Once “as close as brothers,” Microsoft and OpenAI are now locked in a bitter dispute over the definition of Artificial General Intelligence (AGI) and the terms of their contract. These two companies used to collaborate closely in AI, reaping mutual benefits. But as AI has advanced and the potential profits have skyrocketed, their once-close partnership has started to show serious strains.

Why did they “tie themselves together” in the first place?

Let’s go back to 2019. At that time, OpenAI was an ambitious but cash-strapped AI research lab. Developing next-generation AI required enormous amounts of money, especially given the sky-high costs of compute power. That’s when Microsoft extended a lifeline: a $1 billion investment, plus access to its Azure cloud computing infrastructure.

For OpenAI, this money and compute power were nothing short of a lifesaver. In return, Microsoft received “exclusive use rights,” allowing it to embed GPT and other models into products like Office, GitHub Copilot, and Bing Chat—turning them into big money-makers.

The “ticking time bomb” in the deal

However, to prevent AGI from being completely commercialized and misused, OpenAI included a very important clause in the 2019 deal:

If OpenAI achieves AGI, it can immediately terminate Microsoft’s access to its technology.

In other words: if they actually create AGI, Microsoft gets “kicked out.” This clause reflected OpenAI’s original mission—to treat AGI as a benefit for all humanity rather than letting a single company monopolize it.

Crucially, the power to define when AGI has been achieved lies with OpenAI’s nonprofit board. Put simply, “we decide when it’s AGI, and if we say so, you’re out.” Microsoft agreed to this at the time, because back then AGI still felt like science fiction—nobody expected it to happen anytime soon.

Why are they at odds now?

AI’s rapid progress in recent years has shocked everyone. At the end of 2022, ChatGPT’s viral success turned OpenAI into the most influential AI company in the world. Microsoft cashed in: Office 365 with Copilot could justify price hikes, Bing finally got some users, and Azure’s AI business boomed.

But that success also brought problems: OpenAI was now getting closer to its AGI goal.

OpenAI CEO Sam Altman even claimed this year that they have “figured out how to build AGI,” which infuriated Microsoft. Microsoft CEO Satya Nadella shot back, dismissing that claim as “absurd metrics gaming” and rejecting the idea that OpenAI alone could declare AGI achieved.

Why does this make Microsoft so nervous? Because the original deal spells it out plainly: once AGI is reached, Microsoft’s rights to use that technology get cut off. Microsoft has invested $13 billion so far, counting on OpenAI to keep them ahead of Google, Amazon, and Meta in the AI race. If they get cut off suddenly, that entire bet could go up in smoke.

Microsoft’s demands vs. OpenAI’s stance

Reportedly, Microsoft has demanded in negotiations that this “AGI kill-switch” clause be removed so that it can keep using the most advanced models no matter what. OpenAI, for its part, has refused to budge, insisting on preserving its exit rights.

Their core disagreements boil down to two main issues:

1️⃣ Who defines what “AGI” is?

There’s no industry standard for AGI right now. Microsoft doesn’t want OpenAI to have unilateral control over this definition and wants the criteria locked down in advance or jointly determined.

2️⃣ Should AGI be something big companies can keep profiting from?

OpenAI wants to retain the ability to pull the plug at the last moment to prevent Microsoft from fully commercializing AGI. After all, OpenAI’s original nonprofit mission was about managing AGI “safely” for all of humanity.

The cracks go beyond the contract

What’s worse is that these tensions have spilled over into their day-to-day partnership:

In 2024, OpenAI complained that Microsoft’s Azure wasn’t giving them enough GPUs. OpenAI responded by turning to Google Cloud and signing deals with Oracle. This was a huge embarrassment for Microsoft.

Microsoft’s AI lead Mustafa Suleyman has repeatedly complained that OpenAI is “holding back” on sharing details, especially around critical mechanisms like chain-of-thought reasoning. Microsoft feels like it’s being treated as just another customer—even though it’s OpenAI’s biggest investor.

Microsoft hasn’t been sitting still. It’s been developing its own Phi small models and even hired the core team from Inflection to build its own in-house AI capability that could rival OpenAI.

OpenAI is changing too

Adding even more complexity, OpenAI itself is evolving.

Originally, it was a pure nonprofit research lab. To raise funding, it created a “capped-profit” subsidiary so investors like Microsoft could earn returns. Recently, OpenAI announced plans to convert from this profit-seeking structure back into a Public Benefit Corporation (PBC) controlled by its nonprofit board.

The tricky part is that OpenAI can’t make that shift without Microsoft signing off. And Microsoft’s frustration over the AGI provision has become a bargaining chip in those negotiations.

In the end, it’s a battle over money and power

On the surface, this standoff is about “how to define AGI,” but underneath, it’s really a fight over money and control.

Microsoft wants to protect its massive investment and avoid being cut off if AGI is achieved. OpenAI wants to preserve its right to say “no” in the final moment and safeguard its mission of developing AGI for humanity—not just for one giant corporation.

They got into bed with each other because they both needed the partnership at the time. Now OpenAI has more leverage and bigger ambitions. Microsoft is watching OpenAI’s power grow—and realizing it could lose control. That’s not something it can accept quietly.

What happens next?

Negotiations will likely be extremely tough. OpenAI still needs massive funding to keep pushing forward, but doesn’t want to cede strategic control to Microsoft. At the same time, Microsoft isn’t just pushing OpenAI to relax the AGI restriction—it’s also busy building its own backup plans.

Whatever deal they eventually reach (if any), this fight over who gets to define AGI will have far-reaching implications for the entire global AI landscape. After all, if “general intelligence” really is developed, it would be the most powerful technology in human history—deciding who gets to lead it, use it, and govern it will never be an easy question.