Microsoft restructures to build proprietary superintelligence and reduce its reliance on OpenAI models

Microsoft has initiated a sweeping reorganization of its artificial intelligence operations, signaling a profound strategic pivot toward the development of proprietary frontier models and the pursuit of artificial superintelligence.[1][2][3][4][5][6][7] This restructuring marks a significant departure from the company’s recent narrative, which suggested that foundational AI models were becoming a commoditized layer of the technology stack.[4] By centralizing its engineering efforts and elevating the mission of its internal AI research units, the tech giant is positioning itself to own the core intelligence that powers its sprawling ecosystem, moving beyond its previous reliance on external partners like OpenAI to define the frontier of the industry.[2]

The reorganization is characterized by a major leadership shuffle designed to separate the engineering of consumer products from the high-stakes research required to build next-generation models.[7] Mustafa Suleyman, the co-founder of DeepMind who joined Microsoft last year to lead its AI division, will now shift his primary focus to a dedicated superintelligence group.[3][7] This team is tasked with pushing the boundaries of what is possible in AI, focusing exclusively on the research and development of frontier-scale models over a five-year horizon. Meanwhile, the company is consolidating its consumer and commercial Copilot teams into a single unified organization.[6][4][3][8][9][5] Jacob Andreou, a former Snap executive who previously managed product and growth for Microsoft’s AI division, has been promoted to Executive Vice President of Copilot Product Experience.[6][4] Reporting directly to CEO Satya Nadella, Andreou will oversee the design and engineering of Microsoft’s AI assistant, while a broader leadership team including LinkedIn CEO Ryan Roslansky, Perry Clarke, and Charles Lamanna will manage the integration of these capabilities across Microsoft 365 and other core platforms.[3]

This shift in strategy is particularly notable given the public commentary provided by Satya Nadella only months ago. Throughout late 2025 and early 2026, Nadella frequently argued that the industry was entering an era where raw model performance was becoming a commodity. He posited that the real competitive advantage would lie in what he termed context engineering and system-level orchestration—the ability to ground AI in specific enterprise data and workflows.[10] This philosophy was reflected in the launch of products like Agent 365 and the Copilot Cowork tool, which focused on multi-step task execution rather than the underlying intelligence of the model itself. However, the current restructuring suggests a realization within Redmond that the "intelligence" layer is not yet a solved problem and that relying on third-party intellectual property for the foundational core of its software creates unacceptable strategic risks and cost structures.

Internal pressure has likely played a role in this reversal. Reports indicate that Nadella has been increasingly critical of Copilot’s current capabilities, expressing frustration that integrations with key services like Outlook and Gmail have struggled to meet performance expectations.[4] Furthermore, while Microsoft’s partnership with OpenAI remains a cornerstone of its strategy, the financial and operational costs of this relationship are immense. By developing its own proprietary "MAI" series of models and chasing superintelligence internally, Microsoft aims to achieve significant reductions in the cost of goods sold for its AI services.[4] The goal is to create enterprise-tuned model lineages that are more efficient to run at scale than general-purpose frontier models, thereby improving the company’s margins as it attempts to convert its hundreds of millions of Office 365 subscribers into paying AI users.

The scale of this ambition is supported by an unprecedented investment in physical infrastructure.[2][1] At the heart of Microsoft’s superintelligence mission is Project Stargate, a joint venture with OpenAI that has reportedly ballooned from a $100 billion supercomputer plan into a $500 billion roadmap for global "AI Superfactories." These facilities are designed as monolithic entities rather than traditional data centers, featuring custom networking fabrics and millions of next-generation accelerators, including Nvidia’s Vera Rubin platform and Microsoft’s own custom-designed Maia chips. The sheer energy requirements of these facilities, which are projected to consume up to five gigawatts of power per campus, have led Microsoft to secure dedicated nuclear energy sources, such as the recently reopened Crane Clean Energy Center at Three Mile Island. This massive industrial undertaking underscores the company’s belief that the path to superintelligence is paved with raw compute and specialized hardware.

The broader implications for the AI industry are profound. For years, Microsoft and OpenAI have maintained a symbiotic but increasingly complex relationship, with Microsoft providing the capital and cloud infrastructure in exchange for exclusive intellectual property rights. However, as Microsoft builds out its own frontier research capabilities and OpenAI secures massive independent funding from the likes of SoftBank and Nvidia, the two entities are increasingly operating as "frenemies." The restructuring allows Microsoft to hedge its bets, ensuring that even if its relationship with OpenAI evolves or if OpenAI moves toward other cloud providers, Microsoft retains the internal expertise and model weights necessary to power its independent future.

This move also intensifies the competition with other tech titans like Google, Meta, and Amazon. While Google has successfully integrated its research arm, DeepMind, with its product teams through the creation of Google DeepMind, Microsoft is now attempting a similar synthesis but with a clearer separation between the "dreamers" chasing superintelligence and the "builders" shipping product. This structural clarity is intended to accelerate the development of agentic systems—AI that does not just respond to prompts but can plan and execute complex workflows autonomously. As the industry moves from chatbots toward these more capable agents, the underlying model’s ability to reason and handle long-horizon tasks becomes the primary differentiator once again.

Ultimately, Microsoft’s pivot toward superintelligence represents a bet-the-company moment.[7] By moving beyond the "commodity" narrative, Nadella is acknowledging that the race for artificial general intelligence is still in its early stages and that the winner will be determined by who owns the most advanced synthetic mind. The next five years will determine if this restructuring can bridge the gap between Microsoft’s current AI assistants and the transformative, high-reasoning systems Suleyman’s team is now tasked with building. If successful, Microsoft could transition from a software platform that hosts intelligence to a company that defines and controls the very nature of superintelligence itself, cementing its dominance for the next decade of the silicon era.