Moore's Law

Gordon Earle Moore (b. 1929), co-founder of Intel Corporation, articulated the observation in a 1965 Electronics Magazine article that would define five decades of silicon progress. Moore was not a theorist but an empiricist: examining the trajectory of integrated circuit complexity from 1959 onward, he extrapolated a doubling period of 18–24 months. His insight—later named Moore's Law by Carver Mead—became neither prediction nor law but a self-fulfilling industrial target. Moore's modesty about his own discovery was characteristic: he called it an observation, not a law of physics. His co-founder David Noyce and the Intel engineering teams (Andy Grove, Ted Hoff, Federico Faggin) transformed the observation into a manufacturing and design discipline that would outlast Moore himself into the 21st century.

Moore's Law is not an engineering specification but an empirical regularity that *drives* engineering targets. The mechanism: as photolithography wavelengths shrink (from 10 µm in 1965 to 3 nm in 2023), more transistors fit on a fixed die area. Each process node—a marketing term for a generation of fabrication—reduces feature size by ~30%, doubling transistor count within the doubling period. Intel, TSMC, and Samsung organized their R&D roadmaps around this target, treating it as a binding constraint. The physics became harder: quantum tunneling, leakage current, and heat dissipation emerged as limiting factors by the 2010s. Chipmakers responded with 3D stacking (FinFET, GAA transistors), heterogeneous integration, and chiplet architectures—engineering workarounds when planar scaling hit thermodynamic limits. Moore's Law thus became a self-imposed discipline: not a natural law but a manufacturing and design commitment that shaped the entire semiconductor industry's capital allocation, talent recruitment, and strategic planning for 58 years.

Gordon Moore's 1965 observation emerged from a decade of rapid progress in integrated circuit design. In 1959, the first practical IC contained ~10 transistors. By 1965, the Intel 4004 (designed by Ted Hoff and Federico Faggin, fabricated by 1971) would pack 2,300 transistors on a 10 mm² die. Moore, examining the trend, projected that the number of transistors on a chip would double every 18–24 months—a rate driven not by physics alone but by the economics of photolithography improvement and the competitive pressure of the semiconductor industry. For the next 50 years, this observation held with remarkable fidelity. The Intel 8008 (1972) had 3,500 transistors; the 8086 (1978), 29,000; the Pentium (1993), 3.1 million; the Core i7 (2008), 731 million; the NVIDIA H100 (2023), 92 billion. The law enabled the personal computer revolution (1980s), the internet (1990s), mobile computing (2000s), and artificial intelligence acceleration (2010s–2020s). By 2015, however, the physical limits of silicon became apparent. Quantum tunneling, electromigration, and heat density made further shrinking increasingly difficult and expensive. TSMC's 3 nm process (2022) cost $20 billion to develop. The doubling period slowed to 3–4 years. Yet the law's cultural and strategic grip remained absolute: chipmakers continued to pursue it, redefining 'node' names and adopting 3D architectures to sustain the illusion of Moore's Law's continuation.

Moore's Law was not invented; it was observed. Gordon Moore, examining historical data on integrated circuit complexity, noticed a consistent exponential trend and extrapolated it forward. The underlying cause was the relentless improvement in photolithography—the art of printing smaller features onto silicon. Each generation of photolithography tools (from contact printing to projection lithography to extreme ultraviolet) reduced the minimum feature size, allowing more transistors per unit area. The economic incentive was overwhelming: smaller transistors meant lower cost per transistor, which drove demand for chips in computers, consumer electronics, and telecommunications. The semiconductor industry, led by Intel, made Moore's Law a self-fulfilling prophecy by organizing its entire R&D and capital strategy around achieving the doubling target. It became a cultural norm, a competitive necessity, and a roadmap for the digital revolution. Moore's Law thus existed because (1) physics allowed it, (2) economics demanded it, and (3) industry committed to it.

Moore's Law was never a tool used by engineers or technicians in daily work. Rather, it was a strategic compass for semiconductor companies, chip architects, and technology investors. Intel executives used it to set long-term R&D budgets and manufacturing roadmaps. Chip designers used it to project future performance targets and plan for next-generation products. Investors used it to forecast earnings growth and justify valuations of semiconductor stocks. Computer manufacturers (Apple, IBM, Compaq) used it to plan product cycles, knowing that every 18–24 months, a new processor generation would offer double the transistor count and roughly 40–50% more performance. Software developers used it to justify writing increasingly complex code, trusting that hardware would catch up. Venture capitalists used it to evaluate startup pitches: any technology that required more than a few Moore's Law cycles to become viable was considered too risky. By the 2010s, as the law began to slow, its daily use shifted: chipmakers invoked it defensively, explaining delays in new nodes; AI researchers celebrated it retroactively, noting that the exponential growth in compute had enabled deep learning; and technologists debated whether it was truly ending or merely entering a new phase.

Moore's Law was not constructed; it was articulated. Gordon Moore wrote a two-page article in Electronics Magazine on April 19, 1965, titled 'Cramming More Components onto Integrated Circuits.' He examined historical data from 1959 to 1965, plotted transistor count against time on a semi-log graph, and observed a linear trend on the log scale—indicating exponential growth. He extrapolated the trend forward and predicted that the number of transistors per chip would continue to double every 18–24 months for at least another decade. The article was brief, data-driven, and cautious in tone. Moore made no claim to have discovered a law of nature; he simply reported what he saw in the data and what he expected to continue. The 'construction' of Moore's Law as a cultural force happened afterward: Carver Mead named it in the 1970s; Intel's leadership, particularly Andy Grove, made it a corporate mandate; the semiconductor industry adopted it as a shared target; and technology media amplified it into a quasi-religious principle. By the 1990s, Moore's Law had become a self-reinforcing prophecy: chipmakers invested billions to meet the target, which validated the law, which justified further investment. The 'construction' was thus not of a physical object but of an industrial consensus and a strategic discipline.

Moore's Law is not an artifact subject to archaeological discovery; it is a documented historical observation. However, the *evidence* for Moore's Law exists in museum collections and archives: (1) Gordon Moore's original 1965 Electronics Magazine article, held in the Smithsonian Institution archives and the Intel Museum (Santa Clara, California); (2) The Intel 4004 die itself, preserved at the Intel Museum and the Smithsonian's National Museum of American History; (3) Wafer samples from each major process node (10 µm, 3 µm, 1 µm, 0.8 µm, etc.), archived at TSMC, Intel, and Samsung fabrication plants; (4) Photolithography masks and tools from each generation, some preserved at the ASML headquarters (Netherlands) and in technology museums; (5) Original design documents and simulation records from Intel's archives, some declassified and donated to academic institutions. The 'archaeology' of Moore's Law is thus the archaeology of the semiconductor industry itself: preserved in corporate archives, museum collections, and the physical chips that embody the law's trajectory.