THE OBSERVATION

A prediction that shaped an entire industry

roughly doubling every two years

In 1965, Gordon Moore observed that the number of transistors that could be economically packed onto a chip was roughly doubling every year or two, and predicted the trend would continue. That observation, later named Moore's Law, became a target the whole industry organized itself around.

Moore's Law was originally an observation, but the industry chose to treat it as a schedule.

DECADES OF DOUBLING

A straight line, until it isn't

For decades, transistor count per chip climbed in an almost perfectly straight line on a logarithmic scale, doubling again and again on schedule. In recent years that line has started to bend, as each new doubling takes longer and costs more to achieve.

The bend in this curve is the central story of the chip industry today.

WHY IT WORKED SO LONG

Smaller meant better, in every direction at once

smaller, faster, cheaper, all together

Shrinking a transistor made it faster, made it use less power, and let more of them fit on the same die, which lowered the cost per transistor. That rare alignment, where smaller was simply better along every axis at once, is what let scaling continue for so many decades.

Few technologies ever improve on every dimension at once, for as long as this one did.

WHY IT'S BENDING NOW

Physics and economics, both pushing back

Atomic-scale limits Transistor features are now only tens of atoms across

Rising cost per node Each new process node costs more to develop than the last

Diminishing density gains Each new node adds a smaller percentage improvement than before

Both the physics and the economics of shrinking further keep getting harder.

MORE THAN MOORE

Improving the chip without just shrinking it

chiplets, specialization, new materials

Instead of relying purely on smaller transistors, the industry increasingly improves chips other ways: chiplets and 3D stacking from the last part, specialized accelerators built for one task, and new materials that improve on silicon in specific roles.

When shrinking gets harder, the industry finds other axes to improve along.

UNIT 47 STUDY COMPLETE

📈

Ready for the Fab Challenge?

You've covered the trend that built this industry, why it worked for decades, and why it's slowing.