Microsoft is actively working to reduce its dependence on OpenAI, according to recent reports from The Information and other news outlets. Despite having invested over $13 billion in OpenAI since 2019, Microsoft appears to be pursuing multiple strategies to develop its own AI capabilities and test models from other providers, signaling a potential shift in what was once considered an unbreakable partnership.
The technology giant is reportedly developing an in-house reasoning model called MAI, which it plans to sell to other developers, directly competing with OpenAI’s offerings. Microsoft has also been testing alternative AI models from competitors including Elon Musk’s xAI, Meta, and Chinese company DeepSeek as potential replacements for ChatGPT in its Copilot products.
This strategic pivot comes amid several developments that have strained the relationship between the two companies:
- Microsoft’s Copilot has received a lukewarm reception in enterprise settings due to high costs and limited results
- In January, Microsoft allowed OpenAI to break free from exclusive use of Azure for its hosting needs
- OpenAI recently announced a $500 billion plan to build new data centers with Oracle and SoftBank
- Tensions have emerged over OpenAI’s reluctance to share technical documentation about its latest o1 reasoning model
The changing relationship
Microsoft CEO Satya Nadella had previously questioned why the company would develop its own foundational models when it could use OpenAI’s technology. However, this position appears to have evolved as OpenAI demonstrates increasing independence.
Mustafa Suleyman, who joined Microsoft in March 2024 to lead its AI division after co-founding Inflection AI, has been particularly active in pushing for alternatives. During a video call with senior leaders from both companies last fall, Suleyman reportedly expressed frustration over OpenAI’s unwillingness to provide documentation about how it had developed o1.
Despite these tensions, the companies remain contractually connected. Microsoft holds exclusive rights to OpenAI’s models for its own products until 2030, making a complete separation difficult in the near term.
Strategic hedging
Microsoft’s approach appears to be one of strategic hedging: placing multiple bets to ensure it captures value in the AI sector regardless of which company’s models ultimately prevail. This includes:
- Building its own AI models to control costs and technology
- Testing third-party models from various providers
- Maintaining its existing relationship with OpenAI while developing alternatives
- Focusing on the application layer where Microsoft has enterprise dominance
The outcome of this shifting partnership could significantly impact the AI industry landscape, as two of its most influential players reconsider their positions and potentially become direct competitors.
Sources: Gizmodo, Tech Startup