Microsoft Drops OpenAI Models in Copilot
- As of March 8, 2025, Microsoft stands as a primary investor in OpenAI, with investments approximating $14 billion.
- For several weeks, Microsoft has ceased to be the exclusive cloud provider for OpenAI.
- A point of contention reportedly arose between Mustafa Suleyman and former OpenAI CTO Mira Murati.
Microsoft and OpenAI: A Shifting AI Landscape in 2025
Table of Contents
- Microsoft and OpenAI: A Shifting AI Landscape in 2025
- Microsoft and OpenAI: Navigating teh Shifting AI Landscape – A Q&A Guide
- The Microsoft-OpenAI Partnership: Key Questions Answered
- What is the current state of the Microsoft-OpenAI partnership in 2025?
- Why is Microsoft reducing its dependence on OpenAI?
- What is Microsoft’s MAI model, and how does it compare to OpenAI’s offerings?
- How does Microsoft’s AI strategy affect Copilot?
- What external models beyond OpenAI is Microsoft evaluating?
- How has the Azure OpenAI Service evolved?
- What is Microsoft’s “right of first refusal” (ROFR) agreement with OpenAI concerning AI capacity?
- Will the development of MAI hinder the partnership between Microsoft and OpenAI?
- The Microsoft-OpenAI Partnership: Key Questions Answered
- Summary Table : Microsoft and openai Partnership
As of March 8, 2025, Microsoft stands as a primary investor in OpenAI, with investments approximating $14 billion. However, future trajectories suggest a potential divergence. sources indicate that Mustafa Suleyman, head of Microsoft’s AI division, aims to diminish reliance on the California-based startup. Microsoft has reportedly finalized the development of its own model family, internally known as MAI, which promises performance metrics comparable to OpenAI’s offerings.
Microsoft’s AI Strategy: Reducing Dependence on OpenAI
For several weeks, Microsoft has ceased to be the exclusive cloud provider for OpenAI. This allows OpenAI to pursue agreements with other companies if Azure cannot guarantee sufficient computing power. The remaining contractual conditions remain valid untill 2030, including the exchange of information on research and development of models.
A point of contention reportedly arose between Mustafa Suleyman and former OpenAI CTO Mira Murati. Allegedly, the California startup did not provide documentation on the chain of thought followed by the o1 model before responding to user questions, possibly violating contractual agreements.
Copilot with Proprietary Models?
According to sources, Suleyman has planned a “divorce” from OpenAI. Microsoft researchers have completed the development of MAI models that offer similar performance to those of the California startup. Some models are based on reasoning, like o1.Tests are reportedly underway with versions of Copilot that leverage these proprietary models instead of GPT.
Microsoft’s AI division has also tested models from Anthropic (Claude), xAI (Grok), Meta (Llama), and DeepSeek (R1). Suleyman’s goal is to reduce dependence on OpenAI to the point of self-sufficiency. The MAI models are expected to be released by the end of the year and will be accessible to developers via APIs.
The Future of AI Compute and Development
microsoft’s evolving strategy includes meaningful changes to its AI compute partnerships. according to a recent announcement, the new agreement “includes changes to the exclusivity on new capacity, moving to a model where Microsoft has a right of first refusal (ROFR).” This shift aims to “further support OpenAI,” while also allowing Microsoft to diversify its AI capabilities.
The partnership between OpenAI and Microsoft extends beyond compute, encompassing a “multi-year, multi-billion dollar investment to support OpenAI’s independent research and development of safe and powerful AI.” This collaboration also focuses on “building and deploying OpenAI’s technology through Azure, GitHub, and other platforms.”
microsoft’s Azure OpenAI Service provides access to OpenAI’s language models, including GPT-4 and GPT-3.5-Turbo.Developers can “create a resource, deploy a model, and use the API features and responsible AI content filtering” through this service.
This article dives into the evolving relationship between Microsoft and OpenAI as of March 8, 2025. We address key questions about Microsoft’s AI strategy, its investment in OpenAI, the growth of its own AI models, and the potential impact on the future of AI compute and development.
The Microsoft-OpenAI Partnership: Key Questions Answered
What is the current state of the Microsoft-OpenAI partnership in 2025?
As of March 8, 2025, Microsoft maintains a significant investment in OpenAI, approximated at $14 billion. While this partnership remains considerable, Microsoft is actively working to diversify its AI capabilities and reduce its reliance on OpenAI.
Why is Microsoft reducing its dependence on OpenAI?
Several factors contribute to Microsoft’s strategy:
Diversification: Microsoft aims to broaden its access to AI technologies beyond a single provider.
Compute Capacity: OpenAI is no longer exclusively reliant on Microsoft Azure for cloud computing, allowing openai to seek partnerships with other providers if Azure cannot meet its needs.
Model Development: Microsoft has developed its own AI model family, MAI, with performance metrics intended to rival OpenAI’s models.
Contractual Issues: Disagreements have reportedly surfaced regarding openness in model documentation.
What is Microsoft’s MAI model, and how does it compare to OpenAI’s offerings?
MAI (Microsoft AI) refers to Microsoft’s internally developed family of AI models. It is designed to offer performance comparable to OpenAI’s models, with the goal of self-sufficiency in AI capabilities. These models also will be accessible to developers via APIs.
According to The Facts report, the MAI model is training a new large language model estimated to have roughly 500 billion parameters. Microsoft is also training reasoning models that could compete with models from OpenAI, DeepSeek, and Alibaba
How does Microsoft’s AI strategy affect Copilot?
Microsoft is reportedly testing versions of its copilot AI assistant that utilize its proprietary MAI models rather of relying solely on GPT models from OpenAI
What external models beyond OpenAI is Microsoft evaluating?
Microsoft’s AI division has tested AI models from several companies:
anthropic (Claude)
xAI (Grok)
Meta (Llama)
DeepSeek (R1)
How has the Azure OpenAI Service evolved?
Microsoft’s Azure OpenAI Service continues to provide access to OpenAI’s language models, including GPT-4 and GPT-3.5-Turbo. Developers can leverage the service to build and deploy AI-powered applications, utilizing API features and responsible AI content filtering
What is Microsoft’s “right of first refusal” (ROFR) agreement with OpenAI concerning AI capacity?
The new agreement “includes changes to the exclusivity on new capacity, moving to a model where Microsoft has a right of first refusal (ROFR).” This shift aims to “further support OpenAI,” while also allowing Microsoft to diversify its AI capabilities
Will the development of MAI hinder the partnership between Microsoft and OpenAI?
While the long-term implications are still unfolding, the partnership continues with a multi-year, multi-billion dollar investment to support OpenAI’s research and development.
Summary Table : Microsoft and openai Partnership
| Feature | OpenAI | microsoft (Azure OpenAI) | Notes |
| ————————- | —————————————– | ——————————————— | ——————————————————————————————————————————————————– |
| AI Models | GPT series (GPT-4 , GPT-3.5) | GPT series, MAI (in development) | Microsoft investing in proprietary models to reduce dependence on OpenAI; MAI is expected to compete with current models. |
| cloud Infrastructure | Multiple companies | Azure (primary) | OpenAI no longer exclusively reliant on Microsoft Azure |
| Strategic Direction | Independent research and development | Diversifying AI capabilities; reducing reliance | Microsoft retains right of first refusal on new compute capacity while also partnering with other AI providers in the model testing phase and for MAI. |
| Contractual Agreement | R&D information sharing,compute support | Multi-year investment,Azure deployment | Agreements valid until 2030; potential disputes over transparency in model documentation and OpenAI not providing documentation. |