A little-known clause in Microsoft's partnership agreement with OpenAI has become a key factor shaping their AI strategies and determining the future of artificial general intelligence (AGI). According to sources familiar with the contract, OpenAI has the sole right to determine when AGI is achieved — that is, "a highly autonomous system that outperforms humans at most economically valuable tasks," as defined in OpenAI's charter — and this decision effectively deprives Microsoft of access to new models, writes Wired.
This clause first came to light in an interview with Microsoft CEO Satya Nadella in May 2023, when he suggested that if the era of superintelligence arrived, "all agreements are canceled," which was perceived as a philosophical joke at the time. However, further investigation confirmed that the contract explicitly states that if OpenAI announced the achievement of AGI, Microsoft would lose access to the models and the agreement, which was supposed to last until 2030, would be canceled.
This clause has three parts: first, the OpenAI board determines whether AGI has been achieved, and second, it assesses whether AGI is “sufficient”—that is, capable of generating over $100 billion in revenue, enough to reward both Microsoft and OpenAI’s other investors. Microsoft must agree with this conclusion, but cannot unreasonably challenge it—in the event of a dispute, the final word is in the courts. The third part stipulates that Microsoft is not allowed to independently develop its own AGI during the term of the contract. In effect, OpenAI is free to move toward AGI, while Microsoft is limited to commercial products and cannot create its own advanced models.
The clause was included in the contract because the OpenAI team led by Sam Altman truly believed in the feasibility of AGI in the medium term, while Microsoft, led by Nadella, considered AGI a distant scenario and agreed to the terms because the risks seemed insignificant until 2030. However, in 2025, the timelines changed.
Altman has publicly stated that AGI could be here as early as this year, while a fierce global talent race is underway with companies like Meta, Anthropic, and Google DeepMind all vying for top AI researchers with sky-high salaries. For Microsoft, this is an increasingly worrisome scenario: if OpenAI announces AGI, Microsoft will be locked out of new models and forced to start building its own from scratch—despite the fact that its entire AI strategy is now tightly tied to OpenAI’s technologies. Microsoft is now integrating OpenAI’s models into its Copilot for Microsoft 365 and Azure OpenAI Service products, which are key to its commercial AI efforts.
This clause is currently under negotiation as OpenAI seeks to change its governance structure from a model where a nonprofit controls a for-profit division to a public benefit corporation. This would remove profit caps and allow investors and employees to directly own shares. Such a change would require Microsoft’s approval, giving it leverage to demand revisions or even the removal of this clause altogether.