In 1965, Gordon Moore was asked by Electronics magazine to predict what would happen in the semiconductor industry over the next decade. Moore, then director of R&D at Fairchild Semiconductor, examined the data. The number of transistors that could be placed on a single integrated circuit had been doubling roughly every year since the invention of the planar process in 1959. He extrapolated: this doubling would continue for at least ten more years. The article — "Cramming More Components onto Integrated Circuits," published April 19, 1965 — contained a single graph with a trend line that would become the most consequential prediction in the history of technology. Moore projected that by 1975, a single chip would contain 65,000 transistors. The industry hit that target almost exactly.
In 1975, Moore revised the doubling period from every year to every two years, reflecting the increasing complexity and capital requirements of each new process node. Caltech professor Carver Mead coined the term "Moore's Law" shortly afterward, and the label stuck — despite Moore himself pointing out repeatedly that it wasn't a law of physics but an empirical observation about industrial capability. The distinction matters. Gravity operates whether or not anyone believes in it. Moore's Law operates because an entire industry organizes its R&D roadmaps, capital expenditure cycles, and competitive strategies around the expectation that it will continue. It is a prediction that became a coordination mechanism that became a self-fulfilling prophecy.
The numbers are staggering in their consistency. Intel's 4004 processor in 1971 contained 2,300 transistors on a 10-micron process. The 8086 in 1978 had 29,000. The 486 in 1989 had 1.2 million. The Pentium 4 in 2000 had 42 million. By 2024, Apple's M4 chip contained approximately 28 billion transistors manufactured on TSMC's 3-nanometer process — a 12-million-fold increase over the 4004 in roughly 53 years. Plot those data points on a logarithmic scale and the line is remarkably straight. No other industrial trend in human history has maintained exponential growth over five decades.
The economic consequences of this trajectory have reshaped every industry on earth. When transistor density doubles every two years, the cost per transistor halves. In 1970, a single transistor cost roughly one dollar. By 2024, that cost had fallen below one ten-billionth of a dollar. Computation that required a room-sized mainframe costing millions in 1970 now fits in a device that costs $200 and slips into your pocket. This exponential deflation in the cost of logic gates is the substrate on which the entire digital economy was built — from personal computers to the internet to smartphones to artificial intelligence. Every software company, every cloud platform, every social network, every autonomous vehicle project is downstream of the trend Gordon Moore sketched on a napkin in 1965.
What makes Moore's Law intellectually distinctive is that it operates at the intersection of physics, economics, and collective will. The physics sets upper bounds — you cannot make a transistor smaller than an atom. The economics create incentives — whoever reaches the next process node first captures disproportionate margins. And the collective will — the shared expectation across chip designers, equipment manufacturers, materials scientists, and corporate planners that the trend will continue — coordinates the billions of dollars in investment required to make each doubling happen. Intel, TSMC, Samsung, and ASML don't achieve the next node independently. They achieve it because the entire ecosystem plans on the same cadence. Remove the shared expectation and the coordination collapses. This is why Moore's Law is better understood as an industry metronome than as a physical constant.
The law's influence extends far beyond semiconductors. Entire strategy frameworks — disruption theory, technology adoption curves, venture capital return models — assume exponential improvement in price-performance as a background condition. When Marc Andreessen declared in 2011 that "software is eating the world," the implicit assumption was that the hardware underneath the software would continue to get cheaper and more powerful on the Moore's Law trendline. When Jeff Bezos built Amazon Web Services on the bet that computing would become a utility, he was making a Moore's Law wager: that server costs would fall fast enough to make on-demand cloud computing economically rational for every business. The law doesn't just describe semiconductor progress. It underwrites the entire strategic logic of the technology industry.
The counterintuitive consequence of sustained exponential improvement is that human beings consistently underestimate its long-run effects and overestimate its short-run effects. In 1977, Ken Olsen, founder of Digital Equipment Corporation, reportedly said "there is no reason anyone would want a computer in their home." In 1995, Clifford Stoll wrote a Newsweek column arguing that the internet would never replace newspapers, teachers, or physical stores. Both were reasoning linearly about an exponential process. The transistor budgets required for personal computing and internet infrastructure were, at the time of those statements, genuinely insufficient. Two or three doublings later, they were more than sufficient. Moore's Law punishes linear forecasters and rewards those willing to project the curve forward — uncomfortably, counterintuitively — and position themselves for where the exponential is heading rather than where it is today.
Section 2
How to See It
Moore's Law is invisible precisely because it's ubiquitous. You don't notice exponential improvement when you're living inside it — only when you step back and compare across time. The signal is always the same: something that was expensive, slow, or physically large becomes cheap, fast, and small at a rate that violates linear intuition.
The challenge is distinguishing genuine Moore's Law dynamics from unrelated cost declines or one-time technology transitions. Not every price drop is exponential, and not every exponential is Moore's Law. Train your pattern recognition on these signatures:
Technology
You're seeing Moore's Law when a device that required dedicated hardware a decade ago now runs as a software application on a general-purpose chip. In 2005, a professional GPS navigation unit cost $500 and did one thing. By 2012, a $199 smartphone contained a GPS chip that was faster, more accurate, and also served as a camera, music player, calculator, compass, and internet terminal. The GPS function consumed a negligible fraction of the phone's transistor budget. Moore's Law didn't just make GPS cheaper — it made GPS so cheap that it became a rounding error inside a device defined by entirely different capabilities.
Business
You're seeing Moore's Law when a startup can offer for free what an incumbent charged millions for a decade earlier, and the incumbent's margins are collapsing despite no strategic error. Genomic sequencing cost $2.7 billion for the first human genome completed in 2003. Illumina brought the price below $1,000 by 2014 and below $200 by 2024. Companies built on early-generation sequencing economics watched their cost structures become unviable not because competitors outmaneuvered them but because the underlying compute and chemistry costs fell on an exponential curve that legacy pricing couldn't survive.
Investing
You're seeing Moore's Law when a company's capital expenditure on compute increases every year while its cost per unit of compute decreases — and the gap between those two curves is where the margin lives. Amazon spent $48.4 billion on capital expenditures in 2023, much of it on data center infrastructure. Yet AWS's effective price per compute-hour has fallen by roughly 25–30% every three years since 2006. Revenue grows because the falling cost per computation unlocks demand that didn't exist at the previous price point. Investors who model AWS as a hardware business see rising capex and worry. Investors who understand Moore's Law see each dollar of capex buying exponentially more capability.
Strategy
You're seeing Moore's Law when a technology that is currently too expensive for mainstream adoption is exactly 3–4 doublings away from a price point that restructures an entire industry. In 2016, training a state-of-the-art neural network cost tens of millions of dollars in compute. By 2024, equivalent training runs cost a fraction of that figure, and inference costs had dropped even faster. The companies that positioned themselves early — building data pipelines, training talent, and accumulating proprietary datasets while compute was still expensive — captured disproportionate value once Moore's Law (and its AI-specific cousins in GPU performance) brought costs below the threshold where deployment became economically rational for every enterprise.
Section 3
How to Use It
Decision filter
"Is my strategy dependent on the current cost or capability of computation — or on the trajectory? If the cost of compute halves every two years, does my competitive position strengthen or weaken? The answer determines whether I'm riding Moore's Law or being run over by it."
As a founder
Moore's Law is the most reliable tailwind in technology, but only if you build for where the curve is going — not where it is today. The founders who extract the most value from exponential improvement are those who design products that are slightly ahead of the hardware, then let the hardware catch up.
Reed Hastings understood this when Netflix launched streaming in 2007. At the time, broadband speeds were barely adequate for standard-definition video. Buffering was constant. Image quality was poor. The product felt premature. But Hastings had modeled bandwidth cost curves — which follow their own exponential trajectory correlated with Moore's Law through cheaper networking hardware — and knew that within five years, HD streaming would be seamless for most American households. He invested in content licensing and streaming infrastructure while the experience was still mediocre, because by the time the experience was excellent, the infrastructure had to already be in place. By 2013, Netflix had 40 million streaming subscribers. Blockbuster, which had waited for the experience to be ready before investing, filed for bankruptcy in 2010.
The tactical implication: when building a compute-intensive product, price your service based on next year's hardware costs, not this year's. You'll operate at thin or negative margins initially, but the cost curve will bail you out if the product gains traction. AWS, Google Cloud, and Azure all followed this playbook — pricing compute aggressively below current cost to capture market share, knowing that Moore's Law would turn today's loss leader into tomorrow's profit center.
As an investor
Moore's Law creates a specific pattern that separates great technology investments from good ones: the companies that convert exponential hardware improvement into exponential business value are those with software or data layers that amplify the hardware gains rather than merely consuming them.
NVIDIA's trajectory illustrates this. In 2012, NVIDIA was a $10 billion market-cap company making graphics cards for gamers. GPU transistor counts followed Moore's Law, but the company's revenue grew roughly linearly with the gaming market. Then deep learning researchers discovered that GPUs could train neural networks orders of magnitude faster than CPUs. Suddenly, NVIDIA's hardware improvements — driven by the same Moore's Law trajectory as before — unlocked an entirely new demand curve. By 2024, NVIDIA's market capitalization exceeded $3 trillion, not because Moore's Law accelerated but because AI created a software demand layer that converted each hardware doubling into massively more economic value. The transistor trend line didn't change. The value captured per transistor did.
The investor's heuristic: look for companies where Moore's Law is a compounding input, not just a cost reduction. A company that uses cheaper compute to deliver the same product at lower cost captures some value. A company that uses cheaper compute to deliver a product that was previously impossible captures all the value.
As a strategist
Moore's Law is a weapon against incumbents and a threat to anyone who builds a business model on a specific price-performance ratio remaining stable. The strategist's job is to identify which current constraints are temporary — artifacts of today's transistor density that will evaporate within two or three doublings — and which are permanent.
Apple's strategic genius under Steve Jobs was repeatedly identifying the moment when Moore's Law made a previously impractical product category viable. The original iPod in 2001 required a 1.8-inch hard drive that could hold 1,000 songs in a pocket-sized device — possible only because storage density had reached a specific threshold. The iPhone in 2007 required an ARM processor powerful enough to run a full mobile operating system with a multitouch interface — a capability that didn't exist three years earlier. The iPad in 2010 needed a battery-efficient processor that could drive a 9.7-inch display for 10 hours. In each case, Jobs didn't invent the enabling hardware. He identified the moment when Moore's Law delivered the hardware that made the product concept feasible, then executed faster than anyone else.
The strategic framework: maintain a pipeline of product concepts that are currently infeasible due to cost or performance constraints. Estimate how many Moore's Law doublings separate each concept from viability. When the gap closes to one doubling, begin engineering. When it closes to zero, ship.
Common misapplication: Assuming Moore's Law applies uniformly to all technology. It doesn't. Moore's Law specifically describes transistor density on integrated circuits. Battery energy density improves at roughly 5–8% per year — not the 40% per year implied by a two-year doubling. Solar cell efficiency has improved dramatically but follows a different curve entirely. Memory bandwidth has lagged processor speed improvements for decades, creating the "memory wall" that constrains real-world performance regardless of how many transistors you add. Software complexity often grows faster than hardware capability, which is why your laptop feels no faster than the one you bought five years ago despite having four times the transistors.
The founders and investors who get burned are those who assume that every technology input follows the semiconductor curve. Some do. Most don't. Identify which curve each critical input follows — exponential, linear, logarithmic, or S-shaped — and plan accordingly. The discipline is in knowing which inputs are on a Moore's Law trajectory and which are not.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
Moore's Law is background radiation in the technology industry — always present, rarely credited explicitly, yet shaping every major strategic decision of the last half century. The leaders below didn't just benefit passively from exponential hardware improvement. They built their companies' strategies around the specific cadence of the doubling, timing market entries, pricing decisions, and architectural bets to ride the curve rather than fight it.
What connects these cases across different eras, different markets, and different technologies is the shared discipline of planning on a logarithmic timescale when competitors planned linearly. Each leader internalized a version of the same insight: in a Moore's Law world, the future arrives faster than it feels like it should, and the companies that position for the next doubling before it arrives capture disproportionate value when it does.
Moore didn't just observe the trend — he built the company that sustained it. When he and Robert Noyce founded Intel in 1968, the business plan was explicitly constructed around the prediction in Moore's 1965 paper: lead in process technology, capture the cost-per-transistor advantage, and reinvest the resulting margins into the next process generation before competitors could catch up. This created a flywheel that Intel rode for over four decades.
The strategy required extraordinary capital discipline. Intel consistently spent 20–25% of revenue on R&D and another 20–25% on capital expenditure — ratios that would have seemed reckless in any other industry. But Moore understood that in a business governed by exponential improvement, falling one generation behind on process technology meant falling exponentially behind on cost structure. A competitor with a two-year lead didn't have a 10% cost advantage — they had a 50% advantage, because each node halved the cost per transistor. This mathematical reality made Intel's massive reinvestment rates not just rational but existential.
Moore's deeper contribution was institutional. By embedding the law as an organizational planning assumption — chipsets would shrink on a predictable schedule, costs would fall on a predictable curve, new products would ship at predictable intervals — he gave Intel's engineering teams a metronome that coordinated thousands of people across hundreds of projects. The tick-tock cadence that Intel formalized in 2007 (alternating between new architectures and die shrinks) was merely the explicit codification of what Moore had been running implicitly since the 1970s.
If Moore identified the trend, Andy Grove weaponized it. As Intel's CEO from 1987 to 1998, Grove transformed the company from a memory chip manufacturer into the dominant supplier of microprocessors for personal computers — a strategic pivot that only made sense through the lens of Moore's Law.
Grove's most consequential decision came in 1985, when he convinced Moore to exit the DRAM memory business that Intel had founded. Japanese manufacturers had achieved cost parity on memory chips, and Grove recognized that competing on cost in a commodity market — even with a process technology lead — was a losing strategy. Microprocessors, by contrast, were differentiated products where each Moore's Law generation enabled new software capabilities that consumers would pay premium prices for. The margin difference was decisive: memory chips yielded single-digit margins, while processors yielded 60%+ gross margins because the exponential improvement in transistor counts translated into performance gains that software developers immediately consumed.
Grove's "paranoid" management philosophy — captured in his 1996 book Only the Paranoid Survive — was a direct response to the dynamics Moore's Law creates. In an industry where the competitive landscape reshuffles every 18–24 months, complacency is lethal. Grove institutionalized what he called "strategic inflection points," moments when exponential improvement in one dimension (transistor density, bandwidth, storage) crosses a threshold that restructures an entire market. His job, as he saw it, was to identify those inflection points before they arrived and position Intel to benefit from them rather than be disrupted by them. The shift from single-core to multi-core processors in the mid-2000s was an inflection point Grove had anticipated years earlier, even as it only materialized after his tenure.
Jensen Huang co-founded NVIDIA in 1993 to build graphics processing units — chips optimized for the massively parallel floating-point calculations required to render 3D images. For two decades, NVIDIA rode Moore's Law in a straightforward way: each new GPU generation packed more transistors, rendered more polygons, and sold to a larger gaming market. Revenue grew from $158 million in 2000 to $4.7 billion in 2015. Respectable, but not transformative.
Then the curve bent. In 2012, researchers Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton used two NVIDIA GTX 580 GPUs to train AlexNet, a deep neural network that won the ImageNet competition by a margin that stunned the machine learning community. The key insight: the parallel architecture that made GPUs excellent at rendering graphics also made them excellent at the matrix multiplications that neural networks require. Suddenly, Moore's Law applied to GPU transistor counts wasn't just a gaming story. It was an AI story.
Huang recognized the inflection point faster than anyone in the industry. NVIDIA pivoted massive R&D resources toward AI-specific hardware — the Tesla and later A100 and H100 data center GPUs — while competitors were still debating whether deep learning was a fad. By 2024, NVIDIA controlled approximately 80% of the AI training chip market. Market capitalization crossed $3 trillion. The underlying Moore's Law trajectory hadn't changed — GPU transistor counts continued doubling on roughly the same cadence. What changed was that each doubling now unlocked exponentially more economic value because artificial intelligence created demand for computation that scaled faster than Moore's Law could supply it. Huang's strategic insight was that when demand growth outpaces Moore's Law supply growth, the company riding the supply curve captures extraordinary margins.
When Lisa Su became CEO of Advanced Micro Devices in October 2014, the company was eighteen months from bankruptcy. AMD's stock price was $2.65. It had lost $1.2 billion over the previous four years. Intel held over 80% of the x86 processor market, and NVIDIA dominated discrete GPUs. AMD's products were a generation behind on process technology — in an industry governed by Moore's Law, that meant they were exponentially behind on cost-per-transistor.
Su's turnaround strategy was built on a single Moore's Law insight: AMD didn't need to own fabrication. In 2009, AMD had spun off its manufacturing operations into GlobalFoundries. Su leaned into this decision, partnering with TSMC to access leading-edge process technology that AMD could never have afforded to develop independently. While Intel stumbled on its 10-nanometer process — delayed from 2016 to 2019 — TSMC delivered 7-nanometer on schedule in 2018. AMD's Zen 2 processors, manufactured by TSMC, achieved performance parity with Intel for the first time in over a decade. By 2022, AMD's Zen 4 on TSMC's 5-nanometer process surpassed Intel on multiple benchmarks.
The financial results validated the strategy. AMD's stock price rose from $2.65 in 2014 to over $160 by 2024 — a sixty-fold increase. Revenue grew from $5.5 billion to over $22 billion. Su demonstrated that in a Moore's Law industry, access to the best process technology matters more than owning it. Her strategic bet was that the semiconductor industry's capital intensity would concentrate manufacturing in a few foundries, and that the companies best positioned would be those who could design the most efficient architectures for whatever process node the foundry delivered next. The law rewarded the best designers, not the biggest spenders.
Jeff Bezos launched Amazon in 1994 with a bet that Moore's Law would make e-commerce infrastructure — servers, storage, bandwidth — dramatically cheaper every year. His 1997 letter to shareholders made the logic explicit: Amazon would prioritize market share over profitability, investing heavily in infrastructure while prices were still high, because the long-term cost trajectory favored the company that built the largest customer base first. When compute costs halved, the margins would materialize.
But Bezos's most consequential Moore's Law bet was AWS, launched in 2006. The insight was that most companies were bad at predicting their computing needs — they either bought too much hardware (waste) or too little (outages). Moore's Law meant that the hardware they purchased depreciated in value faster than any other capital asset. By offering computing as a utility, priced per hour and scaled elastically, Bezos converted Moore's Law from a hardware depreciation problem into a service margin expansion engine. Each hardware generation that halved the cost per compute-hour flowed partially to customers as price reductions (keeping them loyal) and partially to Amazon as margin expansion (funding the next generation of data centers).
By 2024, AWS generated over $90 billion in annual revenue with operating margins above 30%. The service hosted millions of active customers, including Netflix, NASA, and the CIA. Bezos had built a business model where Moore's Law was the primary margin driver: every doubling in transistor density automatically improved AWS's cost structure without requiring any change to the product. The law didn't just enable the business — it was the business.
Section 6
Visual Explanation
Section 7
Connected Models
Moore's Law doesn't operate in a vacuum. It intersects with strategic, economic, and technological frameworks that either amplify its effects, constrain its predictions, or emerge as natural consequences of its relentless exponential trajectory. Understanding these connections — and knowing which adjacent model to reach for in a given situation — is what separates people who cite the law from people who use it.
Reinforces
[Compounding](/mental-models/compounding)
Moore's Law is compounding applied to transistor density — and the psychological dynamics are identical. Early doublings are imperceptible: going from 2,300 to 4,600 transistors changes nothing meaningful. But the 30th doubling — from 14 billion to 28 billion transistors — enables entirely new categories of computation (large language models, real-time ray tracing, autonomous vehicle perception). Charlie Munger's observation that compounding's power lies in its back end applies precisely: the first two decades of Moore's Law produced calculators and word processors, the last two decades produced smartphones, cloud computing, and artificial intelligence. The critical discipline in both cases is patience — maintaining investment through the early, unremarkable doublings to reach the transformative ones.
Reinforces
[Metcalfe's Law](/mental-models/metcalfes-law)
Moore's Law makes the hardware cheap. Metcalfe's Law makes the network valuable. The interaction is multiplicative: as Moore's Law drives down the cost of connected devices, more people can afford to join networks, which increases the Metcalfe value of those networks, which creates demand for more connected devices, which justifies further Moore's Law investment. The smartphone revolution was this flywheel in action — Moore's Law made a $200 supercomputer possible, Metcalfe's Law made it indispensable because everyone else had one too. Neither law alone explains the digital transformation of the last two decades. Their interaction does.
Tension
Economies of [Scale](/mental-models/scale)
Economies of scale assume that cost advantages come from producing more units of the same thing. Moore's Law generates cost advantages through producing more capable units at the same cost. The tension surfaces in strategic planning: a company optimizing for economies of scale invests in production volume. A company optimizing for Moore's Law invests in process technology. Intel's historic dominance came from choosing the latter — its fabs were never the highest-volume, but they were always the most advanced. When Intel lost its process lead to TSMC in the late 2010s, its scale advantage in production volume couldn't compensate for the cost-per-transistor disadvantage. In Moore's Law industries, process leadership trumps production scale.
Section 8
One Key Quote
"The complexity for minimum component costs has increased at a rate of roughly a factor of two per year. Certainly over the short term this rate can be expected to continue, if not to increase."
— Gordon Moore, 'Cramming More Components onto Integrated Circuits,' Electronics, April 19, 1965
Section 9
Analyst's Take
Faster Than Normal — Editorial View
The debate over whether Moore's Law is "dead" has become an annual tradition at semiconductor conferences, and it misunderstands what the law actually is. If you define it narrowly — as the geometric scaling of planar CMOS transistors — then yes, it ended around 2015 when FinFETs replaced traditional planar architectures. If you define it as the trend Gordon Moore actually described — the exponential increase in the number of transistors on a cost-effective integrated circuit — then it's alive and well, sustained by innovations in 3D stacking, chiplet architectures, advanced packaging, and new transistor geometries that the 1965 paper never anticipated but that serve the same economic function. The law has outlived every one of its obituaries.
The more interesting question is what happens when the law's cadence changes. For fifty years, the technology industry has operated on a roughly two-year planning cycle synchronized with process node advances. Product roadmaps, capital expenditure budgets, competitive positioning — everything aligns to the doubling tempo. If that tempo slows from two years to three or four, the cascading effects on strategy are profound. Companies that depended on next-generation hardware to make their products viable face longer wait times. Startups that assumed compute costs would halve before their runway expired find the math no longer works. Investors who modeled returns on a two-year improvement cycle need to recalibrate. The law's cadence is as important as its direction.
What I find most underappreciated: Moore's Law is as much a story about capital concentration as about physics. The cost of a leading-edge fab was $1 billion in 2000, $10 billion in 2015, and over $20 billion by 2024. ASML's extreme ultraviolet lithography machines — the only tools capable of printing features at 5 nanometers and below — cost $350 million each, and ASML is the sole supplier on earth. This capital intensity has created the most concentrated supply chain in any major industry: TSMC manufactures over 90% of the world's most advanced chips, using machines from a single Dutch company, with key photoresist chemicals sourced from a handful of Japanese suppliers. Moore's Law made this concentration inevitable — when each process generation costs exponentially more, only the players with the largest cumulative investment can afford to continue. The geopolitical implications, particularly around Taiwan, are the most consequential strategic risk in global technology.
For founders and investors, the actionable insight is this: don't plan around Moore's Law continuing. Plan around what happens if it doesn't. The companies best positioned for a post-Moore's Law world are those with software and algorithmic advantages that are independent of hardware improvement. When hardware improvement was fast and cheap, brute-force approaches worked — throw more transistors at the problem. As improvement slows and gets more expensive, algorithmic efficiency and software optimization become the scarce resource. The next generation of valuable technology companies will likely be those that extract the most computation per transistor, not those that simply consume the most transistors.
Section 10
Test Yourself
Moore's Law is cited so frequently — and so loosely — that its meaning has blurred. People invoke it to describe any technology cost decline, any exponential curve, any vaguely impressive improvement trajectory. These scenarios test whether you can distinguish genuine Moore's Law dynamics — exponential transistor density improvement on integrated circuits driving cost-performance gains in computation — from adjacent but distinct phenomena that share the surface pattern of exponential improvement but operate through entirely different mechanisms and follow different trajectories.
Is this mental model at work here?
Scenario 1
A solar panel manufacturer reports that the cost per watt of solar electricity has fallen 99% since 1976, from $76 to under $0.30. The CEO attributes this to 'Moore's Law for solar' and projects continued exponential cost declines based on the semiconductor analogy.
Scenario 2
NVIDIA's H100 GPU contains 80 billion transistors, roughly 35 million times more than Intel's 4004 from 1971. A new AI lab builds its training cluster entirely on H100s, and the CEO tells the board that their models will be twice as capable in two years because Moore's Law will deliver twice as many transistors per chip.
Scenario 3
In 2007, Apple launches the iPhone with a 412 MHz ARM processor and 128 MB of RAM. By 2024, the iPhone 16 Pro contains a 3-nanometer A18 Pro chip with approximately 20 billion transistors, running at frequencies above 4 GHz with 8 GB of RAM. The device's base price has remained between $699 and $999 throughout — roughly flat in inflation-adjusted terms — while its computational capability has increased by several orders of magnitude.
Section 11
Top Resources
The literature on Moore's Law spans semiconductor physics, technology economics, and corporate strategy. The strongest resources combine the engineering foundations with the strategic and economic implications — showing how an observation about transistor density in a four-page trade magazine article became the organizing principle of the world's most valuable industry and the substrate of the modern global economy.
The four-page paper that started everything, published in Electronics magazine on April 19, 1965. Moore's original text is remarkably readable — no jargon, no equations, just a clear empirical observation and a projection backed by data. The key graph, showing component counts rising exponentially against time, has been validated for six decades. Anyone who invokes Moore's Law should read the primary source. It takes fifteen minutes.
Grove's management classic explains how Moore's Law creates "strategic inflection points" — moments when exponential improvement in one technology dimension crosses a threshold that restructures entire markets. Written from the CEO's chair at the company most directly shaped by the law, the book translates semiconductor physics into competitive strategy. Grove's framework for identifying and navigating inflection points remains the best available guide for leaders in any Moore's Law-adjacent industry.
The definitive narrative history of the integrated circuit, from Jack Kilby's and Robert Noyce's independent inventions in 1958–1959 through the rise of the semiconductor industry. Reid traces the physics, the personalities, and the industrial economics that made Moore's observation possible. Essential context for understanding that Moore's Law didn't emerge from theory — it emerged from a specific industrial ecosystem with specific economic incentives.
Miller's history of the global semiconductor industry connects Moore's Law to geopolitics, national security, and the concentration of advanced manufacturing in Taiwan. The book explains why the exponential cost increase of each new process node has created the most concentrated supply chain in any critical industry — and why the strategic implications extend far beyond technology. Required reading for anyone who wants to understand the geopolitical dimension of semiconductor progress.
Huang's articulation of how Moore's Law interacts with parallel computing architectures to produce capability improvements that exceed the raw transistor trend. His concept of "Huang's Law" — that GPU performance for AI workloads doubles faster than Moore's Law predicts — is the most important contemporary extension of Moore's original observation. The blog and associated GTC keynotes provide the clearest view of where the exponential improvement is headed as classical scaling approaches its limits.
Moore's Law — Transistor counts double approximately every two years, driving exponential improvement in price-performance
Tension
Disruptive Innovation
Clayton Christensen's disruption theory predicts that incumbents are displaced by cheaper, simpler alternatives that improve along a different performance axis. Moore's Law predicts that incumbents become stronger with each generation because their products improve exponentially. The tension is real: in semiconductor manufacturing, Moore's Law dynamics have concentrated the industry into fewer and fewer players, the opposite of what disruption theory typically predicts. The resolution is that Moore's Law creates an exception to disruption's usual pattern — when improvement requires exponentially more capital, the incumbents' investment advantages compound rather than erode. TSMC isn't being disrupted by smaller foundries. It's pulling further ahead with every generation.
Leads-to
S-Curve
Every technology governed by Moore's Law eventually hits physical, economic, or architectural limits — the upper asymptote of its S-curve. Single-core processor clock speeds plateaued around 4 GHz in the mid-2000s. Traditional DRAM scaling is approaching physical limits. Classical CMOS transistor scaling ended and was replaced by FinFETs, which will be replaced by gate-all-around, which will eventually be replaced by something else. Moore's Law has survived this long precisely because the industry has jumped between overlapping S-curves — each new transistor architecture beginning its own exponential trajectory before the previous one saturates. Understanding S-curves is essential for predicting when the current Moore's Law trajectory will slow and where the next one will begin.
Leads-to
[Wright's Law](/mental-models/wrights-law)
Wright's Law — the observation that costs decline by a consistent percentage every time cumulative production doubles — provides the manufacturing-side explanation for Moore's Law economics. As semiconductor fabs produce more wafers on each process node, yields improve and per-unit costs fall along a learning curve. The two laws are complementary: Moore's Law describes the design-side trajectory (more transistors per chip), and Wright's Law describes the production-side trajectory (lower cost per transistor as cumulative volume grows). Together, they explain why semiconductor price-performance improves faster than either law alone would predict — each chip generation benefits from both design advances and manufacturing learning.
My honest read: Moore's Law is the most important empirical regularity in modern economic history. Not just in technology — in all of economics. No other trend has persisted for five decades, reshaped every industry it touches, and created more aggregate wealth. The law produced an estimated $13 trillion in global semiconductor industry value by 2024, but that figure massively undercounts its total impact — every software company, every internet platform, every AI model, every smartphone app is a downstream beneficiary.
If you work in technology, your career exists because of a trend line Gordon Moore drew in 1965. Understanding its mechanics, its limits, and its likely trajectory isn't optional for serious practitioners. It's table stakes.
Scenario 4
A biotech startup sequences human genomes for $200 each, down from $2.7 billion for the first genome in 2003. The CEO tells investors this decline — roughly 10 million-fold in 20 years — proves that 'biology follows Moore's Law' and projects that whole-genome analysis with clinical interpretation will cost under $10 within five years.
