OpenAI converts to public benefit corporation, nonprofit keeps $100B stake and control

Summary

OpenAI's restructuring keeps the nonprofit in control with a $100B equity stake in the new Delaware public benefit corporation, while Microsoft and employees/investors each receive roughly 30%. OpenAI must complete the conversion by year-end or lose $19B in committed funding.

The nonprofit emerges as perhaps the world's best-funded research organization. It will hold equity valued at $100B, allowing it to operate as a pure research lab, pay competitive salaries matched against private-sector offers, and theoretically fund grants through secondary market sales. The structure sidesteps the immediate tension between nonprofit governance and for-profit scaling, but creates new ones.

Microsoft partnership

Microsoft and OpenAI signed a nonbinding MOU on partnership terms, breaking a months-long stalemate. Under the original 2023 deal, Microsoft receives 20% of OpenAI's revenue through 2030. A 20% tax on gross revenue off the top is unsustainable for most competitive software businesses. The fact that it held this long speaks to OpenAI's scale and Microsoft's leverage. The new deal structure lets both sides claim victory. Microsoft gets predictable cash flow during OpenAI's highest-margin years, while OpenAI preserves ownership upside beyond 2030.

The nonprofit's role

Whether the nonprofit and for-profit will drift apart over time remains unclear. With 30% of the cap table, the nonprofit operates more like a major shareholder than a traditional research grant-maker. If it accumulates researchers focused on pure science with no commercial pressure, it could become a talent magnet for researchers unbeholden to quarterly metrics. That is strategically valuable to OpenAI because it blocks competitors like Anthropic from claiming the nonprofit-research story and keeps top talent from drifting elsewhere. The friction point is that the nonprofit could theoretically launch new AI companies and convert them to for-profit again, repeating OpenAI's own path, or fund research that competes with the for-profit's priorities.

Structural tensions

California and Delaware attorneys general must approve the conversion. The three-way split among Microsoft, the nonprofit, and employees/investors masks structural oddness. A nonprofit holding $100B in appreciated stock and distributing it through salaries or grants is extreme and precedent-poor. The nonprofit could sell positions to pay researchers, creating a hybrid entity that is both research organization and shadow investment fund, with no clean historical parallel.

The original mistake

Almost everyone involved now agrees starting as a nonprofit was a mistake. Founders like Elon Musk, Peter Thiel, and Sam Altman may have been overthinking it in 2015, or they may have believed AI research couldn't command venture funding at scale. The counterfactual is murky. A traditional C corp might have faced pressure from investors to commercialize faster, and 2017 skepticism about preproduct-fit mega-rounds might have dogged it. The nonprofit framing gave OpenAI runway to do research without quarterly scrutiny. The team consensus now is that a simple Delaware C corp from day one would have been cleaner.

OpenAI proved product-market fit can overcome structural complexity. Even with a botched cap table and nonprofit governance, it shipped ChatGPT and scaled to multitrillion-dollar valuations. The restructuring closes the loop by taking the core insight—research plus scale—and wrapping it in conventional corporate form, while keeping the nonprofit as an endowment arm. Whether that endowment fulfills a research mission or becomes a shadow PE fund remains unwritten.