OpenAI just turned a long, messy year into a tidy new corporate map — and handed Microsoft a front-row seat for the company’s next act. The headline: OpenAI’s for-profit arm has been recapitalized into a public benefit corporation called OpenAI Group PBC, overseen by a newly minted OpenAI Foundation that holds an enormous equity stake and a stated mission to funnel some of that value back into public-interest work. At the same time, Microsoft and OpenAI rewrote their relationship — clarifying who keeps what if and when the industry ever hits the AGI finish line.

What changed

• The for-profit business is now OpenAI Group PBC; the nonprofit parent is the OpenAI Foundation, which will hold a very large equity stake in the for-profit entity. The foundation will start with a multi-billion dollar focus on healthcare, disease research and “AI resilience.”
• Microsoft’s effective ownership stake was re-expressed under the new structure: it now owns roughly 27% on an “as-converted diluted basis,” with that stake being valued at approximately $135 billion in the new arrangement. That’s down from roughly 32.5% in earlier accounting but still enormous.
• The deal clarifies the long-running thorn in their partnership: what happens after AGI — and how we even decide AGI happened. The answer now involves an independent expert panel to verify any AGI declaration, and a set of detailed IP and timing commitments between the two companies.

Why this was necessary (and why it took so long)

OpenAI’s attempt to move from a nonprofit origin story to a structure that could attract truly vast sums of capital has been debated, litigated and scrutinized for more than a year. State attorneys general in Delaware and California poked and prodded the conversion to ensure safety commitments and nonprofit oversight weren’t being disguised as legal theatre — and Elon Musk, who helped start OpenAI as a nonprofit, kept the pressure up with litigation seeking to block or unwind the change. Those negotiations and legal threads needed to be resolved before the recapitalization could be formalized.

The big, slightly awkward fingerprints in the new deal

Two points stand out as particularly consequential.

First, the AGI trigger: OpenAI won’t be the final arbiter of when AGI has arrived. That declaration must be verified by an independent expert panel, which is the single biggest procedural fix for the “who decides AGI?” problem that haunted earlier drafts of the deal. In practice, that adds an external check — but it doesn’t remove all political or strategic incentives from the scene.

Second, Microsoft’s rights were both extended and narrowed in carefully targeted ways. Microsoft now has IP rights to models and products through 2032, and those rights even extend to models built after AGI (subject to safety guardrails), but its access to underlying research and confidential methods will lapse earlier under some conditions. Crucially, Microsoft does not get rights to OpenAI’s consumer hardware — the much-rumored device that OpenAI is building with Jony Ive’s design shop is off the table for Microsoft. OpenAI also loosened the exclusivity of the partnership: it can now work with other third parties on some projects and release open-weight models, and Microsoft no longer has an unconditional right of first refusal to be OpenAI’s compute provider. At the same time, OpenAI committed to a massive incremental Azure purchase commitment. Those concessions balance continuity for Microsoft with breathing room for OpenAI to keep one foot in the open ecosystem.

Money, muscle and the AGI arms race

The economics of this deal matter as much as the legal language. Microsoft’s stake — the $135 billion figure that made business pages light up — gives it huge exposure to OpenAI’s upside while the clarified IP and verification mechanisms attempt to prevent a single partner from unilaterally owning an AGI win. Microsoft also apparently committed to an incremental purchase of Azure services measured in the hundreds of billions (reporting cites a $250 billion figure), which locks the cloud-compute center of gravity around Azure for the foreseeable future. That kind of long-term commitment is effectively infrastructure underwriting for whatever comes next.

At the same time, Microsoft is now expressly allowed to pursue AGI independently or with other partners — which turns the Microsoft-OpenAI relationship from a near-exclusive partnership into something more like a strategic anchor plus open runway for competition. Expect more arms-race headlines: the barrier to another deep AGI push is lower if a company like Microsoft can both buy access to OpenAI IP and run its own parallel initiatives.

The nonprofit’s promise (and skepticism)

OpenAI’s nonprofit foundation is being presented as a fail-safe: a body that will both hold a massive stake in the for-profit and direct billions toward public-interest priorities like health and “AI resilience.” The foundation’s initial focus — a $25 billion set-aside for health and resilience work, per OpenAI’s announcement — is meant to show that the recapitalization isn’t just a cash grab. But skeptics point out that governance and practical oversight still matter: a nonprofit that holds ownership on paper is only meaningful if it actively exercises oversight and resists shareholder pressure when safety conflicts with profit. Regulators and watchdogs have been explicit in probing how real that control will be.

What to watch next

1. The expert panel: who sits on it, how independent they actually are, and what metrics they use to declare AGI. That panel will shape the rules of the game.
2. How the nonprofit actually deploys capital: seeing the foundation move money into concrete research or disease programs will be the real test of its commitment.
3. Microsoft’s competitive moves: with clarified IP access but loosened exclusivity, Microsoft, cloud rivals, and other coalitions can leap into AGI bets. Expect partnerships, acquisitions, and big compute plays.
4. Litigation and regulatory aftershocks: Elon Musk’s legal challenge persists, and state regulators have not been passive. The legal story isn’t finished.

OpenAI’s recapitalization and the companion Microsoft agreement convert a contentious summer of legal scrutiny and corporate hand-wringing into a new operating reality: a for-profit PBC with a nonprofit owner, a clarified but still-complicated relationship with its biggest partner, and structural mechanics meant to govern the weirdest possible future — the arrival of AGI. The new map doesn’t resolve the philosophical or political questions about AI’s risks and rewards; it simply redraws who holds which keys and what happens when one of those keys opens the door. For now, everyone’s counting those keys — and the expert panel that will decide whether the door has actually opened.