If the three years since the release of ChatGPT have signaled OpenAI’s dominance of generative artificial intelligence, it’s worth recalling that the company’s rapid rise would have been impossible without another Big Tech backer.
In 2019, Microsoft agreed to supply OpenAI all the compute it needed, with near exclusivity. In exchange, Microsoft retained the right to use OpenAI’s tech until the arrival of artificial general intelligence, or AGI: the point at which AI systems are able to act like humans and respond to whatever task they’re given, regardless of whether they’ve been trained to solve it.
As generative AI’s capabilities blew past initial expectations, the question of when AGI will arrive became a major point of contention in OpenAI and Microsoft’s agreement. In addition, Microsoft feared that OpenAI’s plans to restructure toward a public benefit corporation and open its platform could dilute its influence.
Then, in late October, the companies announced a new arrangement, one that looks like a win for Microsoft. “Microsoft’s IP rights for both models and products are extended through 2032 and now include models post-AGI, with appropriate safety guardrails,” according to a company blog post announcing the renegotiation.
The AGI clause, then, is no more. So what does it tell us about both companies’ beliefs about the path to AGI—and how long it could take?
In many ways, the announcement felt inevitable. Microsoft had reportedly spent months trying to remove the AGI clause, for fear that AGI’s arrival—or OpenAI’s claim of its arrival—could threaten its access to the most advanced generative AI models on the planet, just as such technologies become commercially invaluable. After investing more than $13 billion, Microsoft wanted to ensure it could maintain access.
Michael Veale, a researcher at University College London, wonders whether the world is reading too much into the change in agreement. He reckons that the firms have planned to “scrap it because it’s too gameable”—or too easy for OpenAI to declare when it has been reached, at its own convenience—“to provide the commercial certainty they want from a contract,” he says.
Veale believes that AGI is too woolly a concept on which to base a long-term agreement with such high financial stakes. It’s “hard to say” what the decision means on the basis of the reporting around the term’s changes, he adds.
Alessandra Russo, a professor in applied computational logic at Imperial College London, is also uncertain about what to make of the changed terms of the agreement and how it interacts with both companies’ timeline for AGI. One thing she is confident about is her belief that AGI isn’t on the horizon. “People are always moving the posts a little bit further back, because of the realization that this technology is not as was initially expected,” she explains.
Catherine Flick, professor of AI ethics at the University of Staffordshire, shares this take, noting that the shift is more an indication of both companies trying to sustain the hype around AGI to justify their investments in the tech, rather than a deep-seated belief that AGI truly is around the corner.
The practical snag with an AGI trigger is verification. There’s no agreed test or arbiter, which makes any clause gameable for the commercial certainty both sides want. As Flick puts it: “How do you even verify AGI?”
One way the two companies have tried to solve this is by agreeing to remove the unilateral declaration of AGI that has previously given Microsoft pause. The companies say that “once AGI is declared by OpenAI, that declaration will now be verified by an independent expert panel”—broadening the number of people who can or would declare it to be so.
Even if that point arrives, Flick argues declaring AGI is “a mutually sustaining confidence boost” designed to steady nerves across the wider AI supply chain and to keep capital flowing while capabilities progress slowly, rather than leaping. It’s also part of the “perpetuation of this hype cycle,” she says, at a moment when definitions blur and milestones remain mobile.
That said, at least publicly, those in charge of the companies are accelerating their timeframes for when AGI might arrive.
In January, Sam Altman published a blog post declaring that the world is getting “closer to AGI,” and that his company was making preparations for how to deal with its imminent arrival. Mustafa Suleyman, CEO of Microsoft’s AI division, isn’t as bullish about the timescale. In a December 2024 interview, he pegged the arrival of AGI “within the next five to seven years.”
Not everyone agrees. “The fact that they’re still talking about AGI,” says Flick, “it’s cloud cuckoo land.”
