For three years, Microsoft’s AI identity was basically a pronoun for OpenAI. The company poured more than $13 billion into Sam Altman’s outfit, got early access to the most capable models on the planet, and rode that advantage to hundreds of billions in added market cap.
Now Microsoft has rewritten the terms of that relationship. A series of contract amendments beginning in October 2025 have given the company something it didn’t previously have: the right to build frontier AI systems on its own, without OpenAI’s blessing or involvement. Mustafa Suleyman, who took over as CEO of Microsoft AI in March 2024, has been the architect of this pivot.
The deal that changed everything
The restructured agreement, confirmed at Microsoft Build 2026 in June 2026, fundamentally alters that dynamic. The licensing arrangement is no longer exclusive. Revenue-sharing terms have been renegotiated. And critically, Microsoft now has contractual freedom to develop its own superintelligence-class systems using proprietary research, data, and custom silicon.
Suleyman has described the renegotiation as “liberating” for Microsoft. The partnership extends until at least 2032.
Seven models and a superintelligence team
Suleyman announced the formation of the MAI Superintelligence Team on November 6, 2025, with a stated goal of creating what he calls “humanist superintelligence,” or HSI. The team’s ambition is to produce Microsoft’s own superintelligence models by 2026.
At Build 2026, the company confirmed it has seven in-house models under active development. Two of the named projects, MAI-Transcribe-1 and MAI-Voice-1, suggest Microsoft is targeting specific high-value application layers rather than just building another general-purpose chatbot.
Suleyman’s background makes this move legible. Before Microsoft, he co-founded DeepMind, which Google acquired in 2014.
Why crypto should pay attention
The AI-crypto intersection has become one of the most active investment themes in digital assets. Tokens tied to AI infrastructure, decentralized compute, and machine learning marketplaces have attracted billions in capital over the past two years. Projects building decentralized alternatives to centralized AI providers, think Render, Akash, Bittensor, and others, position themselves as counterweights to exactly the kind of consolidation Microsoft represents.
Microsoft’s stake in OpenAI was valued at roughly $135 billion following the October 2025 recapitalization round. The fact that Microsoft is now hedging that bet by building independently suggests even the company’s own leadership doesn’t view a single-provider strategy as sufficient.
What investors should watch
Suleyman has set a 2026 target for releasing Microsoft’s own superintelligence models. For crypto markets specifically, the key variable is compute demand. Every new entrant in the frontier AI race needs enormous quantities of training compute. Microsoft, Google, Meta, and now xAI are all competing for GPU capacity and data center buildouts.
The revenue-sharing changes are worth monitoring too. If OpenAI’s economics shift as Microsoft redirects resources toward in-house models, it could affect OpenAI’s ability to fund its own research at current levels.