Satya Nadella (Microsoft) warns about the economic impact of AI

The Impact of AI on Businesses According to Satya Nadella

Satya Nadella, the CEO of Microsoft, recently shared his vision on the transformative impact of artificial intelligence on businesses. In a blog post, he emphasized that AI is poised to redefine the very nature of business. He discusses the formation of a "true cognitive cycle" between people and digital systems, a first in the history of enterprises.

Nadella explains that to thrive in this new environment, businesses will need to rely not only on their human capital but also on "token capital." This term refers to the AI capabilities that companies possess and control. He stresses the necessity for businesses to develop proprietary learning systems and conduct private assessments to verify whether their AI models are genuinely improving outcomes relevant to the business.

He specifies that internal training configurations must be designed to enhance AI models with real company data. Institutional knowledge must become queryable and reusable. "The real opportunity lies not in choosing the best model, but in building a learning cycle where human capital and token capital accumulate," writes Nadella. He adds that while one can delegate a task or even a job, learning can never be delegated. The future of businesses lies in their ability to accumulate this learning between people and AI.

The true test, according to Nadella, is whether a company can replace its core model without losing the knowledge it has built on top of it. Microsoft, which is developing its own AI models, is currently lagging behind its competitors. The company is looking to retain its customers through Azure and its tools, particularly with bundled AI offerings related to Office.

The Threat of Concentrated AI Profits

Nadella warns of the risk that a few AI models could capture the majority of economic benefits, leaving entire industries vulnerable. He describes a scenario where a small number of AI systems capture all economic returns, while the knowledge of businesses is commoditized beneath them.

Nadella's discourse has evolved since March 2025, when he claimed that AI models were becoming commodities. Today, he seems less certain of that assertion. "The last thing we want is a world where every company cedes value to a few dominant models," he writes. According to him, if value is captured by a small number of models, it could lead to an unacceptable economic concentration that political economy would not tolerate.

What could rather be commoditized, Nadella warns, is the knowledge of businesses. He emphasizes that if the commoditization of models progresses more slowly than expected, companies like OpenAI and Anthropic could dominate the market with unmatched models. These companies wrap their models in product ecosystems, blurring the distinction between the model's capability and the agent, which could reinforce the concentration that Microsoft typically criticizes.