- Microsoft AI CEO Mustafa Suleyman told The Verge on April 2, 2026, that revising Microsoft’s partnership agreement with OpenAI “unlocked [Microsoft’s] ability to pursue superintelligence.”
- A mid-March 2026 organizational restructuring moved Suleyman away from day-to-day Copilot oversight, giving him a mandate to build frontier AI models under the MAI Superintelligence team.
- The October 2025 contract renegotiation extended Microsoft’s IP rights through 2032, removed the exclusive AGI clause, and formally allowed both companies to pursue AGI independently.
- Microsoft’s first public output from the superintelligence team — three foundational models including MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 — launched on April 2, 2026.
Microsoft AI CEO Mustafa Suleyman told The Verge‘s Hayden Field on April 2, 2026, that revising Microsoft’s partnership agreement with OpenAI “unlocked [Microsoft’s] ability to pursue superintelligence.” The interview is the first time Suleyman has publicly described the contract revision as a precondition for Microsoft’s independent AI research ambitions, tying together the October 2025 deal renegotiation, the March 2026 organizational restructuring, and the same-day launch of three new foundational models.
What Happened
The Verge interview came hours after Microsoft announced three new foundational AI models — MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 — developed entirely in-house by the MAI Superintelligence team. Suleyman confirmed he had been preparing for this strategic shift for approximately nine months, with the renegotiated OpenAI contract as the enabling condition.
The Microsoft-OpenAI relationship began in 2019 with a $1 billion investment from Microsoft, which became OpenAI’s exclusive cloud provider. By 2023, Microsoft had committed a total of $13 billion and integrated OpenAI models across its product portfolio. The original agreement contained a clause giving OpenAI’s board unilateral authority to declare the arrival of AGI — a declaration that would have terminated Microsoft’s access to any OpenAI models developed after that point.
That clause created structural uncertainty about Microsoft’s long-term access to frontier AI. In parallel, Microsoft was contractually limited in its ability to build competing frontier large language models, restrictions that remained in place until the October 2025 renegotiation.
Why It Matters
The revised agreement, announced by both companies on October 28, 2025, formally allows Microsoft and OpenAI to “independently pursue AGI alone or in partnership with third parties.” That single clause change removed the contractual barrier Suleyman referenced in the Verge interview.
Microsoft’s IP rights now extend through 2032 and explicitly cover models developed after AGI is declared, closing the gap that had threatened to leave Microsoft without access to future-generation OpenAI systems. The power to declare AGI no longer sits with OpenAI’s board alone — the revised agreement requires verification by an independent expert panel.
For Microsoft, the practical effect is a dual-track strategy: maintain the OpenAI partnership for commercial AI products while simultaneously funding an independent research team pursuing what Suleyman calls “Humanist Superintelligence.”
Technical Details
Suleyman defines Humanist Superintelligence (HSI) as “incredibly advanced AI capabilities that always work for, in service of, people and humanity more generally.” He frames the mission as “not some directionless technological goal” but as purpose-driven AI constrained to specific societal applications in areas such as medicine and climate, according to Microsoft AI’s published position paper.
The MAI Superintelligence team, formed in November 2025 and led jointly by Suleyman and Microsoft AI Chief Scientist Karén Simonyan, produced its first commercial outputs on April 2, 2026. The three models released on that date have concrete performance specifications:
- MAI-Transcribe-1: transcribes speech in 25 languages; operates at 2.5 times the speed of Microsoft’s Azure Fast offering; priced at half the GPU cost of comparable state-of-the-art models.
- MAI-Voice-1: generates 60 seconds of audio per second; supports custom voice creation; priced at $22 per 1 million characters.
- MAI-Image-2: image and video generation model; priced at $5 per 1 million text input tokens and $33 per 1 million image output tokens.
All three are immediately available via Microsoft Foundry and the MAI Playground. Suleyman told The Verge these models represent the opening output of a team he describes as having a five-year mandate to “deliver world class models for Microsoft.”
Who’s Affected
The restructuring formalized on March 17, 2026 created a new Copilot Leadership Team with five EVP-level direct reports to Microsoft CEO Satya Nadella. Former Snap senior vice president Jacob Andreou took over as Executive Vice President of Copilot, unifying Microsoft’s consumer and commercial AI assistant divisions under a single leader for the first time. Ryan Roslansky (CEO of LinkedIn), Perry Clarke, and Charles Lamanna joined Andreou and Suleyman in the new structure, according to CNBC’s March 17 report.
Suleyman retains his title as Executive Vice President and CEO of Microsoft AI but now operates without day-to-day product responsibilities. His scope narrows to frontier model research — a move that signals Microsoft’s intention to treat model development as a separate strategic function from product delivery.
For OpenAI, the October 2025 agreement brought its own structural changes. OpenAI contracted to purchase an incremental $250 billion in Azure services. Microsoft will no longer hold a right of first refusal as OpenAI’s compute provider, and OpenAI can now jointly develop products with third parties, provided API products remain on Azure. OpenAI can also release open-weight models that meet defined capability criteria, a significant expansion of its product autonomy.
What’s Next
Microsoft’s five-year model development mandate under Suleyman runs through approximately 2031. The MAI Superintelligence team has begun commercial model releases, but Suleyman’s stated goal — building AI that operates at a level he calls “Humanist Superintelligence” — requires sustained investment at a scale he has publicly acknowledged as potentially running into “hundreds of billions” of dollars, according to Windows Central’s coverage of his public remarks.
The independent expert panel established to verify AGI declarations has not yet been formed publicly. Its composition and the criteria it will use to evaluate AGI claims remain undefined — the most consequential open question in the Microsoft-OpenAI relationship, given that an AGI declaration would trigger the IP rights transition both companies negotiated in October 2025.
Microsoft’s next stated milestones are commercial availability of the three new MAI models via Azure Foundry and expansion of the MAI Playground. Developers can access MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 now through microsoft.ai.
