Restructuring for Independence
In a wide-ranging discussion, Microsoft AI chief Mustafa Suleyman described how the company has spent the past eighteen months reorienting its artificial intelligence efforts. The goal is to build frontier-level systems internally rather than relying solely on external partners. This adjustment follows a new commercial agreement with OpenAI completed last October that preserved access to existing models while explicitly allowing Microsoft to pursue superintelligence on its own.
Suleyman explained that the internal reorganization created a dedicated Superintelligence team and the necessary compute clusters. The change enabled focused work on next-generation models across modalities. Seven new systems were unveiled at the recent Build conference, reflecting progress on that independent path. The emphasis remains on long-term sustainability instead of short-term dependency on any single external source of intellectual property.
I mean, superintelligence is coming. I think it’s just around the corner. And so I think it’s going to be basically the most valuable technology of all time. There’s sort of no way that, long-term, we could be structurally dependent on a third party for providing that IP for all eternity.
Enterprise Focus Amid Consumer Skepticism
Suleyman acknowledged growing public unease with AI, particularly among younger users who report increasing antipathy despite heavier usage. He argued that the industry has yet to deliver consumer products whose benefits clearly outweigh concerns over data use and infrastructure demands. In contrast, enterprise adoption shows clearer product-market fit because companies already control the data and repeatable processes that models can optimize.
Microsoft’s distribution advantage in enterprise software remains unmatched, with nearly all Fortune 500 companies relying on its cloud and productivity platforms. Suleyman stressed that this reach obliges the company to develop its own high-quality models tailored to business needs rather than simply reselling third-party systems. The strategy prioritizes measurable productivity gains in coding, analysis, and workflow automation over speculative consumer features.
Defining Superintelligence and Its Timeline
The conversation clarified distinctions among artificial general intelligence, superintelligence, and the singularity. AGI represents parity with average human performance across most tasks. Superintelligence goes further by exceeding that performance and generating novel knowledge independently. The singularity, in Suleyman’s view, lies decades ahead and involves recursive self-improvement at an accelerating rate.
He noted that current transformer-based architectures continue to deliver gains through additional compute and data, yet further breakthroughs will likely be required to reach true superintelligence. Microsoft’s new MAI-Thinking-1 model demonstrates progress in reasoning benchmarks without relying on distillation from other labs, underscoring the commitment to original research. The company also released specialized models for transcription, imaging, and code that rank among the strongest available.
We want AIs to be controllable, contained, accountable, aligned tools that serve humanity. That’s the project of humanist superintelligence.
Governance and Public Permission
Suleyman repeatedly returned to the need for social license. He pointed out that Microsoft has maintained its net-zero commitments and designed new data centers for liquid cooling and renewable power. Local communities affected by increased electricity demand are to receive compensation to avoid price spikes. These measures reflect an understanding that infrastructure expansion requires ongoing public consent.
The executive framed the broader debate as one of purpose: technology should make people healthier, smarter, and more capable. When systems fail that test, resistance is predictable and justified. Microsoft’s recent partnership with Mayo Clinic to co-train a health-focused foundation model illustrates the preferred direction—applying advanced AI to domains with clear societal benefit while remaining attentive to accountability and oversight.
