In a further indication of Microsoft's shift away from its exclusive partnership with OpenAI—previously observed in the cloud sector—the technology giant is in the process of developing its own artificial intelligence (AI) reasoning models, referred to internally as 'MAI.' This initiative aims to lessen reliance on OpenAI while simultaneously enhancing Microsoft's AI capabilities. A report from The Information reveals that these models are intended to execute tasks that necessitate intricate reasoning and problem-solving, with the goal of matching or exceeding the performance of leading models from OpenAI, a pioneer in the AI field.
Microsoft is currently exploring the integration of these MAI models into its offerings, including Microsoft 365 Copilot. Additionally, the company is contemplating the release of these models as an application programming interface (API) for external developers later this year. This development would enable developers to embed Microsoft's AI functionalities into their own applications, potentially leading to cost savings and greater flexibility compared to relying exclusively on OpenAI's technology.
This strategic move is not unexpected, as reports from December 2024 indicated that despite a substantial investment of $13.75 billion in OpenAI since 2019, Microsoft is actively seeking alternatives to the AI model developed by the creator of ChatGPT, particularly for its 365 Copilot.
Initially, Microsoft 365 Copilot, which debuted in 2023, prominently featured OpenAI's GPT-4 model. However, the current landscape appears to be evolving, likely driven by three primary factors: the desire to avoid monopolistic practices or exclusive dependencies, the need to reduce costs, and the objective to enhance operational speed.
In recent developments, Microsoft’s AI division, led by Mustafa Suleyman, has successfully trained a new family of models that demonstrate performance comparable to the leading models from OpenAI. The team is employing chain-of-thought techniques to develop reasoning models, which enhance the ability to generate answers through intermediate reasoning when addressing complex problems.
Moreover, a recent report reveals that Microsoft has initiated testing of AI models from xAI, Meta, and DeepSeek as potential alternatives to OpenAI’s technology for integration within Copilot.
This advancement follows the introduction of Copilot AI features to Microsoft 365 for Personal and Family subscription plans last month. Previously, these AI functionalities were only accessible through a separate Copilot Pro subscription, which incurred an additional fee of $20 per month.
Earlier this year, Microsoft announced a shift in its partnership with OpenAI, stepping back from its previous exclusive role as the cloud provider for OpenAI’s computing requirements. Under the new agreement, Microsoft maintains a 'right of first refusal' (ROFR), allowing it the first opportunity to supply cloud resources to OpenAI. However, OpenAI is now permitted to explore services from other cloud providers, including Oracle.
This change in the Microsoft-OpenAI relationship aligns with the unveiling of the Stargate Project, a collaborative effort involving OpenAI, Oracle, SoftBank, and Nvidia, which aims to invest up to $500 billion in developing the infrastructure necessary for next-generation AI models in the United States. Concurrently, Microsoft plans to allocate $80 billion towards AI-focused data centers and cloud initiatives in fiscal year 2025, which includes an investment of $3 billion in India.
