Microsoft's Nadella and Hegel's Philosophy

Microsoft CEO Satya Nadella has been pouring trillions of yen into AI. A massive investment in OpenAI, a restarted nuclear plant, giant data centers being built around the world. By any conventional measure, this looks like a decisive bid to win the AI era.

But watching all this, I have a different impression.

"This man is, I suspect, misreading the times."

A Word About Hegel

About two hundred years ago in Germany, there was a philosopher named Hegel. He said something interesting.

The people who move history—kings, revolutionaries, CEOs of giant corporations—act out of their own ambitions and passions. "I'll take the throne." "I'll make my company the biggest." They throw themselves into their work with that drive.

But what they don't realize, in Hegel's view, is that history is using them as tools. They think they're going for the win, but in reality they're being assigned the role of "stepping stone" for the transition to a new era.

Hegel called this the "cunning of reason" (List der Vernunft). History cunningly uses the passions of clever people to prepare the next age.

Nadella's Three Bets, and the Whole Picture

Nadella is pursuing three large bets. Each one is unfolding in a way that pulls the entire world along.

Let's walk through the three bets one by one.

The Nadella Case

Here are three specific things Nadella is doing. All three share a defining feature: they don't stop at being Microsoft's story—they pull the entire world along.

First. The $13 billion investment in OpenAI dragged the entire world into an AI development race.

In 2023, Nadella invested $13 billion (about ¥2 trillion) in OpenAI—an extraordinary sum. With it, ChatGPT spread across the world, and the AI boom caught fire.

What happened next? Google, Amazon, Meta, Elon Musk, Chinese tech firms—everyone started making massive AI investments, terrified of being left behind. Capital and resources from around the world surged into AI all at once.

But the very OpenAI that Nadella set ablaze grew into something Microsoft can no longer control. In 2025, OpenAI restructured into a Public Benefit Corporation, and its valuation reached $500 billion. It is no longer Microsoft's junior partner. It has become an equal—or even superior—negotiating counterpart.

As a result, in the April 2026 contract revision:

• Revenue sharing from Microsoft to OpenAI was eliminated
• Microsoft's right to be the exclusive cloud provider for OpenAI expired
• OpenAI became free to use other cloud providers, including Amazon

In other words, Nadella has already lost his leverage over OpenAI.

So now Nadella is developing AI in-house as well. He has unveiled an in-house model called "MAI" and is pivoting away from his all-in-on-OpenAI posture.

So:

• He pulled the world into an AI race,
• lost his grip on his original bet (OpenAI),
• and is now scrambling to start over with in-house development.

This is about as bad as strategy gets. The man who lit the fire is still wandering around the burning building.

Second. Embedding autonomous agents into Windows and Office.

Since 2024, Nadella has been repeating one direction: "turn Copilot into an autonomous agent."

An "autonomous agent" is a system in which the AI judges and acts on its own, without needing the human to issue every command. For example, if you ask it to "schedule a meeting," it reads emails, checks availability with the other party, puts it on the calendar, and sends out confirmations—that kind of AI.

He is embedding this Copilot into Windows, Word, Excel, Outlook, Teams, and even Notepad and Paint—into a large portion of Microsoft's existing product lineup. The goal: a world where Microsoft's products do work on the user's behalf, on their own.

And, like the first bet, this didn't stop at one company.

After Microsoft embedded Copilot across its products, Google (Workspace), Adobe, Salesforce, Notion—software companies of every stripe began bolting AI agents onto their own services. AI is being loaded onto software around the world, one after another.

And the result is already in.

Among users, Microsoft's products are now being called "Microslop." "Slop" in English means dirty water or scraps of leftover food. The nickname expresses users' biting contempt: "Microsoft's products have been turned into slop by AI bolted on top of them, and now everything is a mess."

• Boot up Windows and Copilot keeps pushing itself in your face, uninvited
• Try to reply in Outlook and the AI starts suggesting things on its own
• Write in Word and the AI tries to rewrite what you've written
• Turn the settings off, and the feature comes back with every update

Users feel the products have become not more convenient, but more intrusive. Because the goal is "an agent that judges and acts on its own," from the user's side it shows up as "moves on its own" and "can't be controlled." This is the heart of Microslop.

The user backlash hasn't stopped at complaints. Users have been coordinating searches to make "Microslop" appear in search engine autocomplete, and after Microslop-themed memes flooded Microsoft's own Discord community, Microsoft locked the community down.

And Microsoft itself has already begun to retreat. It has quietly slowed the Copilot rollout and is pulling AI features back from apps like Notepad. The grand plan to embed AI into every product is starting to collapse from the inside.

Third. The giant data center build-out is starting to seize the world's power and water.

To centralize all AI on Azure (its own cloud), Nadella is building giant data centers around the world.

This too connects directly to the autonomous agent vision. Agents keep running in the background, even when the user is doing nothing. Running an agent for every user, 24 hours a day, requires a different order of computing resources—electricity, water, and semiconductors at a scale that dwarfs anything before.

And Nadella has now turned on nuclear power.

Three Mile Island, the site of the worst meltdown in American history in 1979. Unit 1 of that plant, which had been shut down for economic reasons in 2019, is being restarted with a $1.6 billion investment under a 20-year electricity purchase agreement with Microsoft. On top of that, Nadella joined with then-President Trump to announce the "Stargate" data center build-out at a scale of $500 billion.

And again, it didn't stop at one company.

Watching this, Google, Amazon, and Meta have entered their own race to build nuclear-backed giant data centers. National governments, in the name of "AI sovereignty," have started trying to attract data centers to their own territories. A worldwide scramble for power, water, and land has begun.

The result: society at large is starting to absorb the burden and the risk.

Microsoft had loudly declared itself "carbon negative" (a net reducer of greenhouse gases). But because of the AI infrastructure build-out, the company itself now admits that its greenhouse gas emissions in 2024 were 29.1% higher than the 2020 baseline. For Google, the figure is 48%.

Resources that should be going to food production, healthcare, housing, or climate response are instead being pulled into giant data centers.

And what's more serious is the fragility of concentration.

When a giant data center stops, every operation that depends on it stops with it. An accident, a disaster, a cyberattack, a war—any one of these can paralyze, in one stroke, the companies, government services, hospitals, and banks that use that data center.

In recent years alone, AWS and Azure outages have repeatedly taken down dependent services for hours at a time. A single misconfiguration at AWS once took out banks, government agencies, railways—half the internet. Once AI agents are embedded into every corner of business, the impact of a data center going down will be orders of magnitude greater.

"Concentrated for the sake of efficiency"—and the very fact of concentration becomes the greatest single point of failure. This is the real risk Nadella's strategy is producing.

The Problem Is "Making It Giant"

To be clear: this is not an argument against autonomous agents. Agents running on a personal computer, on a company's own server, or embedded in equipment at a worksite—those distributed forms wouldn't consume much electricity or water, and if one stops, the whole system keeps running. Data doesn't have to leave the user's hands, so privacy is preserved.

The problem is that Nadella chose the direction of "gathering the world's agents into giant data centers and running them there." It is "making it giant" and "concentrating it in one place" that are producing serious problems.

What Hegel Would Say

There is actually one more important premise here.

Hegel's "cunning of reason" only works as an analytical tool when the era itself is in transition. In normal times, clever people simply act cleverly and produce clever results—no paradox arises. A qualitative change in the era is precisely what makes possible the structure in which "those who act rationally according to the logic of the previous era end up playing the role of closing it out."

In other words, the very fact that Hegel fits Nadella's strategy shows that this is a revolutionary moment.

A revolution in what? All of society is groping for how to absorb the new power of AI—that is the revolution underway right now.

AI is something humanity has never encountered at this scale. For society to receive it safely and healthily, several conditions are needed. It must not be monopolized by a single company or a single state. Individuals must keep control. Data must be protected as belonging to the individual. It must be physically sustainable. It must not collapse in a single point of concentration. These are the conditions.

Society is moving toward these conditions unconsciously. Nations are starting to assert AI sovereignty, attention is shifting to small AIs that run locally, open-source models are spreading. These are society's adaptive reactions to receiving AI.

What Nadella is doing is the exact opposite of this adaptive reaction. Treat AI as a single company's product, gather it on Azure, embed it in Office, charge a monthly fee. He is trying to apply the logic of the previous era's IT industry to AI. But AI is not like the previous era's products. Precisely because it touches all of society so deeply, society cannot accept its monopoly by a single company.

And there is one more serious thing. Nadella's business model is premised on "users who want to be automated"—people who want to hand work off to Copilot, who want to delegate judgment to AI, who pay monthly fees to "have it done for them."

But the revolution of the AI era is moving in the direction of eliminating exactly that type of user. As AI spreads, "work that only consists of being automated" stops producing value. What remains is people who use AI to do new things—people who think for themselves, combine their own tools, and act on their own judgment. They do not need an "automation service" like Copilot.

In other words, the more Nadella invests in AI, the more he destroys his own customer base. The more he sells automation, the fewer buyers of automation remain. The users who want to be automated will exit. So Nadella, too, will exit with them.

And the change will come faster than most expect.

Because automation itself is producing new problems. Problems that would not exist without automation come into being precisely because of it.

AI's automatic updates stop systems around the world all at once. Code written by AI flows into production without human review and scatters bugs. The feature of "automatically capturing the screen" creates data leak risks. Automatic concentration in the cloud paralyzes the whole world in a single point of failure. These are problems that would not exist without automation.

And what Nadella is trying to build is a world where everything depends on a single Azure. The larger the scale of automation, the larger the scale of the problems automation creates.

Accidents are already happening. A single AWS error has taken down half the internet. A single CrowdStrike auto-update has paralyzed aviation, healthcare, and finance worldwide. Recall's automatic screen capture has spread concern over user data sovereignty. Problems that automated systems produce precisely because they are automated—this is already empirically demonstrated.

Historical turning points are always accelerated by "accidents." As Three Mile Island changed nuclear policy, Lehman changed financial regulation, Fukushima changed energy policy, and COVID-19 accelerated supply-chain diversification—a large-scale accident triggered by AI automation will accelerate society's reorientation all at once.

And someone like Nadella, who has pushed automation to the limit, will become the symbol of that accident. The larger his strategy, the more violent society's reaction when the accident happens. The change will not be gradual. Triggered by an accident, it will proceed all at once.

That is why, I think, Hegel would put it like this:

"Nadella is acting in the belief that he is becoming the victor of the AI era. But history is using him to push the logic of the previous era to its absolute limit—so that everyone can see that this logic was wrong."

He thinks he is going for the win. In fact, he is playing the role of "the one who closes out the previous era." That is what Hegel meant by the cunning of reason. It can also be read this way: for society to receive AI healthily, a large-scale failure case like Nadella's was needed.

Tellingly, Bill Gates, Microsoft's co-founder, in 2024 sold off his foundation's entire holding of approximately $3.2 billion in Microsoft stock—at exactly the moment when AI infrastructure investments were driving the share price upward.

The founder himself is quietly stepping away from the empire Nadella is building. That, too, may be a sign of the turning point.

Appendix 1: Nadella's Major Investments

Nadella became CEO in 2014. In the 12 years since, Microsoft has spent over $130 billion in acquisitions alone. R&D spending has more than tripled, from $11.3 billion in 2014 to $34.3 billion in 2026.

In other words, Nadella has not "neglected technology investment." He has invested at an extraordinary scale.

The problem is the direction of that investment. Every investment is aimed at "lock-in"—keeping the user inside Microsoft.

Investment / Acquisition	Amount	What it locks in
LinkedIn (2016)	$26.2B	Gathers business-user data into Microsoft
GitHub (2018)	$7.5B	Pulls the world's developers into Microsoft's ecosystem
Activision (2022)	$68.7B	Locks gamer time inside Xbox / Game Pass
OpenAI investment (2023–)	$13B	Exclusive contract keeping OpenAI on Azure
Copilot embedding	(R&D)	Makes it harder for existing users to switch away
Giant data centers	~$500B scale	Concentrates all processing on Azure
In-house chips Maia / Cobalt	(R&D)	Usable only by Microsoft and OpenAI, not sold externally (Apple Silicon, Google TPU, Amazon Graviton are all offered externally)

Appendix 2: Microsoft's Current State in Five Key Areas for a Tech Company

Area	Current state
OS	Windows 11 changes features yearly, with unclear support windows. It enforces a new mechanism called TPM 2.0, cutting off older PCs. The new Recall feature records user activity without explicit consent, raising data-sovereignty concerns. Furthermore Copilot is embedded at the OS level and is difficult to remove or disable—the Copilot button in the taskbar, the Windows-key+C shortcut, Copilot entries in the context menu, all appear everywhere whether users want them or not, and turning settings off only delays the next update bringing them back. Copilot is being designed to become a required Windows feature in the near future, and the option of "Windows without Copilot" is likely to disappear—features usable only on Copilot+ PCs keep increasing, and older PCs will be cut off in the same pattern as TPM 2.0
App distribution	Microsoft Store has effectively failed. Impersonation apps and fakes are rampant, and developers have no rational reason to distribute through Microsoft Store. Furthermore Windows Update silently installs AI features like Copilot—apps users did not choose are added automatically by OS updates and reinstall themselves when deleted. Once Copilot becomes required in the OS, there will be no "install or not" choice left
Development tools	VS Code (released 2015, developed by Erich Gamma's team from existing technology) combined with GitHub (acquired in 2018 for $7.5B) gives Microsoft a hold on the world's developer workflow. However, neither was newly invented under Nadella—both are existing achievements absorbed into Microsoft. The .NET sold to external developers is forced through a rewrite every three years, imposing heavy maintenance on enterprise developers. In addition, WinUI, the latest GUI framework for native Windows apps, has shifted direction multiple times (UWP → WinUI 2 → WinUI 3), still lacks standard controls (like DataGrid) after years, has sparse documentation and samples, and runs only on Windows—not Mac, Linux, or the web—so Microsoft itself does not use it. Microsoft tells external developers "build Windows apps in .NET and WinUI" while not using them itself. External developers have no reason to use them either
Own hardware	Surface lags behind MacBook and iPad in performance, battery, and price. Microsoft has no consumer-grade chip at the level of Apple Silicon. Copilot+ PCs also depend on third-party silicon and have no in-house core. HoloLens has been retired; Xbox is being pushed back by Sony, Nintendo, and Steam Deck. The in-house chips (Maia, Cobalt) exist, but unlike Apple Silicon, Google TPU, and Amazon Graviton, they are not sold externally—usable only by Microsoft and OpenAI, a chip purely for lock-in
Own apps and flagship products	The flagship products are all 30–40 years old—Word (1983), Excel (1985), PowerPoint (1987), SQL Server (1989), Outlook (1997), Visual Studio (1997). Even the "hit products" credited to Nadella's era were not invented from scratch—Teams is a Slack follower built on top of Skype for Business, VS Code came from the Zurich team carrying over existing patterns, GitHub Copilot is the acquired GitHub with OpenAI's technology layered on. Moreover the surfaces users touch (the frontends) are not built on Microsoft's own development tools (.NET, Visual Studio, WinUI)—old Office is in C++, while the rebuilt Outlook, Loop, Teams, VS Code, Designer, and Copilot UIs are all in Web technology (JavaScript/TypeScript-based Electron or WebView2)