles: Sam Altman, Greg Brockman — who serves as president and returned from a brief sabbatical in late 2024 — and Wojciech Zaremba, a co-founder who continues to lead key research functions. The departure of so much founding talent is not, by itself, unusual in a company that has grown at OpenAI's pace — but the concentration of departures to directly competitive AI safety organizations raises questions that the S-1's human capital risk section will be required to address.

18. The Safety Question: AGI, Existential Risk, and the Disclosure No Prospectus Has Ever Had to Make

OpenAI's S-1 will contain a disclosure with no precedent in the history of securities regulation: a formal acknowledgment, in legally operative language reviewed by SEC staff and underwriting counsel, that the technology the company is building may pose risks to humanity at civilizational scale. This is not a theoretical exercise. OpenAI's own leadership has stated in public writing, in congressional testimony, and in company documentation that the development of artificial general intelligence — a system capable of performing any intellectual task that a human can — represents a potential existential risk if developed without adequate safety measures.

The company maintains a Safety and Security Committee, chaired by board member Zico Kolter, that is required by the terms of the California Attorney General's PBC conversion approval to maintain oversight of model development. OpenAI's internal alignment research program has produced published work on interpretability — understanding what is actually happening inside large neural networks — and on reinforcement learning from human feedback, which teaches models to be helpful and harmless. Former members of OpenAI's superalignment team have publicly stated that the company has not fulfilled commitments made about the resources and priority dedicated to safety research. The resolution of that tension — between the speed required to remain competitive and the caution required to remain responsible — is, in the company's own framing, the central challenge of its existence.

19. The Government Relationship: Stargate, National Security, and the AI Race with China

OpenAI's relationship with the United States government has evolved from distant to intimate with remarkable speed. The Stargate announcement — made alongside President Trump at the White House in January 2025 — cast the $500 billion infrastructure project explicitly in national security terms, with Trump saying, 'We want to keep it in this country; China's a competitor.' Subsequent discussions between the Trump administration and OpenAI have reportedly explored the possibility of a government equity stake in the company — a transaction that, if completed, would be without precedent for a startup of OpenAI's size and would raise profound questions about the independence of a company developing technology that the government itself acknowledges could be decisive in the competition with China.

OpenAI has separately proposed a 'Classified Stargate' initiative — dedicated, accredited data centers purpose-built for government AI applications, including classified defense and intelligence use cases. The company has also launched 'OpenAI for Countries,' a program to provide AI infrastructure and support to allied governments seeking to develop domestic AI capabilities. These government relationships are sources of both growth opportunity and political risk: a company whose commercial future is substantially dependent on favorable government treatment has governance complications that private-sector investors will weigh carefully.

20. The Competitive Reality: OpenAI Is Winning Consumer, Losing Enterprise

A dispassionate review of the competitive data in 2026 produces a nuanced picture that differs from both the triumphalist narrative OpenAI's promotional materials favor and the alarm that some competitor press releases imply. In the consumer market, OpenAI remains dominant: ChatGPT commands approximately 64.5 percent of chatbot web traffic globally, with Google's Gemini at 21.5 percent and Anthropic's Claude at a smaller but growing share. In the enterprise market, the picture is dramatically different. Ramp spending data and Similarweb analyses from early 2026 indicate that Anthropic wins approximately 70 percent of head-to-head enterprise deals against OpenAI among new business purchasers. In the AI coding market specifically — which is growing explosively as software development teams adopt AI tools — Claude commands approximately 42 percent market share versus OpenAI's 21 percent.

The competitive implications are significant. Enterprise software has historically been won on trust, integration depth, and customer success — not on the raw capability of any single product version. Anthropic has built a significant enterprise lead in a relatively short period. Google's structural advantage — billions of consumer devices, the world's most widely used productivity suite, and the most sophisticated cloud infrastructure outside of AWS — has not yet translated into AI leadership, but the gap is narrowing. And xAI's Grok, now integrated into the X platform with more than 500 million users, represents a distribution advantage that no conventional startup can replicate.

21. The Path to $1 Trillion: The Bull Case

The bull case for OpenAI at a trillion-dollar valuation begins with the growth trajectory. A company growing at 80x over three years, with 900 million weekly users, a $25 billion revenue run rate, and the most recognized AI brand in the world is, by the standards of traditional technology investing, a generational opportunity. The consumer moat — built on ChatGPT's first-mover advantage, brand recognition, and integration into hundreds of millions of users' daily workflows — is genuine and difficult to dislodge. The enterprise business is the fastest-growing segment and is converting at scale.

The deeper bull case is the agentic opportunity. OpenAI and every other frontier AI laboratory believe that the next phase of AI development involves autonomous agents — systems that can plan and execute complex, multi-step workflows with minimal human supervision. The TAM expansion that agentic AI implies is, in the view of OpenAI's management, orders of magnitude larger than the TAM for AI assistants. If an AI agent can autonomously complete tasks previously performed by a team of knowledge workers — legal research, financial analysis, software development, customer service, medical diagnosis — the addressable market is not tens of billions but tens of trillions. At $852 billion, the market is pricing in a meaningful probability of that outcome. Bulls argue it is pricing in too little.

22. The Bear Case: What Could Go Wrong

The bear case for OpenAI is substantial and coherent. The losses are real: $14 billion in 2026, no profitability projected until 2030, and a business where the cost of compute scales with the growth that bulls celebrate. The competitive environment is intensifying: Anthropic is better-capitalized than it has ever been, growing faster in enterprise, and led by a team whose AI research credentials are at least equal to OpenAI's. Google has structural distribution advantages that, if effectively deployed, could compress OpenAI's market share significantly. Meta has committed over $40 billion in AI capital expenditure in 2026 alone and is building open-source models that threaten to commoditize the API market.

The governance structure — Foundation control over the board, no conventional shareholder voting rights, the unresolved question of Altman's equity interest, and the regulatory scrutiny triggered by the PBC conversion — creates uncertainty that institutional investors accustomed to conventional corporate governance will need to price. And the existential risk disclosure — the company's own acknowledgment that the technology it is building could harm humanity — is a sentence that no prospectus has previously contained, and whose legal and reputational implications for a public company in periods of model failures or adverse AI incidents have no precedent.

23. The Full Financial Picture: What We Know Before the S-1

Revenue: Approximately $25 billion annualized run rate as of February 2026, up from $20 billion at year-end 2025, $6 billion at year-end 2024, and $2 billion at year-end 2023.

Cash Burn: Approximately $14 billion projected loss in 2026, representing approximately $1.17 billion per month. The company does not project profitability until approximately 2030.

Users: 900 million weekly active ChatGPT users, 50 million paid subscribers, 9 million business users.

Employee Compensation: Average stock-based compensation of $1.5 million per employee across approximately 4,000 employees — the highest of any technology startup in history. Median software engineer total compensation is approximately $555,000 per year.

Infrastructure: Stargate project committed to $500 billion in AI infrastructure investment through 2029, with more than three gigawatts of compute capacity already secured.

Investors: Microsoft (26.79%), OpenAI Foundation (25.80%), SoftBank (11.66%), Amazon, Nvidia, and others.

24. The Existential Stakes: Five Things the IPO Cannot Tell You

Behind the financial disclosures and the risk factors, OpenAI's IPO raises questions that no prospectus can fully answer. The first is whether artificial general intelligence — a system that can perform any intellectual task a human can — is years or decades away, and what OpenAI's probability of building it first actually is. The second is whether the safety research the company has invested in is adequate to the risk that its own leadership describes, or whether commercial pressure has systematically subordinated safety to speed. The third is whether the Foundation governance structure will function as designed when confronted with the quarterly earnings pressure, analyst coverage, and shareholder activism that come with being a public company. The fourth is what happens when the first genuinely harmful AI incident — a model used for fraud, manipulation, autonomous cyberattacks, or worse — becomes publicly attributable to OpenAI's systems. The fifth is whether the concentration of transformative AI capability in a handful of American technology companies, some of them receiving investment from foreign sovereign wealth funds, is itself a geopolitical risk that regulators have not yet fully grappled with.

These questions do not have answers in the S-1. They are, ultimately, the questions that will define the decades to come — for the company, for the industry, and for the civilization that is building and deploying this technology.

25. The Bottom Line: What Kind of Bet Is This?

To buy shares in OpenAI at whatever price it sets when it comes to market is to make a specific kind of bet — one that requires holding simultaneously in your mind the extraordinary reality of what the company has built and the extraordinary uncertainty of what it is building next. ChatGPT is the fastest-adopted consumer product in history. OpenAI's revenue growth has no precedent in enterprise software. The talent base it has assembled — even after the departures — includes some of the most important AI researchers working anywhere in the world. The mission, stated plainly since 2015, has not wavered: to ensure that artificial general intelligence benefits all of humanity.

And yet. The losses are real and large and not expected to narrow for years. The governance structure has no precedent in public market history. The competitive environment includes companies with structural advantages that OpenAI does not possess. The legal exposure is substantial and unresolved. The CEO, who has been building one of the most important companies in history for seven years, owns no equity in it — a fact that will require explanation and resolution before investors commit capital.

OpenAI is not SpaceX, which generates real cash from a dominant launch business. It is not Google, with decades of profitability and a trillion-dollar balance sheet. It is something genuinely new: a company organized around a mission that its own leadership describes as the most important in human history, growing at a pace that defies conventional modeling, losing money at a scale that requires either extraordinary conviction or extraordinary caution, and preparing to ask the world's public investors to join a project that may end in the greatest wealth creation in history — or in the most expensive science experiment ever underwritten by a stock exchange. The pricing will come in September. The stakes have been building for a decade.

 PART THREE: FIVE QUESTIONS THAT LOOM BEYOND THE IPO

Can OpenAI Achieve Profitability Before It Runs Out of Patience?

The financial profile of OpenAI today is, in its most reduced form, a race between revenue growth and cost growth. Revenue has grown from $28 million to $25 billion in three years. But compute costs, infrastructure investment, talent expense, and the capital demands of Stargate have grown faster than revenue throughout that period. The 2030 profitability projection assumes that model training costs continue to fall through hardware improvements, that inference efficiency gains outpace the growth in inference demand, and that the custom Titan chip reaches production scale on schedule. Any one of those assumptions could prove wrong. The question for IPO investors is whether they are comfortable funding a company through a four-year runway to profitability at whatever valuation OpenAI ultimately sets.

Will the Foundation Structure Survive the Realities of Public Company Life?

The OpenAI Foundation's control over the board of OpenAI Group PBC is designed to ensure that the company cannot be fully captured by profit motives at the expense of its safety mission. That design has never been tested by the pressures of quarterly earnings calls, activist shareholders, short sellers, or the kinds of governance crises that periodically afflict every large public company. Foundation governance structures are not unusual in nonprofit and hybrid organizations, but they have never been deployed over a company of this potential market capitalization. The first time a major shareholder attempts to assert control over a strategic decision — a merger, an acquisition, a product launch, a safety restriction — the real architecture of OpenAI's governance will be stress-tested. The S-1 will describe that architecture. The public markets will test it.

Is the Talent Exodus a Crisis or a Correction?

The departure of Ilya Sutskever, John Schulman, Mira Murati, Andrej Karpathy, and other founding leaders over an 18-month period has been described variously as a talent crisis and as a natural maturation of a company growing beyond its founding team. Both interpretations contain truth. The more important question is whether the concentration of AI alignment and safety expertise that departed — Sutskever and Schulman in particular were among the most important safety researchers in the world — reflects a systematic prioritization of commercial velocity over safety rigor. That is the allegation some former employees have made publicly. The S-1's human capital section will be read carefully for any acknowledgment of this risk.

What Happens When OpenAI Makes a Mistake at Scale?

Every technology company eventually has a major product failure. Apple had Antennagate, then Maps. Boeing had the 737 Max. Facebook had Cambridge Analytica. The difference between those incidents and the equivalent event for OpenAI is that OpenAI's product is an AI system capable of generating text, images, video, and code with near-human fluency — deployed to 900 million weekly users. A single high-profile safety failure at that scale — a model used for mass fraud, political manipulation, autonomous cyberattack, or the generation of content that causes documented harm — would not merely be a product recall or a public relations crisis. It would be a regulatory event, a litigation event, and potentially a market event with consequences that no conventional company has ever faced. The S-1 will acknowledge this risk. It cannot quantify it. The public markets will have to decide how much of a discount to assign to a future they cannot model.

Does This IPO Change the Course of the AI Race?

The proceeds of OpenAI's IPO — expected to be in the range of $40 to $60 billion — will flow into model development, compute infrastructure, talent acquisition, and international expansion. At that scale, the capital event is itself a competitive moat: it puts resources in OpenAI's hands that Anthropic, xAI, and every other competitor except Google and Microsoft will struggle to match. It also forces a discipline on OpenAI that private company status has, until now, allowed it to avoid: quarterly reporting, analyst coverage, public disclosure of financial performance, and the unrelenting attention of a market that will reward or punish every strategic decision. Whether that discipline accelerates or impedes OpenAI's mission is genuinely unknown. What is certain is that when OpenAI's shares begin trading — on whatever autumn morning the bell finally rings — the AI race will enter a new phase. The most important company in the world will be answerable to the world.