Microsoft and OpenAI built one of the most important partnerships in artificial intelligence. Now, according to reporting cited by The Decoder, the relationship is under strain because the companies are fighting over what happens if OpenAI reaches artificial general intelligence, or AGI.
The dispute is not only technical. It reaches into OpenAI's restructuring plans, Microsoft's investment rights, revenue sharing, intellectual property access and the future of both companies' AI strategies.
The AGI clause has become the core dispute
At the center of the conflict is a contract clause from 2019. Under that term, OpenAI can cut off Microsoft's access to its technology if OpenAI's board decides the company has achieved AGI. The source states that OpenAI defines AGI as systems that outperform humans at most economically valuable work.
That clause was originally designed to prevent technology with world-changing potential from ending up only under the control of one profit-driven company. According to The Information, it has now become a major point of conflict between OpenAI and Microsoft.
OpenAI has not agreed to remove the clause. Microsoft, meanwhile, wants it gone. The company is reportedly withholding approval for OpenAI's restructuring plans and using its veto power as leverage.
The stakes are high because OpenAI is seeking to restructure its for-profit arm as it prepares for a potential IPO. If Microsoft stalls the process, the report says that could derail OpenAI's IPO ambitions while preserving Microsoft's rights to revenue and future profits.
Microsoft reportedly faces its own AGI restriction
The Wall Street Journal reported another contract detail that raises the pressure on Microsoft: the company is barred from developing its own AGI until 2030, according to sources familiar with the matter.
That reported restriction matters because Microsoft is trying to reduce its dependence on OpenAI. The partnership has produced high-profile products such as GitHub Copilot, but the relationship has also moved toward direct competition.
The Decoder's source article describes several sources of tension:
- ChatGPT's unexpected success overshadowed Microsoft's own AI launches.
- By March 2023, Microsoft and OpenAI were competing for the same enterprise customers.
- OpenAI turned to Oracle and Google for extra server capacity through cloud deals outside of Microsoft.
- Microsoft began building cheaper alternatives, including the Phi model family, after OpenAI's more efficient Arrakis model failed.
Arrakis was expected to help Microsoft save money when running OpenAI models in products such as its Bing chatbot. When that model failed, Microsoft had a stronger reason to pursue its own cheaper systems.
The definition of success is also contested
The dispute is not only about whether AGI exists. It is also about how OpenAI's board would decide that the threshold has been met.
The Information reported that during Microsoft's $10 billion investment in 2023, the original clause was expanded to include a definition called "Sufficient AGI." Under that clause, Microsoft's exclusive rights would end if OpenAI's board determines that its AI can generate the maximum profits investors are entitled to.
The source says that figure was $92 billion for Microsoft at the time but is now higher. The AI would not need to actually produce those profits for the clause to take effect. However, Microsoft must approve the methodology used to make that determination.
This creates a practical problem for both companies. OpenAI has an incentive to preserve the board's power to decide when AGI has arrived. Microsoft has an incentive to make sure such a decision does not prematurely end its access to OpenAI's technology.
The CEOs are reportedly far apart on the issue. The Information says OpenAI's Sam Altman has suggested AGI is within sight. Microsoft's Satya Nadella has called that "AGI hype" and "nonsensiical benchmark-hacking." Nadella argues that the real test would be if global GDP grew by 10 percent a year.
Revenue and intellectual property add to the pressure
The restructuring dispute is tied to money as well as technology. OpenAI is seeking a reduction in the 20 percent revenue share it pays Microsoft, according to The Information.
Microsoft wants more than approval rights. It is seeking to eliminate the AGI clause and extend its intellectual property rights with OpenAI beyond the 2030 contract expiration.
One possible compromise being discussed is replacing AGI with "artificial superintelligence" (ASI). That would move the trigger to a later and more demanding threshold, at least in the structure described by the source article.
Intellectual property access remains another unresolved issue. Microsoft hired Inflection AI CEO Mustafa Suleyman, described in the source as a key OpenAI competitor, to reduce its dependency. But Suleyman reportedly ran into problems trying to replicate OpenAI's models.
He reportedly complained for weeks that OpenAI was not meeting its contractual obligations and raised his frustration directly with then-CTO Mira Murati. The source says selective information sharing by OpenAI made it hard for Microsoft teams to understand techniques such as "chain of thought" reasoning.
A partnership that now looks like a rivalry
Microsoft and OpenAI are still deeply connected, and executives from both companies are reportedly meeting daily. But the relationship described in the source article is no longer a simple investor-partner arrangement.
Microsoft wants continued access, stronger intellectual property rights and more certainty beyond 2030. OpenAI wants room to restructure, reduce its revenue share and preserve the governance terms that could limit Microsoft's access if AGI is declared.
The result is a conflict over definitions, incentives and control. The same contract language that once seemed theoretical now shapes a negotiation over some of the most valuable technology in the AI industry.