For every dollar of revenue OpenAI generated in Q1 2026, the company lost an additional $1.22. That is a negative 122% non-GAAP operating margin, a figure that would disqualify most companies from serious IPO conversations.
OpenAI is not most companies. On May 22, it filed a confidential S-1 registration statement with the SEC, setting the stage for what could become the largest technology IPO in history. The target valuation sits between $852 billion and $1 trillion. Goldman Sachs and Morgan Stanley are co-leading the deal, with JPMorgan Chase also involved. The listing window is Q4 2026, possibly as early as September.
The filing arrives at a peculiar moment. OpenAI’s revenue is growing faster than almost any enterprise software company on record, while its losses are accelerating even faster. Its competitor Anthropic just projected its first quarterly operating profit. And the company’s own CFO has raised concerns about financial reporting readiness. The S-1, when it eventually becomes public, will translate that tension into the language of GAAP accounting for the first time.
$25 Billion in Revenue, $14 Billion in Losses
OpenAI’s top line is extraordinary by any measure. Full year 2025 revenue came in at $13.1 billion. By February 2026, the annualized run rate had reached $25 billion, a roughly 92% year-over-year increase. The company now generates approximately $2 billion per month.
The user base behind those numbers is massive: 900 million weekly active users, 50 million consumer subscribers, and 9 million business users. Enterprise contracts account for more than 40% of total revenue, a significant shift from the consumer-dominated model that defined ChatGPT’s first two years. The advertising pilot launched in April hit $100 million in annualized recurring revenue within six weeks, with a full year ad revenue target of $2.5 billion.
None of that growth has closed the gap with spending. The Q1 2026 quarterly loss came in at approximately $6.95 billion, putting the company on track for $14 billion in operating losses for the full year. That figure is worse than the $9 billion full year loss reported in 2025. Revenue is roughly doubling; losses are growing faster.
Profitability is not forecast until 2030 at the earliest. HSBC analysts estimate that OpenAI will need to close a $207 billion capital gap between now and then to remain viable on its current trajectory.
Where $600 Billion in Compute Obligations Go
The cost structure driving those losses is dominated by a single category: AI compute infrastructure. OpenAI has committed approximately $600 billion in semiconductor and data center obligations over the next five years, with some infrastructure partnerships extending to $1.4 trillion over eight years.
These are binding obligations, not aspirational capex plans. They include the $20 billion Cerebras partnership for inference hardware that reduces dependence on Nvidia, the Stargate data center project with SoftBank, and the restructured Microsoft relationship that ended Azure exclusivity in April 2026 but preserved Microsoft’s cloud infrastructure role.
The strategic logic is straightforward: if AI models continue improving with scale and enterprise demand keeps growing, this infrastructure creates a barrier to entry that no competitor can match. If demand growth slows, inference costs drop sharply (as GLM-5.1 demonstrated with zero Nvidia hardware), or a more compute-efficient architecture emerges, the fixed cost structure becomes a liability rather than a moat.
OpenAI’s CFO Sarah Friar has been direct about the risks. According to reporting from multiple outlets, Friar warned internally that the company is “behind on the financial reporting readiness” required for public markets and cautioned that “sluggish revenue could limit the data center buildout.” Those are not the words of a CFO choreographing a triumphant road show. They are the words of a finance executive trying to align ambition with execution.
A Governance Structure Public Markets Have Never Tested
The financial complexity is compounded by a corporate structure with no precedent among major technology IPOs.
OpenAI completed its conversion from a nonprofit-controlled capped-profit entity to a Public Benefit Corporation in October 2025. The original nonprofit, renamed the OpenAI Foundation, retained a 25.8% equity stake along with governance oversight authority. Microsoft holds 26.79% on a fully diluted basis, a position worth approximately $228 billion at the last private valuation, with a revenue share cap of $38 billion through 2030 to 2032.
The CEO equity situation adds another dimension of ambiguity. Sam Altman, who leads what could become the world’s most valuable technology company, has no confirmed equity stake in the restructured cap table. Whether that gets resolved before the roadshow, and what it signals about organizational alignment, will be one of the most closely watched disclosures in the filing. A CEO with zero ownership in a trillion-dollar public company would be without precedent in the technology sector.
Institutional investors will also weigh the November 2023 board crisis, when Altman was fired and reinstated within five days, against the new governance framework. The hybrid model of Foundation oversight combined with public company accountability is entirely untested. Any governance dispute that would previously have been managed privately now becomes a market event with immediate consequences for shareholders.
The February 2026 funding round raised $110 billion at an $852 billion post-money valuation from investors including SoftBank, Microsoft, and sovereign wealth funds. Those investors need a liquid exit. The IPO is not merely a fundraising event; it is the mechanism that either validates or begins to unwind the largest concentration of private capital in technology history.
The 20% Safety Pledge That Became 2%
One risk factor unlikely to appear in the formal S-1 but certain to follow OpenAI through its public life: the distance between its safety commitments and its actual execution.
In 2023, OpenAI publicly pledged to dedicate 20% of its compute capacity to superalignment research, the work of ensuring that increasingly powerful AI systems remain aligned with human values. According to internal reporting that surfaced this month, the company delivered between 1% and 2% of that commitment, on aging hardware, before dissolving the superalignment team entirely.
For a company asking public investors to trust that it can responsibly build artificial general intelligence while burning $14 billion a year, the gap between stated commitments and actual resource allocation is a material governance risk. Institutional investors, ESG-focused funds, and regulators in both the U.S. and EU will measure OpenAI’s safety track record against the expanding frontier AI testing regime, which now includes 40 government evaluations before models reach the public. The dissolved internal benchmarks are no longer the measuring stick.
Anthropic Turned Profitable the Same Quarter OpenAI Filed Its S-1
The timing creates a comparison that OpenAI’s underwriters would probably prefer to avoid. In the same window that OpenAI prepared its trillion-dollar IPO paperwork, Anthropic projected its first quarterly operating profit: $559 million on $10.9 billion in Q2 2026 revenue, representing 130% growth from Q1. Anthropic’s compute cost per dollar of revenue dropped from 71 cents in Q1 to 56 cents in Q2, a 15-cent shift in a single quarter that turned near break-even into a $559 million profit.
Anthropic achieved that result on revenue roughly half of OpenAI’s, with a smaller model lineup, and without an advertising business. The implication is uncomfortable for OpenAI’s investment thesis: a leaner competitor found unit economics that work at scale, while OpenAI’s margins moved in the opposite direction.
OpenAI’s counter will likely center on platform breadth. Between GPT-5.5, the Codex desktop agent, the Deployment Company’s $4 billion consulting arm, ChatGPT’s consumer base, and the advertising platform, OpenAI is building a diversified AI conglomerate. Anthropic is building a focused enterprise AI company. Both are legitimate strategies. Only one of them is asking public markets for a trillion-dollar valuation while losing money on every dollar of revenue.
What the Public S-1 Will Finally Reveal
The confidential filing means the full document will remain private until approximately 15 days before the roadshow, likely sometime in August 2026. When it drops, several questions will have answers for the first time: customer concentration (exactly how dependent is OpenAI on Microsoft as both investor and paying customer?), the detailed compute cost breakdown by product line, the revenue split between ChatGPT subscriptions, API usage, enterprise contracts, and advertising, and the specific terms of the SoftBank Stargate partnership.
For enterprise buyers evaluating AI vendors, a public OpenAI changes the information asymmetry entirely. Quarterly earnings calls will reveal churn rates, expansion revenue, and whether the enterprise business can sustain the infrastructure buildout without perpetual infusions of private capital.
For the broader AI industry, this IPO is a referendum on the central thesis of the current investment cycle: that frontier AI companies can generate returns proportional to the capital they consume. Google has committed $190 billion in annual AI capex. Anthropic just closed a $30 billion funding round at a $900 billion valuation. The total capital deployed into frontier AI infrastructure now exceeds the GDP of all but about 20 countries on earth.
OpenAI’s S-1 will be the first document to translate that spending into the language of public market accounting. The numbers will speak for themselves.