Microsoft, which once strongly backed OpenAI and ignited the artificial intelligence (AI) boom, is now sending a different signal: the AI race no longer revolves solely around building the most advanced frontier models, but increasingly focuses on making AI models interchangeable and widely accessible.
On June 22, Microsoft CEO Satya Nadella called for a change to the current state of the “AI large-model race dominated by a few companies,” warning that the public will not accept a future where a few models “learn for the whole world,” nor will they accept simultaneous warnings of job losses and unlimited expansion.
In a recent interview, he noted that the next phase of artificial intelligence should not continue to be dominated by a few frontier model companies, but should instead shift toward cheaper models, greater user control, and narratives that can win public trust.
Nadella outlined his vision: businesses and consumers will have greater freedom to choose among various AI systems, rather than relying on a few dominant suppliers.
His remarks come as a growing number of companies seek to reduce their dependence on the most advanced proprietary models developed by OpenAI, Anthropic, and Google. Microsoft itself has launched a series of low-cost AI products aimed at helping customers alleviate the cost pressure caused by rapidly rising AI spending.
In the latest interview, Nadella questioned whether the public would accept a future in which a few institutions control the development of technology while simultaneously warning of mass unemployment and safety risks. He argued that people are unlikely to support a world where only a handful of companies can effectively shape how AI learns and evolves.
It is reported that Microsoft recently launched Copilot Cowork, an autonomous AI agent that allows users to choose from multiple AI models, including some lower-cost alternatives, to perform long-running tasks.
At the same time, Microsoft is reportedly considering introducing a fine-tuned version of DeepSeek V4 into its enterprise AI product Copilot as a low-cost alternative to OpenAI and Anthropic models.
This move could expand DeepSeek’s influence and intensify price pressure on existing AI providers. As is well known, Microsoft has been one of OpenAI’s most important partners and investors, having poured billions of dollars into OpenAI and signed a multi-billion-dollar agreement with Anthropic last year.
Despite Nadella’s criticism of the industry’s direction, Microsoft stated that it remains committed to its partnerships with OpenAI and Anthropic and does not view market competition as a winner-take-all scenario. On the other hand, he also questioned the trend of AI becoming a driving force behind corporate layoffs. He argued that companies should not focus solely on layoffs, but rather rethink how work is organized and how employees and AI systems can complement each other. Enterprises need both AI capabilities and skilled workers, and technology providers must offer practical guidance for corporate transformation. Finally, Nadella emphasized that the long-term value of AI lies in helping companies acquire and apply their institutional knowledge, not merely automating tasks. The AI industry must demonstrate that people can still effectively control this technology and continue to benefit from the economic opportunities it creates. He believes that only in this way can AI companies earn public trust, which is crucial for the technology’s next stage of development.
