Six months ago, Microsoft created MAI. In that time, it has launched three foundation models—and its CEO is now openly chasing superintelligence. Something does not add up. If Microsoft's $13 billion OpenAI partnership was meant to be a permanent moat, why is the company simultaneously building the very models that could make that partnership obsolete?
The contradiction is not accidental. Mustafa Suleyman, Microsoft's inaugural CEO of AI, tells The Verge he had been planning the pivot for nine months—before the ink dried on the renegotiated OpenAI contract that officially "unlocked" Microsoft's ability to pursue superintelligence. The restructuring in mid-March, which stripped away some of his operational duties, was the final move in a longer game. Microsoft is no longer merely invested in AI. It is betting the company on arriving at AGI first.
The three foundation models MAI released this week—capable of speech-to-text transcription, audio generation, and image synthesis—represent the first concrete evidence of that ambition. They are not incremental improvements to Copilot. They are foundation models designed to compete directly with what OpenAI, Google, and Anthropic offer. Six months from zero to market is an aggressive pace that suggests Microsoft has shifted from cautious investor to active competitor.
The implications for the OpenAI relationship are murky. The original partnership was premised on Microsoft providing infrastructure and distribution while OpenAI provided frontier models. That arrangement is now under strain. Microsoft has its own models. Suleyman is talking openly about superintelligence—defined as systems that exceed human cognitive performance across most domains. If that goal is achievable within a meaningful timeframe, the calculus changes: why share the prize with a partner when you can capture it yourself?
The risk is substantial. Competing at the frontier requires capital that even Microsoft's scale finds daunting. The renegotiated contract may have given Microsoft more freedom, but it also means OpenAI's incentive to share breakthroughs has diminished. Two organizations that were aligned are now running parallel races toward the same finish line.
For enterprises evaluating AI vendors, the uncertainty matters. Microsoft's foundation models compete on capabilities, but the company's strategic direction remains in flux. Suleyman's superintelligence pivot signals confidence in rapid progress—and that confidence cuts both ways. If he is right, Microsoft will be among the companies that define the next era of computing. If he is wrong, the company will have spent enormous resources competing in a race that may not be won by any single player.
What happens next is straightforward: Microsoft will push MAI's models into Azure, into Copilot, into enterprise contracts. The OpenAI partnership will continue to exist on paper while both parties quietly acknowledge they are no longer the same kind of allies. Suleyman will keep talking about superintelligence. The tension between partnership and competition will sharpen until one side or the other breaks.
The smart money is watching whether Microsoft's models can close the gap with OpenAI's GPT series. If they can, the restructuring was not just a pivot—it was a declaration of intent.