Microsoft's artificial intelligence division has been granted contractual freedom to pursue what CEO Mustafa Suleyman calls "superintelligence" using the company's own researchers and resources, marking a significant strategic shift in its relationship with OpenAI. Speaking at Microsoft Build 2026, Suleyman disclosed that a contractual change with OpenAI approximately six months ago removed restrictions that had previously barred Microsoft from developing its own advanced AI systems.

"We were only sort of set free from our contract with OpenAI about six months ago to formally pursue superintelligence," Suleyman said. The original partnership agreement, which began with Microsoft's investment exceeding $13 billion starting in 2019, had designated OpenAI as the frontier model builder while Microsoft served as the exclusive cloud provider. The deal included caps on how large a model Microsoft could train, measured in computing operations.

The most tangible evidence of this strategic shift came with Microsoft's announcement of seven new AI models developed entirely in-house by its AI Superintelligence Team. The MAI family of models spans reasoning, code generation, image creation, transcription, and voice synthesis. The flagship MAI-Thinking-1 is a 35-billion-parameter reasoning model that Microsoft says matches leading models in its class on software engineering benchmarks. Suleyman emphasized that these models were trained from scratch on commercially licensed data without using outputs from competitors' systems.

Microsoft is also introducing Frontier Tuning, which allows enterprise customers to customize MAI models using their own proprietary data and workflows within secure compliance boundaries. Early results show significant efficiency gains, with one tuned MAI model reportedly matching GPT performance while operating at ten times greater efficiency. Enterprise partners including Mayo Clinic, EY, Land O'Lakes, and Pearson are participating in early implementations.

The company's hardware investments support this independence strategy. Suleyman stated that Microsoft is "the largest buyer of GPUs on the planet" while simultaneously developing its own custom silicon. The company's Maia 200 AI accelerator is already running in production and is reportedly 30 percent more cost-efficient than Nvidia's GB200 chips. When Microsoft optimizes its MAI models to run on Maia silicon, the company sees an additional 1.4x improvement in performance per watt.

Despite this shift toward self-sufficiency, Suleyman emphasized that Microsoft is not abandoning its OpenAI partnership, describing the current situation as offering "abundance, not scarcity." He characterized Microsoft's position as having multiple options through OpenAI, Anthropic, and thousands of models available through its Foundry platform. However, the strategic direction is clear: Microsoft is building capabilities that could eventually operate independently of external AI providers by 2030.