For decades, companies have accumulated capital and talent. Satya Nadella Thinks They Need Something Else Now: Token Capital

The reflections of the CEOs of large companies must always be taken with some caution since they are not usually simple reflections thrown into the air. They seek to give someone their ear: investors, rivals, users or to calm the waters among its employees.

Satya Nadella, has published in their profiles of social networks a text in which he redefined, without much dissimulation, what a company should be in the era of artificial intelligence. It was not just a statement, but rather it almost pointed to become a manifesto in which the focus of the race was changed by develop the best AI model (an area where Microsoft seems to accept the Copilot’s defeat) and defines that the future of companies involves generating a third pillar to their structure: token capital.

The company of the future needs a new type of capital: token capital. Nadella does not conceive AI as a substitute for human employees since, according to his thesis, human capital is the basis for turning AI into a truly disruptive tool that is nourished by knowledgethe criteria, the relationships and the reading of patterns that employees provide. However, it brings a new element to the business equation: token capital. This new element is formed by the AI ​​that a company builds and owns, not in the models that third companies rent to it.

Currently, most companies use AI as a subscription service in which they pay for a model that they use to perform tasks with it. However, when they stop paying for it, all that knowledge and evolution is lost and the company retains nothing of all the knowledge. time and resources you have invested in fine-tuning its use. Nadella maintains that this path leads to transferring value to a few suppliers and the only ones who accumulate advantage are those who sell the models, the knowledge of each company ends up being the raw material that feeds others.

The loop that becomes active. The Microsoft CEO’s idea revolves around what he calls a “learning loop”: a system that feeds back with each decision made and each workflow completed. That is, it is a knowledge base that makes the company’s memory permanent and not lost when changing the AI ​​model or employees. “This cycle becomes the company’s new intellectual property,” highlights the Microsoft CEO. “I look at it as a hill-climbing machine.” The key is that this asset, unlike what happens today, is evolutionary and is built based on training with real company data and internal measurements of its response.

The more you use it and tune it, the more value it has. And, Nadella argues, the company that builds it will soon have something that can’t be bought in any AI model marketplace: a tool that has been “trained” to do a very specific job in a custom context.

AI as a tool, not a monopoly. There is a paragraph in Nadella’s statement that is striking coming from the CEO of a company valued at three trillion dollars. Nadella compares the current risk with what happened in the first phase of globalization: entire industrial sectors were emptied by outsourcing. The macroeconomic figures of the countries they endured the loss of industrial fabric, but the social fabric ended up suffering by adding tension in the labor market.

His warning leaves no room for interpretation: “If all value is concentrated in a few models, political economy simply will not tolerate it. There is no social permission for an AI future that destroys entire industries.” The goal, he says, has to be an ecosystem where each company can build its own learning, not be another cog in an AI monopoly. Actually, this is not new, since it is the same principle with which Microsoft built its platform business in Azure cloudwhich used Microsoft’s infrastructure for companies to generate more value than the platform itself had.

The problem that the manifesto does not solve. However, Nadella’s words also raise a series of contradictions with respect to the latest movements of Microsoft and other large technology companies. The CEO maintains that human capital becomes essential as token capital grows since it is the employees who make a company’s AI learn.

However, his own company has been half doing the opposite. Microsoft fired to more than 15,000 employees during 2025, and in April 2026 it offered voluntary retirement packages to some 8,750 workers in the US, something it had not done in its 51-year history, linking these layoffs to your commitment to AI. It is not an exclusive case of Microsoft. In the first quarter of 2026 they are already more than 92,000 layoffs among the employees of large technology companies and the argument that all companies repeat is the same: AI allows us to do more with less people.

In Xataka | Jensen Huang enters the Samsung salary controversy: “Workers should earn as much as possible”