1. An Economist Asks "What Is Technology, Really?"
W. Brian Arthur is an economist and a thinker in the complex-science tradition of the Santa Fe Institute. His best-known contribution is the theory of increasing returns (which explains why tech industries tend toward winner-take-all—it's also one of the theoretical sources of Kevin Kelly's New Rules for the New Economy).
In 2009, he published The Nature of Technology, trying to answer a simple question no one had seriously tackled before: What is technology, exactly? Where does it come from? How does it evolve?
Arthur's core insight is profound: Technology isn't "invented"—it "evolves" the way life evolves. Every new technology is essentially a recombination of existing technologies.
The jet engine wasn't created from scratch; it's a combination of compressors, combustion chambers, and turbines—all existing technologies. The smartphone wasn't invented from zero; it's a combination of touchscreens, processors, batteries, wireless communications, cameras, and other existing technologies. No technology is truly "new"—they're all new combinations of old ones.
2. Combinatorial Evolution: Why Technology Accelerates
Arthur's insight of "combinatorial evolution" explains something we all feel but can't quite articulate: Why is technology accelerating?
The answer lies in the mechanism of combination: The more technologies there are, the more new combinations become possible.
Imagine: when humans had only 10 basic technologies, the number of possible new combinations was limited. But when you have 10,000 technologies, the number of new combinations they can form is astronomical. The larger the total stock of technologies, the faster new possibilities grow exponentially.
That's why—major technological progress came once every tens of thousands of years in the Stone Age, once every few decades in the industrial era, and today nearly every year. Not because modern people are smarter, but because we stand on a massive "technology component library" with an exponentially growing number of possible combinations.
This is critical for understanding the AI era: AI is not an isolated technology; it's a "general-purpose component" that can be combined with almost every existing technology. AI + healthcare, AI + programming, AI + design, AI + finance... each combination creates a new domain. That's why AI's impact is so broad and so fast—it's a "meta-technology" that can combine with everything, and it will push Arthur's "combinatorial explosion" to another level.
3. Implications for Investors: Look at a Technology's "Combinatorial Potential"
Arthur's framework gives investors a unique lens for judging the value of a tech company: Look at a technology's "combinatorial potential."
The more "general-purpose" a technology is—the more it can combine with other technologies—the greater its long-term value.
Electricity was like that—it could combine with everything (lighting, motors, electronics, computing), so it created a century of value. Semiconductors are like that—they're the fundamental building block of almost all modern electronics. The internet is like that—it connects everything.
And AI is the new generation of "general-purpose combinatorial technology." That's why—companies that provide the "general-purpose component" of AI (Nvidia's compute, foundational models) may have more persistent value than companies that use AI to build a specific application. The former is "a fundamental component that can combine with everything"; the latter is just "one combination."
When I use Arthur's framework to evaluate tech investments, I ask: Is this company providing a "fundamental component that can combine with everything," or is it a "specific combinatorial application"? Fundamental components (compute, energy, networks, data) typically have deeper moats and more durable value than specific applications—because applications are constantly replaced by new combinations, while fundamental components are prerequisites for all combinations.
4. Where I Disagree with Arthur
First, "combinatorial evolution" explains well but predicts poorly.
Arthur's theory perfectly explains "why technology accelerates and where new technology comes from." But it can hardly predict "what the next big combination will be." The number of possible combinations is astronomical; which one succeeds depends on countless contingent factors. So Arthur's framework helps you understand the mechanism of technological evolution, but it doesn't tell you "which stock to buy." It's a worldview, not a stock-picking tool. This is the common flaw of almost all "grand theories."
Second, it underestimates the role of non-technological factors.
Arthur treats technological evolution almost as an autonomous, internally driven process—technologies combine and evolve by themselves. But in reality, which combinations succeed depends heavily on non-technological factors—capital, market demand, regulation, culture, luck. A technologically viable combination may fail because there's no market, no capital, or because regulations prohibit it. Arthur's "technology as autonomous evolution" undervalues the external constraints of business, politics, and society. Technology doesn't evolve in a vacuum.
Third, "combination" understates true breakthroughs.
Arthur says all technology is a recombination of old technologies. But some breakthroughs don't seem like mere combinations—they seem like true paradigm shifts—relativity, quantum mechanics, the discovery of DNA. These feel less like "combinations of existing things" and more like entirely new understandings of the world. Arthur's "combinatorial" view works well for incremental innovation, but may oversimplify paradigm-level breakthroughs. Not all innovation can be reduced to combination.
Fourth, it barely addresses the dark side of combinations.
Arthur's attitude toward technological combination is neutral to optimistic—combination brings acceleration, prosperity. But technology can also combine into terrifying things—nuclear weapons, biological weapons, surveillance systems, AI manipulation. When AI, this "general-purpose component," can combine with everything, it can produce medical breakthroughs and unprecedented manipulation and weaponry. Arthur's "combinatorial evolution" framework shows almost no awareness of these dangerous combinations. As technology accelerates, danger is accelerating too.
5. Arthur vs Schumpeter: Two Understandings of Innovation
Arthur's "combinatorial evolution" and Schumpeter's "creative destruction" are two complementary lenses for understanding technological innovation.
Schumpeter focuses on innovation's economic consequences—how new technology destroys old industries and reshapes competitive landscapes. He looks from the angle of "destruction."
Arthur focuses on innovation's internal mechanism—how new technology emerges from old combinations and why it accelerates. He looks from the angle of "generation."
Schumpeter tells you what innovation will destroy; Arthur tells you how innovation comes about.
Together, they offer investors a complete picture of innovation: Use Arthur to understand "where innovation comes from and why it accelerates" (AI is a meta-technology that can combine with everything, so it triggers a combinatorial explosion); use Schumpeter to understand "whom these innovations will destroy and whom they will elevate" (every combination creates disruptors and disrupted).
The most profound synthesis: We are living in an era of "combinatorial explosion" (Arthur), and this explosion brings with it a "creative destruction" of unprecedented scale (Schumpeter). AI, as a general-purpose component, will combine with everything, creating massive new technologies while destroying massive old business models. For investors, this means—future disruption will be faster and broader than ever before. Standing on the side of "fundamental components" and on the side of "disruptors" is more important than ever.
6. Final Thoughts
My biggest takeaway from this book is an "ecological perspective" on technology—no longer seeing technologies as isolated inventions, but as a self-combining, self-evolving ecosystem.
This perspective changed how I look at tech investing. I no longer ask "what cool thing did this company invent?" Instead, I ask: In the ecosystem of technological combination, what position does this company occupy? Is it providing a "fundamental component that can combine with everything," or is it just "a specific, easily replaceable application"?
That question, more than any financial metric, can judge a tech company's long-term value. Because in Arthur's world: Specific applications get eliminated by new combinations over and over, but "fundamental components" are prerequisites for all combinations, and their value is the most durable.
Electricity, semiconductors, the internet, now AI—history repeatedly shows that those providing "general-purpose fundamental components" are the most enduring winners. Those using these components to build specific applications are mostly passing fads, shining for a few years and then gone.
Arthur has a line I keep coming back to: "Technology is driven by the possibilities it creates for itself."
Every new technology opens up a new set of combinatorial possibilities; every new combination opens up more. This is a self-accelerating, never-ending process.
Understand this process, and you understand why we live in a world that changes faster and faster, and why this acceleration itself is accelerating.
For investors, this is both the greatest opportunity (combinatorial explosion creates massive new value) and the greatest risk (any "application" you hold could be eliminated by the next combination).
Pick your position well—stand on the side of fundamental components, stand on the side of winners who can harness combinations—that's the most important investment judgment in this age of acceleration.
