OpenAI converts structure to PBC and redesigns control over its mission

OpenAI announced on October 28, 2025 the completion of a recapitalization that changes its corporate structure for the second time in a decade. According to the official announcement published at openai.com/our-structure, the nonprofit organization founded in 2015 is now called OpenAI Foundation, and the for-profit subsidiary created in 2019 (which already operated under the nonprofit's control) became OpenAI Group PBC: a Public Benefit Corporation, a corporate category that, unlike a conventional corporation, is legally required to pursue its stated mission and consider the interests of all stakeholders, not just shareholders.

For those who only follow the headline, this sounds like legal formality. For those who design startup capital structures, it's the opposite: this is the moment when OpenAI publicly decided how to separate mission control from investor capital appetite, without giving up either one. The official statement is sparing in its explanations of the reasoning, but the numbers it reveals say a lot about the scale of the legal engineering involved.

What changes in practice: two entities, one mission

OpenAI Foundation and OpenAI Group PBC share, according to the announcement, "the same mission" as stated. But the announcement is explicit about who is in charge: it is the Foundation that holds "special voting and governance rights" and it is the Foundation that appoints all members of the Group PBC's board, and can replace them at any time. In other words: the structure did not distribute power between the foundation and the operating company, it kept power concentrated in the foundation and created, alongside it, a corporate vehicle (the PBC) designed specifically to raise capital at scale and retain talent with real equity.

This is the first design lesson: separating the vehicle that attracts capital from the vehicle that holds final decision-making power is not the same as dividing power. It means concentrating governance in one structure and giving the other the commercial flexibility it needs to operate, including raising billions from investors who demand returns proportional to real shares, not to a capped-profit scheme.

The cap table: 26%, US$130 billion, and a warrant that only pays out in 15 years

The most concrete figure in the announcement is this: upon closing the recapitalization, OpenAI Foundation ended up with a 26% stake in OpenAI Group, valued at approximately US$130 billion based on the company's current valuation. In addition to that stake, the Foundation received a warrant (a right to buy more shares in the future) that only activates if the Group's share price multiplies more than tenfold after 15 years. If that happens, the Foundation gains a significant additional slice of equity.

The design here is deliberate: instead of fixing the foundation's power at a static percentage, OpenAI tied the Foundation's long-term value to the Group's commercial success. According to the announcement itself, this makes the Foundation "the largest long-term beneficiary" of OpenAI's success, which also means it has a direct economic incentive for the company to grow, not just a moral mandate to watch over it from outside.

For the founder negotiating a large round, the real trick lies exactly in this long-term warrant with a multiple-based trigger: it's a way to align whoever controls governance with whoever builds value, without handing operational governance to financial investors. That's different from giving a board seat to every fund that joins a round, a mechanism that's more common (and riskier for anyone who wants to keep control) in the cap tables of Brazilian startups that grow through multiple funding series.

The board: who sits where, and why it matters

The announcement lists the Foundation's board: Bret Taylor as chair, Adam D'Angelo, Paul Christiano, Sue Desmond-Hellmann, Zico Kolter, retired general Paul M. Nakasone, Adebayo Ogunlesi, Nicole Seligman, David Vélez (founder of Nubank, a Brazilian digital bank), and Robin Vince, in addition to CEO Sam Altman. The composition rule is clear: to avoid "fragmented governance," all Foundation directors also sit on the Group's board, with one deliberate exception.

Zico Kolter, who chairs the Safety and Security Committee (SSC), serves exclusively on the Foundation's board and acts only as a non-voting observer on the Group's board. The SSC remains a Foundation committee, not a Group one, and retains governance over safety practices across the entire operation. Within a year, a second Foundation director will move to this same model: an exclusive seat on the Foundation, a non-voting observer on the Group.

This detail is the most replicable part of the entire playbook for anyone building an AI startup's board: safety and mission compliance don't sit on the same board that approves budgets and M&A. That function goes into a committee with separate governance, with veto power or oversight, but without mixing into the day-to-day commercial decisions that require speed.

The counterpoint: what the PBC doesn't solve

The most generous version of this story is the one OpenAI itself tells: a design that preserves the strongest "mission-focused governance" in the sector while enabling unlimited capital. The more skeptical version is the one that prompted nearly a year of dialogue with the offices of the attorneys general of California and Delaware, mentioned in the announcement itself, proof that regulators wanted assurances that the conversion would not hollow out the nonprofit's original purpose in the name of attracting investors.

The structural blind spot is this: a PBC is required to "consider" stakeholder interests, but the law doesn't define a compliance metric, nor does it give third parties real enforcement power against the board for failing that duty. Whoever holds the mission accountable, in practice, remains the Foundation itself, which is controlled by the same CEO and the same directors who also sit on the Group's board. It is not a system of external checks, it is self-governance with a legal seal that raises the reputational cost of ignoring it, but doesn't technically prevent it from happening. For the investor negotiating a billion-dollar round in a scenario like this, this is the question worth asking before signing: what, in practice, stops the mission from losing to growth when the two come into conflict?

What this means for anyone raising a large round in Brazil

No Brazilian startup is going to replicate a foundation structure with a 15-year warrant, this is a design built for a scale of tens of billions of dollars and for regulators in two American states watching closely. But the principle behind it applies at a smaller scale: separating the vehicle that holds strategic control from the vehicle that raises operating capital is a cap table decision made early, not after Series C.

Those negotiating large rounds for generative AI in Brazil, with capital-intensive needs for GPUs and infrastructure, tend to give up operational governance in exchange for a bigger check, because they lack an upfront design of who sits on which committee and with what power. The OpenAI case shows the opposite: define in advance which body holds veto power over safety and mission, keep it separate from the commercial board, and tie the long-term value of whoever controls it to the business's success, not to the static position of a board seat. It is capital structure as a negotiating tool, not as a legal formality signed after the money has already landed in the account.