Microsoft takes 27% stake as OpenAI becomes public benefit corporation
Power to declare AGI shifts to independent, expert panel
Microsoft and OpenAI have agreed to a new deal that changes how they work together and share control over the future of AI.
OpenAI has now officially transitioned into a Public Benefit Corporation (PBC), now operating as OpenAI Group Public Benefit Corporation.
The restructured entity will be governed by the newly renamed OpenAI Foundation, ensuring the company's original nonprofit mission remains intact even as it opens itself further to investors and commercial expansion.
Under the new arrangement, Microsoft will hold a 27% stake in the company, valued at approximately $135 billion.
The revised partnership comes with updated ownership terms, extended IP rights through 2032, and newly formalised rules for how both companies will pursue artificial general intelligence (AGI).
In July, leaked documents revealed a controversial "AGI clause" buried deep in the firms' prior contract. The clause gave OpenAI's board exclusive authority to declare the arrival of AGI, a milestone that, once reached, could have cut Microsoft off from future model access and barred it from independently developing AGI using OpenAI's research.
While Microsoft CEO Satya Nadella publicly downplayed the risk at the time, industry watchers warned that the clause effectively left Microsoft dependent on OpenAI's definitions and decisions. That uncertainty is now over.
Under the new agreement, the power to declare AGI will shift from OpenAI's board to an independent panel of experts, jointly established by both companies. The panel will operate with autonomy and transparency, ensuring no single entity can unilaterally define when (or indeed if) AGI has been achieved.
"The nonprofit remains in control of the for-profit, and now has a direct path to major resources before AGI arrives," said Bret Taylor, OpenAI's chairman.
Rebalanced partnership
The updated framework was described jointly by both firms as "the next chapter" of their collaboration.
While Microsoft's influence remains significant, the Redmond-based company loses its right of first refusal on OpenAI's new technologies, a privilege that previously gave it early or exclusive access to key models.
In exchange, Microsoft retains deep integration across Azure, securing a $250 billion cloud contract that ensures its infrastructure remains central to OpenAI's operations; OpenAI now has the freedom to expand its computing footprint beyond Azure.
Microsoft's intellectual property rights now extend to include post-AGI models, but only under strict safety and governance guardrails.
The new deal allows OpenAI to work jointly with third parties on product development, release open-weight models that meet agreed-upon capability thresholds, and offer its API to US national security clients on any cloud platform.
The deal also loosens Microsoft's prior restrictions on AGI development, allowing the company to pursue AGI independently, either solo or with other partners.
However, this newfound autonomy comes with caveats: if Microsoft develops AGI using OpenAI's technology before AGI is officially recognised by the expert panel, it must operate within defined compute limits and adhere to ongoing revenue-sharing terms until the milestone is verified.
The restructuring of OpenAI's corporate model was closely scrutinised by regulators.
Both the Delaware and California Attorneys General confirmed this week they would not oppose the conversion, contingent on OpenAI's continued commitment to nonprofit oversight and ethical safeguards.