Microsoft is planning to make significant investments in its own AI chip cluster to achieve greater self-sufficiency in artificial intelligence, according to Mustafa Suleyman, Microsoft’s AI chief, during an all-employee town hall. Suleyman emphasized the importance of being able to build world-class AI models in-house, even as Microsoft continues to support OpenAI through its cloud infrastructure and commercial partnership. The move signals a shift in Microsoft’s AI strategy, as the company seeks to reduce reliance on external models and expand its internal capabilities. While Microsoft has long partnered with OpenAI—using its technology in products like Copilot and offering it through Azure OpenAI—the relationship is currently undergoing tense contract renegotiations. Despite this, Microsoft remains committed to supporting OpenAI, as CEO Satya Nadella reiterated during the meeting, noting that both companies are each other’s customers and partners. Suleyman highlighted that Microsoft’s recent MAI-1-preview, the company’s first foundation model trained entirely in-house, was developed using just 15,000 Nvidia H100 chips—a relatively small cluster compared to competitors. He pointed out that models from Google, Meta, and Elon Musk’s xAI were trained on clusters six to ten times larger, underscoring the need for Microsoft to scale up its hardware infrastructure. The company is already exploring multiple avenues in AI, including leveraging open-source models, collaborating with other developers, and building proprietary models. MAI-1-preview currently ranks 24th among text models on LMArena, a key benchmark, indicating that Microsoft still has substantial room for improvement. Suleyman stressed the need for pragmatism: while Microsoft aims to develop top-tier AI models internally, it will also use external models when necessary. The investment in its own chip cluster is a critical step toward achieving that goal, enabling faster training, greater control over AI development, and long-term strategic independence.