Learn it → Apply it → Save as Skill
The resources behind a technology can double on a predictable cycle long after anyone expects it to stop. Moore’s Law is a 60-year case study in what a steady doubling trend does to planning, strategy and patience — and in what happens when that trend finally bends.
The Dr. Wallace’s Favorite Mental Models is designed to turn useful ways of thinking into reusable tools. Each model gives you a fast visual overview, deeper interactive learning, practical application exercises and an AI-ready Markdown skill you can save for future use.
Instead of merely reading about mental models, you can gradually build your own AI skills library: a collection of portable reasoning skills that can help ChatGPT, Claude, Grok and other AI assistants apply the mental models you value.
In 1965, Gordon Moore noticed that the number of components engineers could fit on a chip kept doubling on a steady cycle. That observation became a planning tool: if a resource has doubled reliably for several cycles, it is often rational to plan as if it will double again — while watching for the physical or economic limits that eventually slow any exponential trend.
Important: real doubling periods vary by technology and era. Chip transistor counts doubled roughly every two years for decades — other resources (batteries, bandwidth, storage) double on their own, slower or faster, schedules.
The model is most useful when you need to plan several years ahead for a resource that has shown a genuine, repeated doubling (or halving) pattern, not just one or two data points.
Moore’s Law is an empirical trend, not a physical constant. Treating it well means extending it carefully and checking, every cycle, whether the conditions that drove it are still true.
Confirm the resource has actually doubled (or halved) across several real, comparable cycles — not just once.
Project the trend forward over your real planning horizon, in both the optimistic and conservative directions.
Check whether the economic, physical or competitive conditions that produced the past doubling still hold today.
Plan for the trend continuing, slowing and stopping. Exponential trends do not get a permanent exemption from physics or economics.
Open a use case to see what the model changes.
If your product depends on compute, storage or battery capacity, assume today’s constraints will loosen on a predictable cycle. Build features that are too expensive or slow today but that a few doubling cycles could make practical — and time the launch to when the curve, not just the calendar, gets you there.
Training-compute and AI-accelerator cost-performance have followed their own doubling-style curves, distinct from classic transistor-count Moore’s Law. Separate the trend you are actually relying on (chip density, cluster cost-per-FLOP, model efficiency) from the general folklore of “Moore’s Law for AI,” and check each one against its own recent data.
Chipmakers’ capital spending and pricing power are tied to how close they are to physical scaling limits. A slowing doubling cycle changes the economics of the whole value chain, from foundries to the companies that depend on cheap compute.
Rather than asking “can we build this now,” ask “how many doubling cycles until the constraint that blocks this today is gone.” This reframes a flat “no” into a dated “not yet,” with a rough timeline attached.
Restricting access to leading-edge chips only matters if the doubling cycle keeps moving the frontier. Policy built on the assumption that a technology gap is permanent can be overtaken by a few more cycles of the underlying trend — or undercut faster if the trend is genuinely slowing.
Skills built on top of a resource that is still doubling (cheap compute, storage, bandwidth) are more likely to be reshaped by the next generation of tools than skills built on a resource that has already plateaued.
This is an illustrative example designed to teach the method, using a simplified two-year doubling assumption.
Expand for the distinction between what Gordon Moore actually wrote, what later got attached to his name, and where the trend stands today.
In an April 19, 1965 article in Electronics magazine, “Cramming more components onto integrated circuits,” Moore observed that the number of components per integrated circuit had been doubling every year and projected that pace would continue for at least a decade. In 1975, he revised the estimate to roughly doubling every two years. The name “Moore’s law” was coined and popularized by Caltech professor Carver Mead shortly after that 1975 revision, not by Moore himself.
That figure is commonly traced to David House, an Intel executive, who in 1975 reasoned that combining the doubling of transistor count with faster, more efficient transistors (via what is now called Dennard scaling) implied roughly an 18-month doubling in chip performance. Moore’s own claim was narrower: it was about transistor density, not performance. Moore later said he never stated an 18-month figure, and after Dennard scaling broke down in the mid-2000s, transistor-count growth stopped translating directly into proportional performance gains — which is a large part of why the industry shifted toward multicore chips and specialized accelerators.
This is genuinely disputed rather than settled. Intel’s own leadership has revised the cadence over time (roughly 2.5 years per Brian Krzanich in 2015, closer to 3 years per Pat Gelsinger at the end of 2023), while Nvidia’s Jensen Huang publicly declared Moore’s Law “dead” in September 2022, a claim Gelsinger disputed soon after. The semiconductor industry’s own roadmap body issued its final “Moore’s Law”-framed roadmap in 2016 and shifted toward a “More than Moore” strategy, emphasizing application-specific chips, packaging and parallelism over raw transistor scaling. The honest summary: the original transistor-density trend has slowed and grown more expensive, even as overall computing capability keeps improving through other means.
Moore’s Law is an empirical, economically driven observation about engineering and investment, not a law of nature. It can slow, stop or restart in a different form.
More transistors do not automatically mean proportionally faster chips, especially since Dennard scaling ended in the mid-2000s. Check which specific metric — density, clock speed, cost-per-computation — you actually depend on.
“Doubling every 18 months” (performance), “every 2 years” (Moore’s revised transistor estimate) and “every year” (Moore’s original 1965 estimate) are three different claims from different people. Be specific about which one you are relying on.
Exponential curves that run into physical constraints (heat, leakage, materials) typically flatten into an S-curve. Projecting a straight exponential line indefinitely overstates what is coming.
“Rock’s Law” observes that the cost of building a semiconductor fab has itself roughly doubled every four years. A trend can continue technically while becoming economically out of reach for all but a few companies.
Doubling-style trends show up in batteries, genomics and solar costs too, but each has its own cadence and its own limits. Borrowing the phrase “Moore’s Law for X” is only useful if the underlying data for X actually supports it.
you have only one or two data points. A doubling claim needs several comparable cycles before it is a trend rather than a coincidence.
the resource is a biological, social or physical system known to saturate — population, adoption, or anything bounded by a hard ceiling.
domain specialists already track and publish the actual cadence (as chipmakers now do), rather than reasoning from the 1965 or 1975 figures in the abstract.
This browser tool structures your thinking. It does not send your information anywhere or perform AI analysis.
Save as Skill converts the model into portable Markdown instructions that teach an AI when and how to apply Moore’s Law, what questions to ask, what evidence to seek, what mistakes to avoid and how to structure its analysis.
Save the skills you find useful and organize them into your own reusable library. The Markdown is intentionally model-neutral, so it can be used wherever an AI system accepts uploaded files, project knowledge, custom instructions or reusable text context. Platform capabilities differ, so this is a portable skill file rather than a claim of one-click installation.
Yes — informally, “Moore’s Second Law” (also called Rock’s Law) observes that the capital cost of building a semiconductor fabrication plant roughly doubles every four years. Capability has kept scaling even as the cost of staying at the leading edge has made it an increasingly concentrated industry.
“Moore’s Law for AI” is a loose borrowing of the phrase, usually describing separate trends such as training-compute growth or cost-per-FLOP decline. These trends have their own drivers and are not the same claim Moore made about transistor density in 1965. Treat each one as its own empirical trend to verify, not as an extension of the original law.
Historically, new transistor geometries (such as FinFET and later gate-all-around designs), new materials, and packaging innovations (such as chip stacking) have extended the trend past points where it looked like it would stall. A slowdown in one approach does not rule out a different engineering path picking up the pace again.