The World We Thought We Lived In

The Immaterial Century

In 1991, the United States won a war without fighting it. The Soviet Union, the only rival capable of contesting American primacy, collapsed under the weight of its own material contradictions. It had too many tanks, too many factories, too many people organized around the production of physical things in an era that had already decided physical things were not the point. America, by contrast, had Microsoft, Goldman Sachs, and Hollywood. It had ideas, capital, and narrative. It won by being less heavy.

The lesson was not lost on the generation that inherited the world afterward. If the great contest of the twentieth century had been settled in favor of the side that bet on abstraction, then abstraction was the correct bet going forward. The most sophisticated minds in economics, strategy, and technology spent the following three decades constructing institutions, companies, and doctrines around a single organizing principle: value lives in software, and hardware is a commodity. A necessary cost. A solved problem. Something you outsource to people willing to do it cheaply, in places you don't need to think about.

This was not stupidity. It was a rational extrapolation from observed evidence, and for a long time the evidence cooperated. The companies that came to define American power in the decades after the Cold War were, almost without exception, companies that made nothing you could hold. Google indexed the world's information from servers nobody saw. Facebook owned the social graph of three billion people without owning a single building those people cared about. Amazon ran the logistics of the global economy and was still understood primarily as a software company. Apple, the one apparent exception, outsourced every physical operation it possibly could and kept only the design, the software, and the story. The message across all of them was identical: the material world is overhead. The immaterial world is where power actually compounds.

Governments absorbed this lesson too, though more slowly and with considerably less elegance. Trade policy across the 1990s and 2000s was constructed on the premise that deep economic integration made conflict irrational. Why contest territory when you can profit across it? China's WTO accession in 2001 was not simply a trade deal. It was a civilizational bet: that a nation brought fully into the global economic system would find the costs of disrupting that system too high to bear. Offshoring manufacturing was reframed not as hollowing out but as intelligent specialization. The software-defined nation-state farming out its physical production to lower-cost partners while retaining the high-value cognitive work. Elegant in theory. Compelling in spreadsheet. Completely correct until it wasn't.

For thirty years this worldview was not merely coherent but empirically vindicated at every turn. Living standards rose across most of the world. Great power conflict receded to the point where serious strategists began treating it as a historical artifact rather than a persistent condition. The most valuable real estate on earth was not a port or an oil field but a data center in a suburb of Seattle. A person in a dorm room could assemble something worth more than the GDP of a midsize country before turning twenty-five. The dematerialization of value seemed not just real but irreversible. A permanent feature of the modern world rather than a contingent feature of a specific historical moment.

And the moment is over.

Why It Worked

It is tempting, in retrospect, to treat the software-defined era as a prolonged collective delusion. A thirty-year hallucination shared by economists, strategists, and technologists who should have known better. This is the wrong reading, and it leads to the wrong conclusions.

The software-defined worldview worked because it was, for a specific and identifiable window of history, genuinely true. Not approximately true. Not true-enough-to-be-useful. Empirically, measurably, defensibly true in ways that compounded across decades and produced real outcomes for real people at a scale that is difficult to overstate.

Start with the economics. Between 1990 and 2015, global extreme poverty fell by more than half. That reduction was not delivered by ideology or institutional goodwill. It was delivered by manufacturing supply chains, most of them anchored in Asia, producing physical goods at costs that made them accessible to people who previously could not afford them. The software-defined worldview did not cause this directly, but it created the conditions for it. The logic of comparative advantage, applied globally through trade liberalization, allowed nations to specialize in what they produced most efficiently and trade for everything else. A factory worker in Shenzhen assembling components designed in Cupertino and shipped to consumers in Ohio was not a symptom of dysfunction. It was the system working as intended.

The geopolitical case was equally compelling. The period between 1991 and roughly 2014 was, by historical standards, extraordinarily peaceful among major powers. Not peaceful in the sense that conflict disappeared. Wars continued, brutally, in places the architects of the liberal order preferred not to look too carefully. But the specific nightmare of great power conflict, the kind that had consumed the first half of the twentieth century and threatened to incinerate the second, receded to a point where credible scholars began writing seriously about whether it had become structurally impossible. The argument had real teeth. Nations deeply integrated into global supply chains faced costs of conflict that earlier great powers did not. When your manufacturing base, your consumer markets, and your financial system are all threaded through the same global network as your rival's, the calculus of confrontation changes. Mutual economic destruction is a meaningful deterrent. It worked, for a while, better than anyone had a right to expect.

Then there was the innovation argument, which was perhaps the most seductive of all. The software-defined era did not just redistribute existing wealth. It generated new categories of value that had never existed before. Search. Social networks. Cloud infrastructure. Smartphones. Streaming. These were not refinements of prior technologies. They were genuinely novel, and they emerged almost entirely from environments where capital was cheap, talent was mobile, and physical constraints were treated as engineering problems to be abstracted away. The venture capital model, which funded speculative bets on intangible assets with no physical collateral, produced returns that made traditional capital allocation look primitive. The message from the market was consistent and loud: the immaterial world generates compounding returns, the material world generates margins.

None of this was fabricated. The wealth was real. The poverty reduction was real. The relative peace among great powers was real. The innovation was real. The people who built the intellectual scaffolding of the software-defined era were not fools or frauds. They were pattern-matching correctly on the data available to them, constructing a worldview that explained observed outcomes and generated accurate predictions across multiple domains for multiple decades.

This is exactly what makes what happened next so serious.

The software-defined worldview did not fail because it was intellectually weak. It failed because it was so empirically successful for so long that it stopped being treated as a model and started being treated as a law. The difference between those two things is the difference between a map and the territory. The map was accurate. The territory changed.

And buried inside every spreadsheet, every trade agreement, every fabless business model, every national security doctrine written during those thirty years was an assumption so foundational that almost nobody thought to write it down.

The Assumption Buried Inside

Every coherent worldview rests on a foundation of things it does not bother to defend. Not because the foundation is indefensible, but because it has never been seriously challenged, and unchallenged premises have a way of migrating out of the category of assumptions and into the category of facts. The software-defined era had many such premises. This is the one that mattered most:

Not stated. Not argued. Not written into any treaty or doctrine or business plan in explicit terms. Simply assumed, the way you assume the floor will hold your weight when you walk across a room. Someone would always fabricate the chips. The factories would persist. The supply chains, however complex, however geographically concentrated, would continue to function because they always had and because the economic incentives for all parties to keep them functioning were, in the aggregate, overwhelming. Hardware was a commodity problem. Commodity problems get solved by markets. Markets are reliable. Therefore hardware is reliable. The syllogism was never examined because the conclusion always held.

For a long time, the conclusion held because the assumption was being continuously validated by a specific set of historical conditions that nobody recognized as conditions. Taiwan remained stable. ASML, a single Dutch company that manufactures the lithography machines required to produce leading-edge chips, continued to operate without interruption. The rare earth supply chains threading through China continued to flow. TSMC, which by the 2020s was fabricating the majority of the world's most advanced semiconductors within a geography that the People's Republic of China considers an unresolved territorial question, kept expanding capacity. The system was fragile in ways that were visible to anyone who looked directly at it. Almost nobody looked directly at it.

The first serious crack appeared not where geopolitical analysts expected but in hospital parking lots. In 2021, automobile manufacturers began halting production lines across the United States, Germany, and Japan because they could not source semiconductors. Not advanced chips. Not cutting-edge processors. Basic microcontrollers, the kind embedded in door locks and dashboard displays, components so mundane they had been treated for decades as interchangeable commodities available on demand. They were not available on demand. They were available only through supply chains so lean, so geographically concentrated, and so optimized for efficiency rather than resilience that a single disruption propagated globally within months. Ford lost an estimated two and a half billion dollars in profit that year. General Motors lost roughly the same. The global automotive industry collectively surrendered more than two hundred billion dollars in revenue.

The policy class noticed, registered concern, and largely failed to understand what they were actually seeing. The chip shortage was framed as a supply chain problem, a logistics failure, a pandemic disruption that would self-correct as conditions normalized. This framing was not wrong about the immediate cause. It was wrong about the underlying structure. The shortage was not an aberration. It was a preview. A brief, low-stakes demonstration of what happens when the physical substrate of the modern economy becomes unavailable, in an industry peripheral enough to absorb the blow without systemic collapse. The demonstration was not heeded with the seriousness it warranted.

Because the automotive shortage was noise compared to the signal underneath it.

The signal is this: the production of leading-edge semiconductors (the chips that power artificial intelligence, advanced weapons systems, financial infrastructure, and communications networks) is not distributed across the global economy in any meaningful sense. It is concentrated, to a degree that has no precedent in the history of strategically critical resources, in a handful of facilities on a single island that sits at the center of the most consequential territorial dispute on earth. TSMC's most advanced fabrication plants are located in Taiwan. The machines required to build those plants are manufactured almost exclusively by ASML in the Netherlands. The specialized gases, chemicals, and raw materials required to run them thread through supply chains spanning dozens of countries, with critical chokepoints at every link. Remove any single node and the entire system does not slow down. It stops.

This is not a supply chain problem. It is a sovereignty problem dressed in the language of logistics, and the distinction is not semantic. Supply chain problems have market solutions. You add redundancy, diversify suppliers, build inventory buffers, and over time the system stabilizes. Sovereignty problems do not have market solutions because markets cannot resolve the underlying question, which is not how to move goods efficiently but who controls the physical capacity to produce the foundation of modern technological power and what they are willing to do with that control.

The software-defined worldview had no framework for this question because it had no framework for the possibility that hardware could become leverage. In the model, hardware was always becoming cheaper, more abundant, more commoditized, more irrelevant to the actual sources of value. Moore's Law made the chips faster and smaller every two years. Globalization made them cheaper. The trend lines were all pointing toward a world where the physical substrate mattered less and less. Instead, a specific convergence of technological, economic, and geopolitical forces produced the opposite outcome. The chips became so powerful, so central to so many critical systems, and so difficult to produce at the frontier that the physical capacity to make them became the most concentrated and consequential form of leverage in the modern world.

The thirty years of software-defined thinking did not see this coming because it was structurally incapable of seeing it coming. You cannot identify a threat to an assumption you have stopped treating as an assumption.

The floor, as it turns out, does not always hold your weight.

We built the most important technology in human history on top of a single point of failure, told ourselves it was a supply chain, and moved on. Part II is what that actually looks like from the ground.