An Underestimated Analogy
Many people think the development of AI can be analogized to the invention of the computer. It sounds reasonable—computers changed the world, and AI is changing the world too, so AI is the "next-generation computer."
But Jensen Huang has a perspective I find quite illuminating: this analogy may vastly underestimate the impact of AI.
Why? Consider the numbers first.
The global IT industry is worth roughly 100 trillion. However powerful computers may be, they are fundamentally tools—excellent tools, but tools that serve humans nonetheless. You need people to write code, operate systems, and maintain servers for computers to be useful. After decades of development, the IT industry has essentially operated within that $1 trillion domain.
AI agents are different. As Agent technology matures, it ceases to be merely a tool that assists humans and becomes capable of delivering results independently, without human intervention. Vibe coding is a vivid example—in one afternoon, someone with no frontend knowledge can "chat" a complete website into existence using natural language. This means the AI industry is not competing with traditional IT for that 99 trillion of economic activity—tasks that previously only human labor could accomplish.
This is a qualitative change.
Historical Leaps in Productivity
To understand the magnitude of this "qualitative change," it is worth reviewing the key leaps in productivity throughout human history.
The Agricultural Revolution moved humanity from nomadism to settlement, but its core drivers were land and labor. To produce more food, you needed more land and more people. The ceiling of productivity was locked by natural resources.
The Industrial Revolution introduced machines, allowing capital to "substitute" a portion of human labor for the first time. A single textile machine could match the output of dozens of hand-weavers. But machines still needed people to operate, maintain, and manage them. So the formula of the Industrial Revolution was: capital buys machines, machines amplify human labor, human labor produces goods. Humans remained in the chain, merely shifting from "direct producers" to "machine operators."
The Information Revolution—the invention of the computer—further boosted efficiency, but it was essentially the same model. Excel lets one accountant do the work of five; ERP lets one factory manager coordinate production lines across ten factories. The computer is an excellent amplifier, but it amplifies human capability. Without humans in the middle, computers produce no commercial value on their own.
AI breaks this rule. For the first time, "autonomous output without human intervention" becomes possible. It does not merely amplify human labor; it directly replaces human labor in certain links of the chain. This is not quantitative change; it is qualitative change.
A New Mode of Economic Transformation
Following this logic, you will find a deeper change taking place.
In traditional economic models, converting capital into output requires massive amounts of human labor in the intermediate stages. You have money; you must hire people, train them, manage them, and have them execute before things get made. So the chain of productivity transformation has always been: capital → labor → output.
But AI is rewriting this chain. Once capital is invested in computing power, electricity, and models, it can transform directly into productive force to a considerable degree, without requiring such a large proportion of human labor. This will give rise to an entirely new economic model and production relations—the role of capital is vastly magnified.
This change is already happening. Consider the data: Microsoft's 2025 capital expenditure exceeds 80 billion might need to hire hundreds of thousands of people to operate it. But in the AI era, that $80 billion buys computing power, and computing power becomes productive force directly.
Looking back at history, shocks of this magnitude are rare. In agrarian societies, productivity relied mainly on land and labor; capital had little room to maneuver. After the Industrial Revolution, machines multiplied the efficiency of capital, but you still needed large numbers of workers to operate and maintain them. Now, AI may enable capital to convert almost directly into output.
Say AI is the next computer? I think a more accurate analogy is the steam engine—even more impactful. The Industrial Revolution took a century to reshape the world; the AI revolution will likely compress that process into a decade.
What This Means for Industries
If you accept the premise that "AI is not a computer but an industrial revolution," the subsequent implications are clear.
First, the valuation logic for AI companies should be completely different. Traditional IT companies hit a ceiling at the 99 trillion of economic activity. This explains why the market valuations of companies like Nvidia and OpenAI seem "irrational"—using the IT industry's framework to understand the AI industry is destined to underestimate it.
Second, every traditional industry will be rebuilt from the ground up. Not in the sense of "adding an AI feature," but fundamentally redesigning the value chain. Take the legal industry: previously, a due diligence project required more than a dozen junior lawyers to spend two weeks poring over documents; now AI can complete the same work in a few hours. This is not a story of "a law firm bought a new tool," but a story of "the firm's human resource structure must change completely."
Third, the logic of entrepreneurship has changed as well. In the past, the core moat of a startup was the team—whether you could recruit the best engineers and designers. Teams are still important, but AI has turned the variable of "execution capability" from "strongly correlated" to "weakly correlated." The real moats have shifted to industry understanding, insight into user needs, and the ability to iterate rapidly.
Choices on the Crest of the Wave
As the power of capital is further magnified, the resulting questions become very practical.
Every leap in productivity throughout history has been accompanied by changes in governance. The steam engine brought the factory system and labor laws; electricity gave rise to modern corporate governance; the internet drove data privacy legislation. For this round of AI, what is the corresponding governance framework? There is no answer yet.
Marx's extrapolation in Das Kapital of the unlimited expansion of capital's power has been moderated by various institutional designs over the past century and more. But if AI truly weakens the necessity of human labor in production to a large extent, will those previous balancing mechanisms still suffice? This is not a distant philosophical question, but a real challenge that is already unfolding.
In 2026, we may truly stand on the crest of a great transformation unseen in a century. This is not alarmism—when a technology simultaneously changes the source of productive forces and the structure of production relations, its impact extends far beyond the technical realm.
As someone working in this industry, I believe understanding the depth of this transformation is more important than chasing any specific technological hot topic. Tools will iterate; frameworks will update. But the reconstruction of production relations may happen only once in a century. And when computing power can be converted directly into productive force, whoever controls computing power controls power—this is no longer just a business issue.