Microsoft is reshaping its artificial intelligence strategy in a major way. The company is working to reduce its reliance on OpenAI while building its own advanced AI systems. Even though the two companies remain partners, Microsoft is clearly preparing to stand on its own in the fast-moving AI race.
The shift became more visible after internal developments and a reworked partnership agreement in late 2025. Together, these moves show a company focused on long-term AI control and flexibility.
Partnership Restructured as Both Companies Gain Freedom
In October 2025, Microsoft and OpenAI signed a revised agreement that changed the structure of their relationship. Microsoft converted its earlier profit-sharing rights into a 27 percent ownership stake in a newly formed OpenAI Group PBC. The deal valued the company at around $135 billion.
Microsoft warns of rising AI spending risks as data center costs raise investor concern
The agreement extended the company’s rights to use OpenAI’s technology through 2032, including access to highly advanced systems. At the same time, the restructuring gave both sides more room to operate independently.
OpenAI gained the ability to seek computing power beyond Microsoft’s Azure cloud platform and approach new investors. Microsoft, in turn, secured the right to develop artificial general intelligence projects on its own or with other partners. The updated structure allowed each company to build future technologies without being tightly restricted by the other.
Microsoft Expands Its Own AI Capabilities
Alongside the partnership changes, Microsoft has accelerated work on its internal AI systems. The company previewed its in-house model, MAI-1-preview, in August 2025. The system uses a mixture-of-experts architecture and was trained on approximately 15,000 Nvidia H100 chips. These chips are among the most powerful processors used for training large AI models. The scale of computing involved shows how serious the company is about developing its own foundation models.
The company has indicated that MAI-1-preview may be integrated into certain Copilot text features. Copilot operates within products such as Microsoft 365 and other productivity tools. Using internal AI models in these services would reduce dependence on OpenAI-powered systems.
Security researchers uncover APT28 campaign exploiting newly disclosed Microsoft Office flaw
Microsoft’s AI division is being led by Mustafa Suleyman. He has described the mission as building “true AI self-sufficiency.” According to him, the company aims to create its own frontier foundation models at a massive scale, supported by what it calls “gigawatt-scale compute.” That phrase refers to extremely large data centers dedicated to training and running AI systems.
At the same time, the company has broadened its AI supplier base. Its cloud platform now hosts models from companies such as xAI, Meta, Mistral, and Black Forest Labs. This approach gives customers access to multiple AI options rather than relying on a single provider.
Reports also suggest that the company tested Claude models from Anthropic for specific Microsoft 365 Copilot tasks. Internal benchmarks reportedly showed strong performance for certain Office-related workloads. In order to evaluate those systems, it even accessed them through a rival cloud provider.
Market Reaction Reflects Investor Concerns
Investors have closely watched these developments. During a recent earnings discussion, analysts pointed out that OpenAI represents about 45 percent of Microsoft’s backlog of future sales. That concentration raised questions about durability and exposure.
Shortly after, the company experienced a historic single-day decline in market value, losing $357 billion. The drop was not directly tied to operational performance but reflected market sensitivity around dependency risks.
The company’s push toward building internal AI models and expanding partnerships appears to address those concerns. By reducing reliance on a single AI provider, Microsoft lowers potential risks connected to business concentration.
Meanwhile, development of its in-house AI systems continues at a rapid pace. Mustafa Suleyman has indicated that the company expects to have its own superintelligence-level model ready within the year.
Microsoft’s current strategy rests on three main elements: maintaining contractual access to OpenAI technology, developing powerful internal AI models, and broadening its ecosystem of AI partners.
The company continues to integrate AI into tools used daily by businesses and consumers. Copilot and other AI-powered features remain central to its product offerings. However, the technology behind those features may increasingly shift toward Microsoft-built systems.
The restructuring of the OpenAI partnership, the launch of MAI-1-preview, and the expansion of alternative AI models together mark a new phase in Microsoft’s artificial intelligence strategy.
