OpenAI Files for US IPO as AI Giants Race to Public Markets

On May 22, 2026, OpenAI confidentially filed its S-1 prospectus with the Securities and Exchange Commission, setting in motion what could become the largest technology IPO ever [1]. The company, which began as a nonprofit AI research lab in 2015, is now targeting a public market valuation above $1 trillion, with Goldman Sachs and Morgan Stanley leading the deal [2]. A September 2026 listing is the current target window.

The filing arrives amid an unprecedented wave of AI companies heading to public markets — Anthropic, Cerebras, and SpaceX are all preparing their own offerings — but OpenAI's case is uniquely fraught. It involves an unusual corporate structure, staggering losses, unresolved copyright litigation, and a central tension that no prospectus can fully resolve: whether the company that claims to be building the most consequential technology in human history can serve shareholders and humanity at the same time.

The Numbers: Revenue Explosion, Profit Nowhere in Sight

OpenAI's top-line growth is extraordinary. Revenue grew from $2 billion in 2023 to $6 billion in 2024, then roughly $13 billion in 2025. By March 2026, the company hit a $25 billion annualized run rate, generating approximately $2 billion per month [3]. ChatGPT now has over 900 million weekly active users and more than 50 million paying subscribers, with another 9 million business seats [2].

Source: OpenAI Financial Disclosures / Fortune

Data as of Mar 31, 2026CSV

But the cost structure tells a different story. In Q1 2026, OpenAI reported a negative 122% non-GAAP operating margin — meaning for every dollar of revenue, the company spent $2.22 [2]. In 2025, OpenAI spent roughly $22 billion against $13 billion in revenue, producing a $9 billion net loss [4]. Internal projections show the company expects a $14 billion loss in 2026, with cumulative losses from 2023 through 2028 reaching an estimated $44 billion before cash flow turns positive around 2029 or 2030 [5].

Source: Fortune / OpenAI Financials

Data as of Nov 12, 2025CSV

At the targeted $852 billion to $1 trillion valuation, OpenAI would be priced at roughly 34-40x its 2026 annualized revenue — a multiple that exceeds the peak of the 2021 SaaS IPO boom, when high-growth cloud companies typically debuted at 20-30x forward revenue. OpenAI's bulls argue the comparison is apples-to-oranges: no SaaS company has ever grown this fast at this scale. Its bears note that no SaaS company has ever burned this much cash, either.

Source: CNBC / TechCrunch

Data as of Apr 1, 2026CSV

The Corporate Structure: A Nonprofit Shell Around a Profit Machine

OpenAI's path from nonprofit to public benefit corporation is one of the most unusual corporate journeys in Silicon Valley history. In 2019, it created a "capped-profit" subsidiary that limited investor returns to 100x their investment [6]. By October 2025, under pressure to attract larger capital commitments and prepare for a public offering, OpenAI completed a full restructuring: the for-profit arm became OpenAI Group PBC (a Delaware public benefit corporation), while the original nonprofit was renamed the OpenAI Foundation [7].

Under the current structure, the OpenAI Foundation holds a 26% stake in OpenAI Group PBC, worth roughly $130 billion at the March 2026 valuation. Microsoft holds approximately 27% on a fully diluted basis. The remaining 47% is split among employees and other investors [7]. All board members of the PBC are appointed — and can be removed — by the Foundation [8].

One striking detail: CEO Sam Altman holds zero direct equity in OpenAI, according to May 2026 court filings [9]. Neither Altman nor the company has disclosed the mechanism, if any, through which he would receive equity tied to the IPO. Separately, the Trump administration has discussed a possible government stake in OpenAI, though no concrete terms have been reported [10].

As a public benefit corporation, OpenAI Group is legally required to balance shareholder returns against its stated mission. But critics argue this is weaker than it sounds. "Delaware law is clear that the directors of a PBC owe no duty to the public at all," one governance analysis noted, arguing that the PBC framework gives management broad discretion to define what "public benefit" means in practice [11].

The Investor Stack: Who Got In, and at What Price

OpenAI's cap table reads like a who's-who of global capital. The company closed a record $122 billion funding round on March 31, 2026 — the largest private fundraise in history — at a post-money valuation of $852 billion [12]. The lead commitments came from Amazon ($50 billion, though $35 billion is contingent on an IPO or the achievement of artificial general intelligence), Nvidia ($30 billion), and SoftBank ($30 billion). Other participants included Andreessen Horowitz, Abu Dhabi's MGX sovereign wealth fund, D.E. Shaw Ventures, TPG, and T. Rowe Price [13].

The valuation ramp has been steep. OpenAI was valued at $29 billion in January 2023, $86 billion in October 2023, $157 billion in an October 2024 tender offer, $300 billion by December 2025, and $852 billion by March 2026 [14]. Early investors who entered at the $29 billion mark are sitting on nearly 30x paper returns.

For retail investors considering the IPO, the structure of earlier rounds matters. Amazon's contingent commitment means a significant slug of capital only converts under specific conditions. The standard lockup period for insiders after an IPO is 90 to 180 days, but specific terms have not been disclosed in the confidential filing [15]. Liquidation preferences from prior rounds — where certain investors are entitled to get their money back before common shareholders see returns — could dilute the upside for ordinary public shareholders, particularly if growth slows or the company needs additional capital raises.

The Microsoft Question: Dependency, Decoupling, and Revenue Concentration

For years, Microsoft was OpenAI's exclusive cloud provider, its largest investor, and its primary distribution partner. That relationship was restructured in April 2026, in what amounted to a pre-IPO disentanglement [16].

Under the new terms, Microsoft lost its exclusive license to OpenAI's intellectual property. OpenAI can now serve its products across any cloud provider, including Amazon Web Services and Google Cloud — a key factor in the $50 billion Amazon commitment [17]. Microsoft continues to receive a 20% revenue share from OpenAI, but those payments are now capped at $38 billion through 2030. In exchange, Microsoft stopped paying a revenue share to OpenAI [18].

Azure remains OpenAI's primary cloud provider, and OpenAI products still ship first on Azure unless Microsoft decides otherwise [17]. The question for IPO investors is how much of OpenAI's revenue still depends on the Microsoft distribution channel — through Azure OpenAI Service, Copilot integrations, and enterprise licensing — and what happens if that relationship further loosens. The S-1, when made public, will likely need to disclose customer concentration data, which analysts expect to show that Microsoft-linked revenue represents a significant share of OpenAI's total.

The Competitive Moat: Eroding Faster Than the Valuation Suggests

The most difficult question the IPO raises is what, exactly, investors are paying a trillion dollars for. On standard AI benchmarks, the gap between OpenAI's proprietary models and open-weight alternatives has narrowed sharply. On MMLU — a widely used knowledge benchmark — the gap between open and closed models collapsed from 17.5 percentage points to 0.3 points in a single year [19].

DeepSeek's R1 model, released in January 2025, matched or exceeded OpenAI's o1 reasoning model on key math benchmarks while being trained at a fraction of the cost — an estimated $5.5 million for DeepSeek V3 versus the hundreds of millions OpenAI reportedly spends per training run [20]. By early 2026, Meta's Llama 4 open-weight models rivaled proprietary offerings on many tasks [21].

Independent ML researchers are divided on OpenAI's defensibility. The consensus view is that raw model intelligence is no longer a durable moat. OpenAI's remaining advantages are concentrated in areas like multimodal capabilities (image, video, and audio reasoning), safety fine-tuning, enterprise SLAs, and the distribution advantages of the ChatGPT brand [19]. Whether those advantages justify a valuation premium over companies like Meta — which gives its models away for free — is a question the market will have to answer.

The Employee Equation: $1.5 Million Per Head, and a Retention Cliff

OpenAI's average stock-based compensation hit $1.5 million per employee across its roughly 4,000-person workforce in 2025 — seven times the average equity compensation Google offered ahead of its 2004 IPO [22]. The company has eliminated the standard one-year vesting cliff for new hires and is offering retention bonuses of up to $1.5 million for senior employees, with most technical staff receiving at least $300,000 in bonuses vesting over two years [23].

In early 2026, OpenAI facilitated a $6.6 billion secondary tender offer that let more than 600 current and former employees sell vested shares to outside investors at an implied $400 billion valuation [24]. These tenders function as a retention tool — giving employees partial liquidity without requiring a full exit.

The retention risk materializes after the IPO lockup expires. Once shares become freely tradeable, early employees who hold equity worth tens of millions of dollars will face a choice: stay and keep building, or leave and diversify. This post-lockup cliff has historically led to significant departures at companies like Facebook and Uber. For OpenAI, where a small number of core researchers disproportionately drive model improvements, even a handful of departures could meaningfully affect the company's competitive position.

The Safety Argument Against Going Public

The strongest case against OpenAI's IPO is not financial — it is structural. Several former executives have quit the company citing concerns about safety culture eroding under commercial pressure. "Over the past years, safety culture and processes have taken a back seat to shiny products," one departing executive stated [25].

The concern is that public market pressures will accelerate this trend. Public companies face quarterly earnings expectations, activist investor campaigns, and competitive pressure to ship products quickly. These forces are difficult to reconcile with the deliberate, sometimes slow pace of safety research — testing models for dangerous capabilities, red-teaming outputs, and occasionally choosing not to release a model that performs well but carries unresolved risks.

"If you think AI companies are unethical now, wait until they go public," argued a Futurism analysis, noting that fiduciary duties to shareholders could "further undermine any lingering commitments to the ethical development of AI" [26]. OpenAI's own executives have acknowledged that their technology carries risks including enabling bioweapons development, coordinated cyberattacks, and — in the extreme case — existential threats.

Defenders of the IPO counter that public markets bring transparency. The S-1 filing and subsequent quarterly disclosures will force OpenAI to reveal financial details, risk factors, and governance structures that have been hidden behind private company opacity. A public benefit corporation structure, they argue, gives the board legal cover to prioritize safety even when it conflicts with short-term profitability [8].

Regulatory and Legal Exposure

OpenAI enters public markets carrying a substantial regulatory and litigation portfolio. The Federal Trade Commission has been investigating the company since July 2023 over potential violations of consumer protection law, specifically around data collection and privacy practices [27]. Italy's data protection authority fined OpenAI €15 million for GDPR violations — the first regulatory penalty against a generative AI company in Europe [28].

On the litigation front, the company faces consolidated multi-district copyright litigation in the United States. Twelve cases were combined in April 2025, including The New York Times' high-profile suit alleging that OpenAI used millions of articles to train its models without authorization [29]. In January 2026, a federal judge ordered OpenAI to produce 20 million ChatGPT conversation logs to plaintiffs — a discovery ruling with significant implications for the company's legal exposure [30]. As of mid-2026, three judges have ruled on the fair use question central to these cases: two in favor of AI companies, one against. No further summary judgment decisions are expected before summer 2026 [29].

The EU AI Act, which imposes compliance obligations on providers of general-purpose AI models, adds another layer of cost. While specific dollar estimates have not been publicly disclosed, underwriters will likely flag these contingent liabilities in the S-1's risk factors section. The combination of active FTC scrutiny, unresolved copyright litigation, and evolving European regulation creates a legal overhang that could persist for years after the IPO.

What the S-1 Must Answer

When OpenAI's S-1 becomes public — likely in the weeks before the listing — it will be one of the most closely read corporate filings in history. Investors, regulators, competitors, and AI safety researchers will all be looking for answers to questions that OpenAI has largely been able to avoid while private.

Among them: How much revenue comes from Microsoft versus independent customers? What are the specific terms of Amazon's contingent $35 billion commitment? How does the PBC structure interact with the Foundation's board appointment power in practice? What dollar range have underwriters assigned to the copyright litigation and regulatory investigations? And, perhaps most fundamentally: what is the path to profitability for a company that currently loses $1.22 for every dollar it earns?

The answers will determine not just OpenAI's stock price, but the market's broader appetite for AI companies that promise transformative technology while delivering transformative losses. The AI industry's public market era has begun. What investors are willing to pay for a piece of it — and what they're willing to overlook — will shape the field for years to come.