// moore's law

The Law That
Built the World

In 1965, Gordon Moore observed that transistor counts doubled roughly every year. This observation became a self-fulfilling prophecy — and the most important economic force in tech history.

1965
Moore's first paper
~2yr
Revised doubling period (1975)
2025
Law is slowing — but not dead
// the data

Is Moore's Law dead? This page presents the data to judge for yourself — an interactive chart comparing actual transistor counts against the ideal doubling curve from 1971 to 2026, alongside the history of Gordon Moore's original observation, Dennard Scaling, and what the semiconductor industry is doing next.

Moore's Law: Predicted vs. Actual

What Gordon Moore said

The number of components per chip that minimizes cost doubles roughly every year — and that rate of increase will continue for at least a decade. — Gordon Moore, paraphrasing his 1965 paper "Cramming More Components onto Integrated Circuits"

Gordon Moore was a chemist turned engineer who co-founded Fairchild Semiconductor in 1957 — one of the pioneering companies in silicon integrated circuits. In 1965, Electronics Magazine invited him to write a piece predicting the next decade of chip development. Moore plotted just five data points (the transistor counts of chips from 1959 to 1964), drew a line through them, and extrapolated forward. What he saw was a doubling roughly every year. He published it almost as an offhand observation — "I had no idea this was going to be an accurate prediction," he later said. In 1968 he co-founded Intel. In 1975, at an IEEE conference, he revisited the data and revised his forecast to roughly every two years — because one of the key drivers of the original pace, rapidly growing die sizes, was beginning to slow. As he later put it, that factor had been contributing about half the progress: once it faded, the overall rate had to come down. Caltech professor Carver Mead popularized the term "Moore's Law" in the early 1970s — a name Moore himself never much liked.

Moore's Law was never a law of physics — it was a business strategy that became a roadmap. Chip companies planned R&D cycles around it, creating a self-fulfilling prophecy for 50 years.

Since around 2010, the pace has slowed — and by 2016 the industry's own roadmap had abandoned Moore's Law as a guide. Physical limits — atoms are only so small — mean transistors can no longer shrink as fast. The industry has responded with 3D stackingInstead of shrinking transistors further, chips are now stacked vertically like floors in a building. AMD's 3D V-Cache and Intel's Foveros technology bond multiple dies on top of each other, dramatically increasing transistor density without needing a smaller process node., new materialsSilicon is reaching its physical limits. Gate-All-Around (GAA) transistors — already in production at 2nm — wrap the gate on all four sides for better control. Further out: carbon nanotubes, gallium nitride, and 2D materials like MoS₂ promise to extend scaling for another decade., and heterogeneous architecturesRather than one large general-purpose chip, modern designs combine specialized dies: CPU cores, GPU cores, NPUs for AI, and memory controllers — each optimized for its task — connected via high-speed interconnects on a single package. Apple's M-series and AMD's EPYC chiplets are prime examples..

Transistors per chip vs. Moore's Law projection
// logarithmic scale · proportional time axis · highest transistor count chip per year (CPU or GPU) · dashed line = ideal doubling every 2 years
// related trends

Three Trends Linked to Moore's Law

Dennard
Scaling (died ~2006)
Smaller transistors used to mean lower power. This stopped working, causing the "power wall" and the end of clock speed growth.
Koomey
Law (energy efficiency)
Computation per unit of energy doubled every 1.57 years historically (1946–2000). Since then the rate has slowed to ~2.6 years — still improving, but no longer as fast.
Amdahl
Law (parallelism limit)
Adding more cores has diminishing returns if software isn't parallelized. The challenge of the multicore era.
// frequently asked

Common Questions

Moore's Law is slowing but not dead. Since around 2010 the pace has slowed, and by 2016 the industry's own roadmap had abandoned it as a guide. The industry has responded with 3D stacking, new materials like Gate-All-Around transistors, and heterogeneous architectures to continue improving performance.
Gordon Moore co-founder of Intel, first published the observation in 1965 in Electronics Magazine, plotting just five data points and extrapolating the trend. He revised the doubling period from one year to roughly two years in 1975.
Dennard Scaling was the principle that as transistors shrank, power density stayed constant — meaning chips got faster without using more power. This stopped working around 2004, causing the thermal wall and the end of clock speed growth.
Moore's Law started visibly slowing around 2010. By 2016, the semiconductor industry's own International Technology Roadmap for Semiconductors (ITRS) had abandoned Moore's Law as a planning guide due to physical limits at small transistor sizes.