# Moore's Law **The model:** The observation, originally made by Intel co-founder Gordon Moore in 1965, that the number of transistors on an integrated circuit doubles approximately every two years. The result has held with remarkable accuracy for more than five decades — driving the exponential growth in computing power that has reshaped the global economy. The "law" isn't a physical law; it's a self-fulfilling industry roadmap that the semiconductor industry has organized itself around. The mental-model importance extends far beyond chips. Moore's Law is the canonical example of *sustained exponential growth in a technological capability*. Combined with [[Compound Interest]] reasoning, it predicts dramatic late-period changes that human intuition consistently underestimates. The first 30 years of the Law (1965-1995) produced computers that improved each generation; the last 30 years (1995-2025) have produced AI, smartphones, the internet, and now generative AI — capabilities that earlier generations couldn't have predicted from their position on the exponential curve. ## The Mechanism **Doubling every ~2 years.** Original formulation: doubling every year (1965); revised to every two years (1975). The actual pace has fluctuated but the long-run trend has held. Doubling every 2 years implies 32x improvement per decade and 1,000x per 20 years — exponential growth on a scale that linear intuition can't accommodate. **Driven by deliberate planning, not natural progress.** Moore's Law has held partly because the semiconductor industry organized roadmaps, capital expenditures, and R&D priorities around achieving it. The International Technology Roadmap for Semiconductors (now the IEEE International Roadmap for Devices and Systems) explicitly planned each generation's targets. Moore's Law was descriptive in 1965, predictive in 1975, and prescriptive thereafter. **Physical limits are real and approaching.** Moore's Law cannot continue forever — quantum effects become significant below 1-nanometer scale, and current chips are already in the 2-3nm range. The semiconductor industry has been working around the limits through new architectures (multi-core, GPU parallelism, specialized AI chips, 3D stacking) that maintain capability growth while abandoning the strict transistor-density formulation. Whether "Moore's Law" continues depends on how you define it. **Cost reductions compound with capability gains.** Each doubling of transistor density is also roughly a halving of cost per transistor. The compounding effect on computing-per-dollar over six decades has been dramatic — by some estimates, more than a trillion-fold improvement. This is the underlying driver of why computing has become essentially free as a per-operation cost. ## KB Applications ### AI Capability Trajectories Contemporary AI capabilities are a direct downstream consequence of Moore's-Law-driven computing growth combined with algorithmic improvements. GPT-class models require computing capacity that wouldn't have existed at any price 20 years ago. Each new model generation requires 10-100x more computing than the previous, partly enabled by continuing computing-power growth. The 2023-2026 generative AI revolution sits squarely on the exponential curve of accumulated Moore's-Law gains; it could not have happened earlier and may accelerate further as the underlying computing capability continues to grow. ### Why Tech-Right Politics Has Material Foundation The tech-right's political emergence is partly a story about who captured the trillion-dollar value created by Moore's-Law-driven economic growth. The semiconductor and software industries — the most direct beneficiaries — concentrated enormous wealth in a small number of individuals (the tech billionaires). The political alignment of that wealth has been a major theme of 21st-century U.S. politics. Without Moore's-Law-driven concentration of tech-sector wealth, the contemporary tech-right coalition wouldn't have its material base. ### China-U.S. Technology Competition The [[CHIPS and Science Act]] (2022) and the broader U.S.-China semiconductor competition are explicit policy responses to Moore's Law dynamics. Whoever controls the leading edge of semiconductor manufacturing controls the upstream constraint on AI, defense computing, and many other technologies. The U.S. has been losing ground to Taiwan (TSMC) for two decades; the CHIPS Act was an attempt to rebuild domestic capacity. The geopolitical stakes are substantial because Moore's-Law-driven capability differences compound into substantial strategic gaps. ### Why "Tech Doesn't Solve Politics" Misunderstands the Compounding A common left-coded argument: "more computing won't solve political problems." Partly true (politics has irreducible normative content); partly misleading (computing dramatically changes the action space available to political actors). The 2010s social-media revolution wasn't a "political problem solved by tech" but a "political problem created by tech." Moore's-Law-driven compounding produces capability levels that change what political problems even look like — sometimes ameliorating them, sometimes creating them, almost never leaving them unchanged. ### Why Exponential Growth Outlasts Linear Intuition A specific cognitive failure mode the Moore's-Law frame highlights: humans systematically underestimate the long-run effects of sustained exponential growth. Each individual year's change is modest; the compound effect over decades is dramatic. The same intuitive failure shows up in climate-change reasoning, in AI capability extrapolation, in financial compound-interest underestimation, and in demographic-trend underappreciation. Moore's Law is the canonical case study in the failure mode. ## Implications 1. **Exponential trends are real and consequential when they continue.** When you can identify a process that's actually compounding, projection should use exponential models even if your intuition resists. 2. **Self-fulfilling prophecies can be productive.** Moore's Law worked partly because the industry expected it to. Coordinated expectations can produce coordinated outcomes that no individual could have produced alone. 3. **Physical limits eventually bind.** Exponential growth eventually meets physical or resource constraints. Identifying when constraints will bind, and how the system will adapt, is part of honest analysis. 4. **Don't conflate technology and political progress.** Moore's Law is about capability; whether the capability is used well is a political question that the technology doesn't answer. ## See Also - [[Gordon Moore]] — the namesake; the 1965 article that articulated the prediction - [[Andy Grove]] — the Intel CEO who operationalized Moore's Law as a manufacturing roadmap - [[Compound Interest]] — the mathematical engine - [[Creative Destruction]] — what Moore's-Law-driven progress produces in market terms - [[Path Dependent]] — Moore's Law's continuation has been partly path-dependent on industry coordination - Lost Einsteins — the population of potential innovators that Moore's-Law-enabled tools could activate