In the Era of Profitable Layoffs: Microsoft CEO Who Laid Off 15,000 People Discusses "Human Capital"

Management Illusions, Collapsing Organizations: Part 5

A few days after seeing the news that Meta, alongside its massive AI investments, was mobilizing part of its engineering organization as a "source for generating AI training data," an op-ed by Microsoft CEO Satya Nadella caught my eye.

These were words penned by the head of Microsoft himself, at a time when massive AI investments were putting pressure on profits. I was interested in the implications of how a leader of one of the world's most prominent companies thinks about the concept of experiential knowledge.

The implications were certainly there, but they left me with three questions.

The Essence of the Op-Ed: Human Capital and Token Capital

In mid-June 2026, Microsoft CEO Satya Nadella published a long-form op-ed on X (formerly Twitter) titled "A frontier without an ecosystem is not stable."

His argument is that companies now need to accumulate two types of capital. The first is "human capital," consisting of employees' knowledge, judgment, relationships, and creativity. He claims that even as AI evolves, its value will not diminish, but rather increase. The second is "token capital," consisting of accumulated workflows and organization-specific judgments. He argues that owning the "learning loop" on top of general-purpose AI models, rather than the models themselves, is what will become a new asset that grows through compound interest.

In June 2026, when this op-ed was published, Microsoft stock was in a historic downturn. On June 25, it hit a 52-week low of approximately $349.20, a decline of over 20% year-to-date. It was a time when the market had begun to cast a skeptical eye on the company's massive annual AI capital expenditure of approximately $190 billion.

Read as the words of a CEO, it makes sense. He is justifying the massive investment through the logic of interdependence between humans and AI. Up to this point, it was a plausible story.

Can Experience Really Be Transferred? The Double Wall and the Shadow of Judgment

So, can AI really replace experience?

My honest impression is that Nadella's argument intentionally mixes what "can" be done with what "cannot."

Let me break down the work I have been involved in into three layers. The first is the codified procedures, SOPs, and regulatory requirements themselves. The second is the pattern of judgment regarding "where and to what extent to apply those procedures"—the intuition of knowing which risks are acceptable and which deviations cannot be explained to authorities. The third is the unspoken intuition or the sense tied to memories of past failures that explains why a particular judgment was reached.

What Nadella calls "token capital = learning loop" can realistically capture only the first layer and part of the second. AI agents execute tasks, score the results in an evaluation environment, and accumulate successful patterns—this mechanism can statistically approximate the "input-output correspondence" of judgment. If enough cases are stored, superficial reproduction becomes possible.

However, the third layer—the substance of "why that judgment was made"—is, in principle, difficult for others, including AI, to grasp. When anticipating trouble based on past experience, that judgment is not the application of explicit rules, but an intuition derived from the totality of embodied experience. If one tries to reproduce this, AI may be able to "imitate the correct answer," but it does so without understanding "why it is the correct answer." Therefore, when faced with a new type of deviation not found in the training data—the very exception where quality assurance provides the most value—the imitation fails.

This is the first wall: the wall that cannot be transferred even if it is observed. And there is another wall even further ahead.

A significant portion of a veteran's value actually lies in their inaction—the times they "deliberately chose not to act" or "sensed danger and stayed away." Inaction is not observed. Therefore, action-based learning loops can only capture the contours of experience. This is the second wall: the wall that is not observed in the first place.

Perhaps what is happening in the third layer is not the transfer of information at all. When you encounter someone else's experience, their judgment is not transplanted into you as is; rather, something else rooted in your own experience begins to stir within you—it is not a copy, but an inspiration. What the learning loop collects are the traces of actions—"what was done"—not the substance behind the judgment—"why it was done" or "why they deliberately chose not to act"—which is what triggers the inspiration.

This is, of course, only about the current level of technology. A mechanism where AI, through dialogue with the individual, helps unearth the basis for judgments that even the person themselves hadn't realized is, in principle, possible. Whether this wall will lower in the future is something no one knows yet.

Why did Nadella dare to write this?

In the same year he argued that human capital is precious, the same company let go of 15,000 people while remaining profitable.

In 2025, Microsoft cut over 15,000 jobs while its business performance was strong. While the simultaneity of these two facts can be demonstrated, it cannot be definitively stated that this is a direct cause-and-effect relationship. Even so, faced with this gap, one cannot help but ask for whom the philosophy of 'turning experience into an asset' is intended.

So, why did Nadella dare to write this essay?

First, let's look at the platform itself. Mr. Nadella rarely uses X for daily updates. His posts are primarily focused on key moments linked to earnings announcements or conferences. If the message were intended for internal engineers, he should have used internal meetings or memos. The fact that he chose X, which is open to the public, indicates that it was not intended for an internal audience.

Furthermore, this essay was a long-form piece of several thousand words. Some media outlets have reported that the number of views after publication exceeded 28 million. Judging from Nadella's conventional posting patterns, this can be considered an exceptional, deliberately chosen communication. Moreover, the timing of the publication was in the midst of a historic decline in Microsoft's stock price, with a shareholder class-action lawsuit filed questioning the slowdown in Azure's growth and excessive capital investment.

The remaining candidates are the market/investors or industry executives. These two overlap. One analysis of the essay points out that the argument that 'models are commodities and replaceable' could be seen as effectively neutralizing the pricing power of competing AI model companies and strengthening the monopoly of their own cloud infrastructure. If so, this essay is not only an explanation of the 'true ROI of AI investment' to investors, but also guidance to corporate executives to 'avoid dependence on frontier models and build your own learning loop.'

Thinking about it this way, isn't this essay calling out not to the veterans themselves, but to the executives and organizations that are in a position to recognize, accept, and utilize veterans? The value of experience is determined not by those who possess it, but by those who recognize it. If so, this essay might have been a wake-up call directed at those who do the recognizing.

Why Nadella wrote this is something only he knows. However, the fact that the scope can be narrowed down by process of elimination is itself grounds for reading this essay not as a mere statement of philosophy, but as a communication designed to reach multiple audiences simultaneously. And—even without knowing why he wrote it—the fact remains that this essay 're-questioned the importance of recognizing experience to those who do the recognizing.'

Three Questions

I would like to summarize the three questions once again.

I understand the gist of the essay. So, can experience really be transferred? Why did Nadella dare to write this? And for whom is this assetization intended? These are the three questions.

The outline of the first question has become visible up to this point. Accumulating two types of capital—human capital and token capital—and owning the learning loop in-house makes sense logically.

However, what could be transferred was only the shadow of judgment. Even if codified procedures or repeatedly observable patterns can be reproduced, the substance of why a decision was made cannot be transferred to anyone at this time.

And the second question—why Nadella wrote this—is something only he knows. There is the simplest possibility that Nadella truly believes in the scarcity of human intent. However, there is one more thing I would like to add to this reservation. The final question—for whom is this assetization intended?

What Nadella wants to sell is Azure, the foundation for companies to own their learning loops. The more he extols the value of humans, the more his company's products, as vessels for storing that value, will sell. If so, there remains room to read this essay as commercial position-taking.

Furthermore, it is possible that Nadella's argument and the 'rejuvenation' of Japanese companies are not actually contradictory. They might even be two sides of the same coin. Japanese companies reducing the number of veterans is a move to view human experience as a cost and cut it. It is not impossible to view what Nadella is saying as a move to suck up human experience as an asset, transfer it to a system, and then let the humans go. The former discards experience, and the latter squeezes it out before discarding it. The destination—an organization without veterans—might be the same.

Experience cannot be transferred at this time. If so, does experience have any meaning? Next time, I would like to return to my own experience of writing this series and face this question.
Part 6 (Afterword): Accumulated Experience Is Not Valued—The Value of Experience Only Exists When There Is Someone to Recognize It

Thank you for reading to the end.
Please check the entire series here.

References

Nadella, S. (2026). A frontier without an ecosystem is not stable. X. Reports on the core concepts of human capital, token capital, and learning loops.https://x.com/satyanadella/article/2066182223213293753

Barchart (2026). Microsoft Stock Hitting New Lows Could Be a Long-Term Opportunity. Reports the 52-week low of approximately $349.20 on June 25 and the year-to-date decline rate.https://www.barchart.com/story/news/3035369/microsoft-stock-hitting-new-lows-could-be-a-long-term-opportunity

The Next Web (2026). Microsoft shareholder suit over Azure, AI spending. Reports on the shareholder class-action lawsuit (filed June 12, 2026) regarding Azure's slowing growth and AI spending.https://thenextweb.com/news/microsoft-shareholder-suit-azure-ai-spending

CNBC (2025). Microsoft's Satya Nadella says job cuts have been 'weighing heavily' on him. Reports on the reduction of over 15,000 employees in 2025 (May and July) and Nadella's statement.https://www.cnbc.com/amp/2025/07/24/microsoft-satya-nadella-memo-layoffs.html

36Kr (2026). Microsoft CEO's Long Essay: Future to Feature Two Types of Capital - Human Capital and Token Capital. Reports that the essay has been viewed over 28 million times.https://eu.36kr.com/en/p/3853872162165763

ProMarket (2026). Satya Nadella's AI Warning Is a Sales Pitch. Reports an analysis of the 'Commoditize your Complement' strategy.https://www.promarket.org/2026/07/01/satya-nadellas-ai-warning-is-a-sales-pitch/

Polanyi, M. (1966). The Tacit Dimension. University of Chicago Press. The fundamental proposition of tacit knowledge: 'We can know more than we can tell.'