Microsoft Strives for "Self-Sufficiency": Developing In-House Top Models to Gradually Reduce Dependence on OpenAI

IT Home, February 12 — This afternoon, according to the Financial Times of the UK, Mustafa Suleyman, head of Microsoft’s AI division, stated that Microsoft is gradually reducing its dependence on OpenAI by developing its own top-tier models, aiming for true self-sufficiency.

In an interview with the Financial Times, Suleyman said that the restructuring of their relationship with OpenAI in October prompted Microsoft to reorient its technological strategy. As a tech giant with a market value of $3 trillion (Note: approximately 20.75 trillion RMB at current exchange rates), Microsoft has decided to independently develop the most advanced AI technologies instead of relying entirely on external partners. “We have gigawatt-level computing power and world-class AI training teams, and we must build cutting-edge foundational models.”

Microsoft is heavily investing in collecting and organizing data resources to support advanced model training. “This is what true self-sufficiency means for Microsoft.”

As one of the earliest and largest investors in OpenAI, Microsoft previously relied on OpenAI models to provide core capabilities for its AI products. Last year, Microsoft agreed to a corporate restructuring with OpenAI, retaining a stake valued at $135 billion (about 933.972 billion RMB at current rates) and securing rights to use advanced models until 2032.

The agreement also grants OpenAI more freedom in financing and infrastructure collaborations, potentially leading to more direct competition with Microsoft in the future.

Microsoft is also diversifying risk by investing in model companies like Anthropic and Mistral, while accelerating internal model development. Suleyman stated that self-developed models are expected to be launched this year.

Microsoft plans to invest $140 billion in capital expenditures during the fiscal year ending June, focusing on AI infrastructure development. The substantial investment has raised concerns about an AI “bubble,” with Microsoft’s stock price dropping over 13% in the past month.

Suleyman said, “The current situation is unprecedented, and the market is assessing the next five years. However, everyone believes that these investments will ultimately translate into revenue and profit growth.”

Currently, Microsoft is promoting AI applications in healthcare, aiming to create a “superintelligent healthcare system” to address staff shortages and long waiting times. Last year, Microsoft released an AI diagnostic tool, claiming it outperforms doctors in certain tasks.

Suleyman emphasized that Microsoft’s goal is “human-centered superintelligence,” AI technology always under human control, to address concerns about laboratories rushing to develop systems that are difficult to regulate.

He stated, “Only when we are confident that we can control and ensure the system remains subordinate should we introduce such systems into reality. Like previous technologies, these tools aim to enhance human well-being and serve humanity, not to dominate it.”