OpenAI consolidates controlling foundation over PBC, owner of 26% of the company

On October 28, 2025, OpenAI formalized the recapitalization that had been under negotiation for nearly a year with the attorneys general of California and Delaware. The result, described on the official Our Structure page, completely redesigns who holds power and who profits at the most valuable company in the AI sector. A year later, the design already serves as a reference model for how AI startups should think about governance and fundraising.

The change has two sides. The nonprofit entity founded in 2015 became the OpenAI Foundation. The for-profit subsidiary created in 2019 (which already existed and was already controlled by the nonprofit) is now OpenAI Group PBC, a public benefit corporation. The difference between a PBC and an ordinary corporation is not cosmetic: by law, it is required to advance its stated mission and consider the interests of stakeholders beyond shareholders, not just maximize returns.

What Changes in Governance in Practice

The Foundation is not a passive partner. It holds special voting and governance rights that allow it to appoint all members of the OpenAI Group board and replace them at any time. The Safety and Security Committee, responsible for overseeing security practices across the entire operation, remains subordinate to the Foundation, not the Group.

To avoid what OpenAI itself calls "fragmented governance," nearly all Foundation directors will also sit on the Group's board. The exception is Zico Kolter, who chairs the safety committee and will serve only on the Foundation, acting as a non-voting observer on the Group's board. Within a year, a second Foundation director is expected to follow the same path: leaving the Group's board and becoming a non-voting observer.

In short: the Foundation concentrates veto and appointment power, but tries to maintain a human bridge between the two structures so that the mission doesn't become a dead letter on one side and money doesn't become an isolated decision on the other.

The Numbers Behind the Design

OpenAI states that the Foundation holds a 26% stake in OpenAI Group, valued at approximately $130 billion based on the Group's current valuation. That percentage, applied to the stated figure, points to an implied valuation of about $500 billion for OpenAI Group as a whole (OpenAI does not disclose that total figure directly; it is a calculation derived from the ratio and the absolute value the company itself published).

Beyond the 26% stake, the Foundation holds a warrant: if OpenAI Group's share price rises more than tenfold within 15 years, it receives a significant additional volume of shares. This makes the Foundation, according to the company itself, the largest long-term beneficiary of OpenAI's success, larger even than the traditional financial investors on the cap table.

All equity holders in OpenAI Group now hold the same type of common stock, which participates proportionally in value growth. This eliminates a complexity that existed in the previous structure, in which the for-profit subsidiary's capital had more atypical profit-sharing arrangements relative to the nonprofit's control.

Why This Wasn't a Branding Move

It's tempting to read the recapitalization as a public-relations gesture: "the world's most powerful AI commits to purpose." But the history OpenAI itself describes contradicts that reading. The process took nearly a year of direct dialogue with the offices of the attorneys general of two American states, the kind of exchange that only happens when there is real regulatory scrutiny over whether a nonprofit is improperly transferring value to a commercial structure.

This matters for anyone in Brazil considering copying the model without understanding the original reason for it. The structure was born as a response to a specific legal question, raised by American regulators, about what happens when a nonprofit organization creates and then scales a billion-dollar commercial subsidiary. Brazilian startups that never had a prior nonprofit arm don't face that problem, so they don't need the same solution.

What This Means for Fundraising in Brazil

The angle that matters to founders and investors of AI-native companies in Brazil is not replicating OpenAI's exact architecture, which requires billion-dollar scale and decades of patient capital built into the design of the 15-year warrant. What's worth observing is the principle: major institutional investors, especially funds with impact mandates or regulatory exposure, will start asking how a company's governance handles the conflict between stated mission and financial return.

A simple PBC, without a controlling foundation, already signals that commitment in a cheaper and faster way to set up, and it's the path that makes sense for most AI startups in the country, especially since setting up a control structure through a Brazilian nonprofit entity (or a hybrid one, with a US entity) carries legal and compliance costs that few seed rounds can sustain.

Those raising Series B or C rounds around AI and safety theses, however, should expect investors to ask: who controls the board if the company grows too fast in the wrong direction? OpenAI's answer (a foundation with veto power and the status of largest long-term beneficiary) has become the benchmark against which other answers will be measured, even though most companies will opt for leaner versions.

The Counterpoint Missing from the Hype

The best critique of this model isn't about OpenAI, it's about imitation. Setting up a controlling foundation to signal purpose, without the regulatory history, without the capital to sustain a 15-year warrant, and without the scrutiny of attorneys general, can turn into corporate theater: an expensive layer of governance that solves no real problem for the company, only delays decisions and complicates the cap table for investors who prefer simplicity.

There is also a tension that OpenAI's own design doesn't resolve: the Foundation has the right to appoint and replace Group directors at any time, but nearly all Foundation directors also sit on the Group's board. In practice, the same person sits on both sides of the table in most cases. This tends to reduce day-to-day friction between mission and profit, which can be read both as governance efficiency and as a dilution of the very control the structure promises.

For those deciding the corporate structure of a Brazilian AI-native company today, the concrete lesson isn't to copy the foundation. It's to understand that sophisticated investors will demand a clear, documented answer with real legal teeth to the question OpenAI spent nearly a year answering to two American states: who's in charge when mission and money collide.