Microsoft Signals Plan to Reduce OpenAI Dependence

- Microsoft has invested over $13 billion in OpenAI, resulting in an approximately 27% stake in the company. Despite this, a revised agreement in October 2025 prompted Microsoft to pursue its own advanced AI technology more independently. - The new head of Microsoft AI is Mustafa Suleyman, who co-founded Google's DeepMind and joined Microsoft in March 2024. He is leading the charge for "true self-sufficiency," which includes developing in-house foundation models slated for release this year. - Under the restructured deal, Microsoft's intellectual property rights for OpenAI's models are extended through 2032, even including post-Artificial General Intelligence (AGI) systems. In exchange, OpenAI gained more freedom to work with new investors and infrastructure partners. - To diversify its AI portfolio beyond OpenAI, Microsoft is also investing in competing AI firms, including Anthropic and France's Mistral. - Microsoft is developing its own family of AI models, reportedly known internally as MAI. This is part of a broader strategy to have alternatives to OpenAI's technology for its Copilot products. - The company's commitment to building its own AI infrastructure is substantial, with projected capital expenditures of $140 billion for the fiscal year. This includes building massive datasets and developing its own AI-accelerator chip, the Maia 200. - While Microsoft's Copilot and other AI services currently rely on OpenAI models, developing its own models offers more strategic independence and financial efficiency. - The strategic shift also comes as OpenAI faces a challenging financial situation, requiring constant cash infusions to operate and being on the hook for massive future compute spending contracts. OpenAI has also started to diversify its infrastructure partners, making deals with Oracle and AWS.