OpenAI Has Filed Confidentially for an IPO — and Almost Nothing About It Is Normal
OpenAI has submitted a confidential S-1 registration filing to the SEC, initiating a path toward a public market debut that would value the company at approximately $852 billion. If that valuation holds at listing, it would make OpenAI one of the largest IPOs in American corporate history — surpassing Saudi Aramco's 2019 record on a US exchange basis and dwarfing every prior technology listing.
But the headline number is arguably the least interesting part of the filing. What the S-1 reveals about OpenAI's finances, governance, and long-term structure raises questions that no IPO prospectus has ever had to answer before.
The Financials: $14 Billion in Losses on the Path to AGI
OpenAI is burning money at a scale that would be alarming for almost any other company. The S-1 reportedly discloses approximately $14 billion in cumulative operating losses, driven by the extraordinary cost of training frontier models, running inference at consumer scale, and building the safety and alignment research infrastructure that the company argues is essential to its mission.
Revenue has been growing rapidly — the company passed $10 billion in annualized revenue in early 2026 — but the gap between revenue growth and cost growth remains wide. Compute expenses alone, predominantly paid to Microsoft Azure under the existing partnership agreement, constitute the single largest line item in OpenAI's cost structure.
The S-1 is expected to frame these losses not as a sign of business model weakness, but as the necessary capital investment for building artificial general intelligence — a goal the company has argued since its founding is worth the cost regardless of short-term profitability. Whether public market investors accept that framing at an $852 billion entry price is the central question of the offering.
The Governance Structure: A Nonprofit That Controls a For-Profit Giant
The most structurally unprecedented aspect of the filing is OpenAI's proposed post-IPO governance. Unlike any publicly traded company in the Fortune 500, OpenAI is proposing a structure in which the OpenAI Nonprofit — the original entity founded in 2015 — retains ultimate control over the commercial arm, even after it becomes a publicly traded corporation.
The mechanics of this structure are still being worked out in regulatory discussions with the SEC and the California Attorney General's office, but the broad outline is as follows:
- The nonprofit board retains authority over the company's core mission, including the definition of what constitutes "safe" AI development and the power to override commercial decisions that conflict with mission alignment.
- Public shareholders would hold economic interests — dividends, capital appreciation, buyback participation — but would not control the fundamental strategic direction of the company.
- Sam Altman and the current management team would continue to operate the business, accountable to the nonprofit board rather than to a conventional shareholder-elected director structure.
This governance design is without precedent in public market history. The closest analogies — Berkshire Hathaway's dual-class structure, or Google's Class B supervoting shares — still ultimately deliver control to humans with economic skin in the game. OpenAI's structure delivers control to a nonprofit entity whose legal mandate is not shareholder return but human benefit.
Why This Structure Exists — and Why It Matters to Investors
OpenAI's unusual governance isn't an accident or a legal afterthought. It's a direct consequence of the company's founding premise: that building AGI is too consequential to be governed purely by profit motive. The nonprofit wrapper was designed to ensure that if — or when — OpenAI achieves a breakthrough of civilizational significance, the decision about how to deploy that breakthrough is not made by the shareholder with the largest position.
For institutional investors considering the offering, this creates a novel risk calculus. On one hand, the nonprofit structure provides a meaningful argument against the kind of short-term extractive pressure that has damaged other technology companies after IPO. Management cannot be voted out by an activist hedge fund that wants to monetize the model weights more aggressively than safety allows.
On the other hand, public investors would be buying into a company where their ability to influence strategy — through proxy voting, board seat campaigns, or SEC activism — is structurally limited. They are, in effect, buying a financial instrument attached to an entity whose fundamental decision-making sits outside the normal governance mechanisms that public market investors rely on.
The Microsoft Question
Any analysis of OpenAI's IPO must grapple with its relationship with Microsoft, which has invested approximately $13 billion in the company and holds a preferred revenue-share arrangement on OpenAI API traffic routed through Azure. The S-1 is expected to detail the terms of this relationship and any planned modifications as OpenAI transitions to a public company structure.
Microsoft's position is complex. It is simultaneously OpenAI's largest infrastructure partner, its largest single investor, and — through Copilot — a direct commercial beneficiary of OpenAI's models. The IPO process will require OpenAI to publicly disclose the financial terms of this relationship for the first time, which will reshape how the market values both companies.
The parallel situation at Anthropic — where Google's deep investment creates similar questions about independence — suggests that the investor community is becoming increasingly comfortable with frontier AI labs having large strategic investors, provided the governance is structured to prevent mission capture.
Valuation: Is $852 Billion Defensible?
At $852 billion, OpenAI would be priced at roughly 85 times its current annualized revenue — a multiple that makes even the most generously valued SaaS companies look cheap. The bull case rests on three arguments:
- Market size: If AI genuinely automates a significant fraction of knowledge work, the total addressable market is measured in the tens of trillions of dollars annually.
- Model leadership: OpenAI's GPT-5 and Codex product lines have demonstrated genuine enterprise traction at a scale no competitor has matched so far.
- Network effects: The more developers build on OpenAI's APIs, the harder it becomes to switch — creating switching costs that compound over time.
The bear case is equally straightforward: the model landscape is rapidly commoditizing, as evidenced by DeepSeek's permanent repricing of enterprise AI. If inference costs continue to fall and open-source models continue to close the quality gap with frontier proprietary models, OpenAI's pricing power — and thus its path to profitability — becomes structurally harder to defend.
What Happens Next
A confidential filing means OpenAI is not yet locked into a specific timeline. The company can amend the S-1, delay the offering, or withdraw entirely based on market conditions or ongoing regulatory negotiations. The typical window between confidential filing and public listing is six to twelve months, which would put a potential IPO in the late 2026 to mid-2027 window — assuming no major complications arise from the governance structure review.
What is certain is that the OpenAI IPO, when it happens, will be unlike anything public market investors have encountered before. It is simultaneously a bet on the most transformative technology in a generation and an exercise in accepting governance terms that would be unthinkable in any other context. The S-1 is just the beginning of a very long conversation.
