OpenAI Has Been a Public Benefit Corporation Since October 2025: What Changes for Those Who Depend on the API

On October 28, 2025, OpenAI completed a restructuring that had been under negotiation for nearly a year with the attorneys general of California and Delaware. According to the official announcement "Our Structure", the nonprofit founded in 2015 became the OpenAI Foundation, and the commercial arm created in 2019 became OpenAI Group PBC, a public benefit corporation. More than a year after the announcement, what has changed has already had time to show up in contracts: it's time to look at this with the mindset of someone paying the bill, not someone watching the keynote.

What Actually Changed Hands

Before, the nonprofit directly controlled a commercial arm with profit limited by contract (the so-called capped-profit model, which restricted investor returns to a multiple of invested capital). After the recapitalization, according to OpenAI, "all holders of shares in OpenAI Group now hold the same type of traditional common stock, which participates proportionally and grows in value with OpenAI Group's success": in other words, the profit cap for investors was removed from the design.

The Foundation swapped its direct control for a 26% stake in OpenAI Group's capital, valued by the company itself at approximately US$130 billion based on the Group's current valuation. Dividing that number by the stake, one can infer an implied valuation for OpenAI Group of around US$500 billion: OpenAI doesn't disclose that total directly, it's a rule-of-three calculation from the official figure.

Who's in Charge Now: Control Without Ownership

The Foundation isn't just another shareholder with 26%. According to the announcement, it retains "special voting and governance rights" that allow it to appoint all members of OpenAI Group's board and replace them at any time. In practice, money and corporate control were separated: whoever holds the largest equity stake (market investors) doesn't choose who sits on the board; whoever chooses the board (the Foundation) holds a minority stake in the capital, but gains a warrant granting it extra shares if the share value rises more than tenfold within 15 years.

This design creates a specific incentive: the Foundation, which is supposed to be the guardian of the safety mission, is now also the largest long-term beneficiary of the company's commercial growth. The Safety and Security Committee (SSC) remains tied to the Foundation, not to the Group, which at least formally preserves a governance body separate from direct commercial pressure.

is required to advance its stated mission and consider the broader interests of all stakeholders, ensuring the company's mission and commercial success advance together.

OpenAI, official announcement "Our Structure"

In American legal theory, a PBC is a middle ground between a traditional corporation and a nonprofit: the board CAN prioritize the mission over short-term profit, but is rarely OBLIGATED to in practice, because case law enforcing PBC duties is still thin (few shareholders sue a company for not fulfilling its mission enough). This means the "interest of all stakeholders" clause functions more as a reputational shield than as a deal-breaking constraint.

Why the End of the Profit Cap Matters to Those Paying for the API

The original capped-profit model was born as a safeguard: limiting how much investors could earn theoretically reduced the incentive to monetize the API aggressively at the expense of safety or the mission. With traditional common stock and no cap, pure commercial incentive once again drives pricing decisions, roadmap prioritization, and which customer segment gets attention first: an enterprise with a nine-figure contract tends to carry more weight than a Brazilian developer paying per token on the self-serve plan.

That doesn't necessarily mean higher prices tomorrow: OpenAI Group still competes with Anthropic, Google, and Meta, and competitive pressure holds prices in check as much as corporate structure does. But it changes the calculation for whoever decides internally: before, there was a formal cap limiting how much the company needed to grow to satisfy investors; now there's no cap, and the appetite for capital to train ever-larger models is public and growing.

What Changes in Single-Vendor Risk

For Brazilian CTOs and founders who treat the OpenAI API as critical infrastructure, the practical question is: does the new structure make the company more or less predictable as a long-term supplier? Three concrete points:

• Governance split between mission and capital. The Foundation controls the board, but now has a direct financial interest (via the 10x-in-15-years warrant) in commercial growth: the institutional brake that existed in the original design has gotten weaker.
• Uncapped need for capital. Without a return limit for investors, the pressure to raise multibillion-dollar rounds and sustain a valuation in the hundreds of billions increases dependence on API revenue growing fast, quarter after quarter.
• Brazilian presence on the board. David Vélez, founder of Nubank (a Brazilian digital bank), is among the independent directors of the OpenAI Foundation, a detail worth watching, since it's the only seat with direct visibility into the Brazilian market at a table that decides global product priorities.

In summary: the change doesn't create a brand-new risk, but it removes one of the few formal safeguards that set OpenAI apart from any other big tech company chasing revenue growth at any cost.

What to Negotiate in Enterprise Contracts From Now On

For those renewing an enterprise contract with OpenAI in 2026, three clauses deserve more attention than they did two years ago: fixed-period price protection (price lock), a data and prompt portability clause for switching to another provider, and a model-continuity SLA (a guarantee that a version won't be discontinued without a migration window). None of these is new in an infrastructure contract, but OpenAI has historically negotiated with less pressure on these points because it was, in fact, the only mature option: that has changed with Anthropic and Gemini competing at technical parity.

The Counterpoint: Structure Might Mean Stability, Not Threat

The strongest argument in OpenAI's favor is that the old design, with its profit cap, was becoming a real obstacle to raising the capital needed to compete in frontier model training: a company without the capital to train competitive models is a worse risk for the customer than a more commercial, but well-capitalized, company.

Under this reading, the end of the profit cap isn't OpenAI abandoning its mission: it's OpenAI recognizing that it can't fulfill the mission of ensuring AI benefits humanity without first winning the capital race against competitors who don't face this kind of constraint.

This reading is plausible, but it doesn't solve the concentration problem: it only shifts the bet from "OpenAI will fulfill the mission because it's contractually obligated to" to "OpenAI will fulfill the mission because the board, which it appoints itself, decides to." For those building a business on top of this API, the practical conclusion is the same regardless of which reading prevails: diversifying models and vendors has stopped being a theoretical precaution and become a risk-management item, with a set date for review, at every contract renewal.