Where the Microsoft CEO Is Wrong: People and Companies Could Lose Out as AI Advances

In his recent manifesto, Microsoft CEO Satya Nadella argues that the winners in the AI era will be companies that integrate their own expertise into training cycles—creating “token capital” on top of human capital. According to his logic, value will be distributed widely: every industry and every country will retain its competitive sovereignty. German Kaplun, co-founder of TMT Investments PLC, argues in his column that he agrees with half of these points but disputes the rest.

It’s a nice idea, but I don’t buy it. Technological revolutions all follow the same pattern: powerful monopolization at the top on one hand, and explosive market expansion at the bottom on the other. The internet gave rise to Google, Amazon, and Facebook, while simultaneously creating millions of businesses that simply wouldn’t exist without it. Smartphones have brought Apple and Alphabet hundreds of billions while simultaneously opening up the app market to hundreds of thousands of developers. AI will be an amplified version of this pattern.

A handful of platform systems will capture a disproportionate share of the value. It’s already clear: OpenAI, Anthropic, Google DeepMind, xAI, and Meta AI—this is not a competitive market, but a race with increasing returns to scale, where computing power, data, and talent are concentrated in a few places. The rest will be API consumers, not independent players.

Nadella is right in saying that companies that incorporate expertise into their own training cycles will gain a real competitive advantage. If your model improves based on your own data—becoming smarter with every transaction, diagnosis, or trade—you’re building an asset that can’t be bought from OpenAI.

But how many companies will make it that far? Only a handful will become major players due to competition from existing players, who will be able to quickly occupy adjacent niches. However, narrow segments—such as industry-specific diagnostics, specialized retail, and local jurisdictions—are of no interest to the giants. It is precisely in these segments that a dense layer of small businesses will emerge, having integrated niche expertise into the AI cycle.

The resulting structure will look familiar: many small businesses, a reasonable number of medium-sized companies, and a few major players. What will be new is that entry into each level will require data and a training cycle as a basic asset. This is similar to the globalization of the 1990s, but not in the way Nadella wants it to be.

At that time, GDP was growing, but industrial regions lost their economic base forever. Many cities and regions that had relied on manufacturing lost jobs after factories were relocated to Asia, and they have never fully recovered since then. We are still living with the political consequences of this. AI risks repeating this process, but faster and on a larger scale. Yes, the aggregate figures will look good, but the internal redistribution will be painful. Many people, industries, and regions will become poorer. The process is repeating itself in a new chapter of history. It’s different, but it’s still repeating itself.

Predictions are rarely wrong about the direction of change, but they are often wrong about the pace of change and about who exactly will emerge as the winner.

This article was AI-translated and verified by a human editor