OpenAI IPO: S-1, PBC Structure and Valuation

Last updated: June 2, 2026 – Based on OpenAI’s official PBC structure, reported IPO-preparation coverage, SEC S-1 rules, and OpenAI’s GPT-5.5/Stargate infrastructure disclosures.

What is confirmed about the OpenAI IPO?

The confirmed answer is narrower than the market chatter: there is no public OpenAI S-1 and no official OpenAI IPO date as of June 2, 2026. OpenAI has not published a Rule 135 notice saying it confidentially submitted a draft registration statement to the SEC. It has not disclosed a ticker, exchange, share count, price range, underwriters, or offering date.

What is confirmed is the corporate setup that makes an IPO possible. OpenAI completed a recapitalization announced on October 28, 2025. The nonprofit is now the OpenAI Foundation. The for-profit is OpenAI Group PBC, a public benefit corporation. The OpenAI Foundation continues to control OpenAI Group through special voting and governance rights, holds a 26% equity stake valued at approximately $130 billion at the time of the recapitalization, and can appoint all members of the OpenAI Group board. Microsoft holds roughly 27%, while current and former employees and investors hold the remaining 47%.

That structure is IPO-relevant because public markets need ordinary equity, board rights, disclosure accountability, and a clear entity whose shares can list. OpenAI’s old capped-profit/nonprofit-controlled structure was difficult to map onto a conventional IPO. OpenAI Group PBC is still unusual, but it is much closer to a public-market-ready issuer.

What is being reported?

Axios reported on May 20, 2026 that OpenAI was working on a confidential IPO prospectus that could be filed shortly, although timing remained fluid. Axios also reported that OpenAI and SpaceX were working with overlapping banks, including Goldman Sachs, Morgan Stanley and JPMorgan Chase, and that OpenAI and Anthropic were widely expected to pursue public listings later in 2026 if markets allowed.

That is credible reporting, not the same thing as an official filing. The distinction matters. A confidential draft S-1 can exist without being visible on EDGAR, but if the company chooses to publicly announce it, the announcement usually contains narrow Securities Act language. Anthropic did that on June 1, 2026. OpenAI had not done that by the time this article was updated.

Why an OpenAI IPO would be different

An OpenAI IPO would not be a normal software listing. It would combine at least six stories into one public-market event: ChatGPT subscriptions, enterprise software, developer API usage, Codex and agentic work, model-layer research, and an infrastructure buildout measured in hundreds of billions of dollars.

The hardest public-market question is not whether OpenAI is important. It is whether OpenAI can convert importance into durable economics. Strong models are expensive to train and expensive to serve. Enterprise adoption may be huge, but public investors will ask whether gross margins improve as usage scales or whether every capability jump requires another wave of compute spending.

The second unusual factor is governance. OpenAI Group PBC is controlled by the OpenAI Foundation. That may be attractive for mission continuity and safety oversight, but public investors will want to know how voting rights, board rights, mission obligations, Microsoft rights, employee equity, and Foundation control interact when commercial incentives and public benefit obligations conflict.

The valuation question

Private-market reporting around OpenAI has moved very quickly. Axios reported in February 2026 that OpenAI secured $110 billion in new funding from Amazon, Nvidia and SoftBank, giving OpenAI an $840 billion valuation including the capital raised. Bloomberg separately reported a $110 billion financing at a $730 billion value. Those are private financing values, not an IPO price.

An IPO valuation would depend on the public S-1: audited or reviewed financials, growth rate, gross margin, loss profile, compute obligations, customer concentration, related-party economics, and demand from institutional investors. The same headline number can look cheap or expensive depending on whether the company has software-like margins, cloud-like capex intensity, or something more volatile than either.

The Microsoft problem: partner, owner, supplier, competitor

Microsoft is the most important relationship the OpenAI S-1 would need to explain. OpenAI’s official structure page says Microsoft holds roughly 27% of OpenAI Group. Microsoft has also been OpenAI’s strategic cloud partner, distribution partner and product channel through Azure OpenAI and Microsoft Copilot.

That creates four disclosure questions:

• Economics: How much revenue flows through Microsoft-related channels, and how is it shared?
• Compute: What are the minimum cloud commitments, capacity rights and termination provisions?
• IP: What model, product or technology rights does Microsoft retain after the recapitalization?
• Competition: When does Microsoft become a distributor, customer, infrastructure supplier or rival?

For public investors, this is not just a governance footnote. It goes directly to revenue quality, concentration risk, gross margin, and strategic control. If OpenAI is valued like the core AI platform layer, the S-1 has to show that it is not merely a high-growth application business sitting inside another company’s cloud economics.

Stargate and the compute wall

OpenAI’s infrastructure story is centered on Stargate. OpenAI announced on January 21, 2025 that Stargate would invest $500 billion over four years in new AI infrastructure for OpenAI in the United States, with SoftBank, OpenAI, Oracle and MGX as initial equity funders. SoftBank and OpenAI were described as lead partners, with SoftBank holding financial responsibility and OpenAI holding operational responsibility.

Any OpenAI IPO prospectus must explain how Stargate affects the business. Is Stargate a supplier, joint venture, financing vehicle, related party, capacity guarantee, or strategic infrastructure partner? What are OpenAI’s obligations? Who owns the assets? Who bears construction risk? How much capacity is contracted versus optional? How does Stargate interact with Microsoft Azure, Oracle Cloud, Nvidia systems, Amazon participation and future data-center sites?

This is why an OpenAI S-1 could reshape the AI market. It would turn the “AI infrastructure boom” from a story about press releases into a balance-sheet and footnote problem.

GPT-5.5 and the product story

The product case for an OpenAI IPO is stronger after GPT-5.5. OpenAI released GPT-5.5 on April 23, 2026 and positioned it as a model for real computer work: agentic coding, online research, data analysis, document and spreadsheet creation, software operation, and multi-step tool use. OpenAI reported strong scores on Terminal-Bench 2.0, GDPval, OSWorld-Verified, BrowseComp and FrontierMath, along with real internal use of Codex across engineering, finance, communications, marketing, data science and product management.

For IPO purposes, the important part is not the benchmark table alone. It is the direction of the business. OpenAI is trying to move from “chatbot subscription” to “AI labor layer”: Codex for engineering, GPT-5.5 for knowledge work, ChatGPT Enterprise for organizations, API usage for developers, and agents that perform work across tools.

That supports a large revenue story. It also raises the risk story: autonomous agents can fail, misuse can scale, cybersecurity capability cuts both ways, and regulators will expect stronger controls as models become more capable.

What an OpenAI S-1 should reveal

If OpenAI publicly files an S-1, the market should read it less like a normal SaaS IPO and more like the first full prospectus for a frontier-model operating system. These are the sections that matter most.

OpenAI vs Anthropic: why the race matters

Anthropic has already announced a confidential draft S-1. That gives Anthropic pole position in the public IPO narrative, at least on official timing. OpenAI still has the larger consumer brand through ChatGPT, a deeper Microsoft relationship, and arguably the broadest developer and enterprise surface. But if Anthropic’s S-1 becomes public first, it may define the initial investor framework for frontier AI economics.

The first public frontier-lab S-1 will set the vocabulary: cost per unit of intelligence, inference margin, training capex, model-risk disclosure, cloud concentration, enterprise adoption, safety governance and partner economics. OpenAI’s eventual S-1 would then be judged against that framework. That is why the timing race matters even before shares trade.

Regulation and safety disclosure

OpenAI’s S-1 risk section would likely be one of the most scrutinized technology risk sections ever published. GPT-5.5 already came with expanded safeguards for cybersecurity and other high-risk domains. The EU AI Act’s GPAI obligations and broader implementation dates also make model providers disclose and manage systemic-risk categories more formally than they did in the GPT-4 era.

Likely risk-factor categories include:

• Frontier model misuse: cyber abuse, automated fraud, misinformation, biological or chemical risk, and agentic tool misuse.
• Regulatory compliance: EU AI Act, U.S. state AI laws, sector rules, privacy laws, export controls and government procurement requirements.
• Intellectual property: training-data litigation, output disputes, licensing costs and indemnity obligations.
• Safety governance: how Foundation control, Safety and Security Committee oversight and commercial pressure coexist.
• Reliability: hallucinations, bad tool calls, autonomous-agent errors and enterprise liability.
• Compute supply: chip availability, cloud capacity, power, data-center construction and model-serving efficiency.

The likely IPO timeline

If OpenAI confidentially files or has already filed without public notice, the public does not see the actual draft until the company files the public S-1. SEC guidance says an issuer conducting an IPO must publicly file the registration statement, the first nonpublic draft and all draft amendments at least 15 days before it conducts a roadshow, or at least 15 days before effectiveness if there is no roadshow.

The best near-term watch item is not a rumored valuation. It is a public OpenAI S-1 on EDGAR or an official OpenAI notice that a confidential draft has been submitted. Everything else is preparation until the disclosure documents appear.