Microsoft is embarking on a strategy to achieve “true AI self-sufficiency,” according to the company’s AI chief, Mustafa Suleyman. This involves developing its own advanced foundation models and reducing its reliance on OpenAI, even as the two companies maintain their existing partnership. The move comes after a renegotiated agreement in October 2025 that allows Microsoft to pursue its own artificial general intelligence (AGI) while retaining a 27% stake in OpenAI and continued access to its models through 2032.
Currently, OpenAI’s models power Microsoft’s Copilot, its widely used consumer AI assistant, as well as features integrated into Office and Azure. However, Suleyman explained to the Financial Times that dependence on a single provider introduces risk. Any disruption at OpenAI could directly impact Microsoft products that rely on its technology. Developing its own models, optimized for Copilot, Office, and Azure, provides Microsoft with increased strategic flexibility and financial efficiency.
“Three or four months ago, after renegotiating our partnership, we decided it was time to deliver on true AI self-sufficiency,” Suleyman stated. “After all, it’s the most important technology of our era.”
This isn’t a complete severing of ties. OpenAI remains Microsoft’s primary partner for many applications. However, diversifying its AI sources is a protective measure against potential issues such as technical failures at OpenAI, strained relationships, or abrupt strategic shifts by its partner. Microsoft is effectively hedging its bets.
A Diversification Strategy Already Underway
Microsoft is already actively diversifying by contracting with Anthropic to host its Claude model on Azure, alongside Meta’s Llama, Mistral AI, and other providers. Simultaneously, the company is making substantial investments in its own models, most notably MAI-1.
This multi-source approach is transforming Microsoft from a customer of AI models to an orchestrator of them. The company will be able to switch between providers based on performance, cost, or the specific needs of each product, eliminating the single point of failure currently inherent in its relationship with OpenAI. This allows for a more dynamic and resilient AI infrastructure.
A Massive Infrastructure to Support Independence
Achieving AI self-sufficiency requires a colossal infrastructure, which Microsoft is rapidly deploying. The company has unveiled Maia 200, its in-house AI acceleration chip, and is constructing the Fairwater data center network, which houses some of the world’s most powerful supercomputers. These investments represent hundreds of billions of dollars across the industry.
Microsoft’s deliberate choice to not solely rely on OpenAI ensures it will be well-positioned regardless of future outcomes. The company benefits if OpenAI continues to dominate through its 27% stake and prioritized access. If a competitor emerges, Microsoft has alternatives readily available. And if its own models prove superior, Microsoft controls the entire supply chain. This comprehensive strategy reflects the inherent uncertainty surrounding generative AI, where today’s leaders can be quickly overtaken by unexpected breakthroughs.
The development of MAI-1, previewed in August 2025, demonstrates Microsoft’s commitment. The company described it as an “in-house mixture-of-experts model” pre-trained and post-trained on approximately 15,000 NVIDIA H100 GPUs, with plans to integrate it into certain Copilot text use cases. This signals a clear intent to build models at a significant scale, utilizing the same hardware infrastructure as its competitors.
Microsoft’s Maia 200 chip is designed as an inference accelerator to “dramatically improve the economics of AI token generation,” effectively challenging NVIDIA’s software dominance by pairing custom silicon with a software package intended to reduce CUDA’s influence. This move highlights Microsoft’s ambition to control not only the models themselves but also the underlying compute, security, and billing processes, ensuring they remain “Microsoft-shaped” regardless of which model is currently leading the market.
Microsoft’s strategy isn’t simply about building a competitive AI; it’s about building a resilient and strategically advantageous position in a rapidly evolving technological landscape. The company is preparing for a future where AI dominance is far from certain, and having multiple options – and the infrastructure to support them – is paramount.
