Microsoft is working to reduce its dependence on OpenAI for its flagship AI product, Microsoft 365 Copilot. This shift is part of a broader strategy to lower costs and improve performance for enterprise users. While Microsoft has long partnered with OpenAI, including integrating its GPT-4 model into 365 Copilot when it was first announced in March 2023, the company is now exploring other AI models.
Microsoft has been working to incorporate both internal and third-party AI models into 365 Copilot, sources familiar with the matter told Reuters. One of the main concerns for Microsoft has been the cost and speed of using OpenAI’s models, which could affect enterprise users’ experience.
Also read: Google wants US govt to break Microsoft and OpenAI partnership, here’s why
A Microsoft spokesperson emphasised that OpenAI remains an important partner, especially for the most advanced AI models. “We incorporate various models from OpenAI and Microsoft depending on the product and experience,” the spokesperson said.
To achieve this goal, Microsoft has been training its own smaller AI models, including Phi-4. Additionally, the company is working on customising open-weight models to improve the performance of 365 Copilot. By incorporating these new models, Microsoft hopes to streamline the AI processes and reduce operational costs, which could lead to potential savings for customers.
Also read: Microsoft calls OpenAI its “competitor” because of SearchGPT
Microsoft’s move to rely less on OpenAI comes as part of a broader strategy to manage costs in an increasingly competitive AI landscape. While the company remains committed to OpenAI for some of its most cutting-edge models, this new direction reflects Microsoft’s growing desire to diversify its AI offerings and reduce dependence on a single partner.
In the long run, this strategy could make 365 Copilot more affordable and efficient, benefiting both Microsoft and its customers.
Also read: Elon Musk wants court to stop OpenAI’s for-profit transition, here’s why