Microsoft has commenced a vigorous search for novel artificial intelligence startups, aiming to lessen its reliance on OpenAI, its primary collaborator of recent years, in the future. The corporation is evaluating both outright acquisitions and establishing strategic alliances with promising AI developers. Insiders suggest that Microsoft’s objectives extend beyond mere technological access to encompass the retention of top-tier AI talent. Within the company, there is an ambitious directive set to craft one of the premier AI models as early as next year.
During the spring, Microsoft explored the possibility of acquiring Cursor, a startup specializing in AI tooling for software development. Nevertheless, the potential deal was abandoned due to concerns that regulatory bodies might object to the purchase, given Microsoft’s existing ownership of GitHub Copilot.
Currently, Microsoft is engaged in discussions with the startup Inception, which was founded by a team emerging from Stanford University. This company is pioneering unconventional language models that leverage diffusion technology—an approach more commonly associated with image and video generation. This methodology allows for the simultaneous processing of multiple semantic elements, potentially offering a significant speed enhancement for the AI. Inception has already secured substantial backing, including capital from Microsoft’s venture arm, M12. According to sources, the startup’s valuation could surpass $11 billion.
Meanwhile, the AI market is currently experiencing a boom. Major corporations are vying fiercely for developers and promising projects, with compensation for AI researchers escalating to tens of millions of dollars. Among Microsoft’s rivals is noted to be Elon Musk’s SpaceX, which is aggressively scaling up its own AI division, xAI.
Reports also indicate that OpenAI is gradually starting to transcend its exclusive partnership with Microsoft. The company has reportedly gained the latitude to work with alternative cloud providers, including Amazon. Concurrently, Microsoft is striving to cultivate its indigenous AI capabilities so as not to be solely dependent on OpenAI’s technology stack.
Within the broader industry, the race to engineer ultra-powerful models featuring trillions of parameters continues unabated. Whereas leading models a few years ago clocked in around one trillion parameters, premier laboratories are now developing systems targeting approximately 10 trillion parameters.
About the Author
