Mustafa Suleyman says Microsoft is building its own AI superintelligence

Microsoft is accelerating its artificial intelligence strategy as it moves toward what it calls “true AI self-sufficiency”. Speaking to the Financial Times, Microsoft AI Chief Mustafa Suleyman said the company is actively reducing reliance on OpenAI while still honouring existing agreements.

Suleyman said his personal mission is to build superintelligence at Microsoft. The company has already taken steps in that direction. Microsoft AI introduced MAI-1-preview as an internal mixture-of-experts model which they trained on 15,000 Nvidia H100 GPUs in August 2025.

Microsoft builds its own AI models.

Microsoft plans to integrate the model into selected Copilot text features. This marks a clear move toward replacing OpenAI systems inside Microsoft products.

At the same time, Microsoft has widened its AI ecosystem. It now hosts models from xAI, Meta, Mistral and Black Forest Labs in its data centres. Reports suggest the company even tested Anthropic’s Claude for Microsoft 365 Copilot tasks, citing stronger performance in certain Office workloads.

In October 2025, Microsoft and OpenAI signed a reworked agreement. Microsoft converted profit-sharing rights into a 27 per cent ownership stake in OpenAI Group PBC, valued at about $135 billion. The deal extended Microsoft’s access to OpenAI intellectual property through 2032.

However, the agreement also gave both companies more independence. OpenAI gained flexibility to seek computing resources beyond Azure. Microsoft secured the right to pursue artificial general intelligence alone or with new partners.

Wall Street has reacted cautiously. Jefferies analyst Brent Thill highlighted OpenAI’s 45 per cent share of Microsoft’s future sales backlog, raising concerns about exposure. Shortly after, Microsoft shares saw a sharp single-day decline.