Need to Be Self-Sufficient in AI: Mustafa Suleyman on Microsoft’s Superintelligence Quest
For years, Microsoft has operated under significant constraints in the artificial intelligence race. The tech giant, despite its massive resources and technical prowess, was essentially sidelined from pursuing advanced AI models through 2030, bound by a restrictive agreement with OpenAI that positioned the startup as the undisputed leader in frontier AI development. Microsoft’s role was relegated to playing infrastructure provider, cloud partner, and financial backer—a position that, while profitable and strategically important, kept the company from directly competing in the race to build superintelligence.
Now, that dynamic has fundamentally changed. In a significant renegotiation of its partnership with OpenAI announced in November 2025, Microsoft has secured the freedom to independently pursue artificial general intelligence (AGI) and superintelligence, either alone or in partnership with third parties. This shift has prompted Microsoft to launch what company insiders are calling the most ambitious AI initiative in the company’s history: the Microsoft AI Superintelligence team.
Speaking to Business Insider, Mustafa Suleyman, Microsoft’s Chief AI Officer, crystallized the company’s new strategy with a single, powerful statement: “Microsoft needs to be self-sufficient in AI. And to do that, we have to train frontier models of all scales with our own data and compute at the state-of-the-art level.”
The Constraints That Held Microsoft Back
Understanding the significance of this moment requires understanding just how restrictive the previous OpenAI partnership was. According to sources familiar with the original agreement, Microsoft was explicitly barred from developing its own AGI systems through 2030. This clause reflected OpenAI’s desire to maintain exclusive control over frontier AI development while leveraging Microsoft’s unparalleled cloud computing infrastructure and capital resources.
The restriction forced Microsoft into a supporting role. Rather than building its own cutting-edge foundation models that could compete with OpenAI’s GPT series, Microsoft focused on smaller-scale AI projects, post-training OpenAI’s existing models for specific applications, and channeling billions of dollars into infrastructure and investment to support OpenAI’s ambitions.
This arrangement was awkward and ultimately limiting. Microsoft possessed $300 billion in annual revenues, massive cloud computing capacity, and some of the world’s most talented AI researchers, yet was contractually prohibited from deploying these assets to build frontier-grade AI systems. Meanwhile, OpenAI, despite its technical brilliance, struggled with profitability and sustainability questions, creating friction in the relationship.
The tension between the two companies intensified in recent months as OpenAI announced plans to transition to a for-profit structure to raise capital and prevent hostile takeovers. Microsoft, concerned about protecting its investment and influence, initially blocked OpenAI’s restructuring plans. The stalemate ultimately prompted both companies to renegotiate their fundamental relationship.
The New Deal: Freedom at Last
The renegotiated partnership, finalized in November 2025, represents a historic turning point. Rather than maintaining the previous restrictions, the new agreement grants Microsoft explicit permission to independently pursue AGI development. Notably, the agreement includes provisions that if Microsoft uses OpenAI’s intellectual property to develop AGI, the resulting models will be subject to compute thresholds—significantly larger than current system sizes—but Microsoft retains the right to build its own systems without OpenAI’s involvement.
This freedom has immediately catalyzed Microsoft’s most aggressive push into frontier AI development. The company announced the formation of Microsoft AI Superintelligence, a dedicated team tasked with building what Suleyman describes as a “world-class, frontier-grade research capability in-house”.
The team’s mission extends far beyond simply replicating OpenAI’s work. Instead, Microsoft is pursuing what Suleyman terms “Humanist Superintelligence”—a fundamentally different philosophical approach to advanced AI development. Rather than building uncontrolled autonomous systems capable of operating with minimal human oversight, Microsoft’s vision centers on systems that are “carefully calibrated, contextualised, within limits” and explicitly designed to serve human purposes.
The Humanist Superintelligence Philosophy
Suleyman has been increasingly vocal about his concerns regarding an unbounded race to AGI without consideration for human control and alignment. In multiple interviews and blog posts, he has articulated a vision of AI development that prioritizes human agency and safety over raw capability maximization.
“We can’t build superintelligence just for superintelligence’s sake,” Suleyman wrote in announcing the superintelligence team. “It’s got to be for humanity’s sake. It’s not going to be a better world if we lose control of it.”
This stance explicitly rejects what he calls “the narrative of the AI race to AGI,” which he views as irresponsible and directionless. Instead, Microsoft’s approach frames the pursuit of advanced AI as “a wider and deeply human endeavor to improve our lives and future prospects”. The team will focus on specific, tangible applications that demonstrate AI’s potential to solve real-world problems: medical diagnostics achieving 85 percent accuracy on difficult cases compared to 20 percent for human doctors, personalized education systems, and clean energy innovations.
Suleyman acknowledges the fundamental tension inherent in this mission. “How are we going to contain, let alone align, a system that is—by design—intended to keep getting smarter than us?” he posed rhetorically. His honest answer: “No AI [control mechanism] has a fully reassuring answer to this question”.
Despite this sobering reality, Suleyman argues that building advanced AI systems with explicit focus on human-centered applications and safety considerations remains the best path forward. The company is making significant investments in responsible AI practices, with recent appointments including Trevor Callaghan, a former legal director at Google and DeepMind, as Vice President of Responsible AI.
The Competitive Landscape: Microsoft Enters Direct Competition
By launching its superintelligence team, Microsoft is entering direct competition with OpenAI, its longtime partner. The move also positions the company against other major AI competitors, including Google, Anthropic, Meta, and Elon Musk’s xAI. Each of these companies has announced its own superintelligence or AGI initiatives, creating what many in the industry view as an intense race for advanced AI capabilities.
Suleyman has attempted to downplay the competitive framing, but the reality is inescapable: Microsoft is now a direct competitor to OpenAI in the frontier AI space, even as the two companies maintain partnership arrangements for cloud services and model deployment.
To succeed, Microsoft is making massive capital and resource commitments. The company is investing heavily in compute capacity through partnerships with Nvidia and is simultaneously developing its own custom AI chips designed specifically for AI model training and inference. These custom chips are a critical component of Microsoft’s broader strategy to reduce reliance on third-party hardware and achieve what Suleyman calls “the most performant infrastructure in the world”.
Self-Sufficiency Through Scale and Distribution
Microsoft’s advantages in the pursuit of superintelligence differ substantially from OpenAI’s. While OpenAI has historically led in pure research capability, Microsoft brings massive scale, distribution channels, and existing customer relationships.
“We have the data, we also have the distribution, and we have the user interface,” Suleyman noted, suggesting that Microsoft’s vast installed base of customers across Windows, Microsoft 365, Azure, and other platforms provides an unparalleled avenue for deploying advanced AI systems once they’re developed.
The company’s $300 billion in annual revenues and existing enterprise relationships mean that even frontier-grade AI models developed by the superintelligence team will have immediate practical applications across Microsoft’s product portfolio. This differs markedly from OpenAI’s approach, which has historically focused on research breakthroughs that are then incorporated into consumer-facing products like ChatGPT.
A Pragmatic Approach to Model Selection
Despite the newly competitive posture, Suleyman has emphasized that Microsoft will maintain a pragmatic, technology-agnostic approach to model development. The company will use whatever models and technologies best serve its products and customers, whether those are open-source offerings, Anthropic’s Claude, OpenAI’s GPT models, or Microsoft’s own MAI models.
“There’s no reason for us to be religious about that,” Suleyman said, signaling that Microsoft won’t be bound by ideology in its model selection strategy. This pragmatism contrasts with some competitors that have committed fully to internal model development at the potential expense of other available technologies.
Timeline and Challenges Ahead
Suleyman has set realistic expectations for the superintelligence team’s development timeline. He expects that it will take “a good year or two” before Microsoft’s internal research team produces frontier-grade models capable of competing with the most advanced systems currently available.
The challenges are substantial. Building frontier-grade AI models requires not only massive computational resources but also access to high-quality training data, innovative research methodologies, and solutions to fundamental problems in AI development. Two areas Suleyman highlighted as particular focuses are transfer learning—enabling AI models to teach other models new knowledge—and continual learning, which allows AI systems to add new knowledge to existing neural networks without forgetting previous learnings.
The Broader Implications
Microsoft’s superintelligence initiative marks a watershed moment in the AI industry. The renegotiated partnership with OpenAI effectively ended Microsoft’s subordinate role in frontier AI development and launched the company into direct competition with all major AI players.
For OpenAI, the shift represents a fundamental change in its relationship with its largest backer. While the partnership will continue in cloud services and model deployment, OpenAI no longer enjoys Microsoft’s exclusive focus or protection from competitive AI development.
For the broader AI industry, Microsoft’s “Humanist Superintelligence” philosophy offers an alternative to what some view as an irresponsible race to build ever-more-capable systems without adequate consideration for safety and alignment. Whether this approach proves more successful than competitors’ methodologies remains to be seen, but it signals Microsoft’s determination to shape the future of AI development according to its own vision and values.
Microsoft’s AI Independence
Mustafa Suleyman’s declaration that “Microsoft needs to be self-sufficient in AI” captures the essence of a historic transformation in the company’s AI strategy. After years of operating within constraints imposed by its OpenAI partnership, Microsoft is now pursuing its own path toward superintelligence.
The company brings immense resources, technological capability, and infrastructure to this pursuit. It brings also a different philosophical approach, emphasizing human control, safety, and real-world applications over abstract capability maximization.
Whether Microsoft’s superintelligence team succeeds in building frontier-grade AI systems that can compete with OpenAI, Google, and other players remains uncertain. What is certain is that the AI landscape has shifted fundamentally, and Microsoft is no longer a supporting player in someone else’s story, but a primary protagonist in the race to define the future of artificial intelligence.
Click Here to subscribe to our newsletters and get the latest updates directly to your inbox