In 1965, Electronics magazine asked the head of research at Fairchild Semiconductor Corp. to sketch out the next decade of the fledgling chip industry. So Gordon E. Moore decided to plot the heady growth in the number of transistors on chips, from a mere four transistors in 1961 to more than 200 on a chip then being envisioned. What he found was amazing: The transistor count per chip was doubling every year. Moore boldly predicted in print that it would do so for the next 10 years--the first appearance of what would come to be called Moore's Law.
This was a singular moment in high-tech history. Moore's Law captured the essence of semiconductor technology: relentless, geometric growth in chip power. The results eventually would make electronics the world's biggest industry and Moore's next startup, Intel Corp., the world's richest chipmaker. Considering that he then steered Intel to the pinnacle of high technology, "it's kind of funny that Moore's Law is what I'm best known for," says Moore, who is now chairman emeritus. "It was just a relatively simple observation."
Even to some fellow chipmakers, Moore's Law seemed too good to be true--and it was. In 1975, the increase in chip-transistor counts was slowing, so Moore plotted the trend again. He decided that chip densities would double only every two years. But this time, he was too cautious. The actual growth rate over the past 37 years has essentially split the difference, doubling every 18 months or so. This year, says Moore, the industry will produce about 1 quintillion transistors. "That's at least as many as all the ants on earth," he marvels.
The transistor count will continue soaring for the next 15 to 20 years, as chipmakers shrink circuit linewidths from 0.35 microns today to 0.07 microns in 2011. That would enable Intel to make chips 200 times more powerful than its current speed champs.
This rapid shrinking in feature size can't go on forever. "There are fairly fixed limits, such as the atomic structure of matter, on how tiny transistors can become," Moore explains. But physical limits won't signal an end to silicon innovation, he argues. The chief reason is the often overlooked corollary of Moore's Law: The cost of a given amount of computer power drops 50% every 18 months. Each time that happens, the market explodes with new applications that hadn't been economical before. "The market is phenomenally elastic," says Moore. "Anytime we cut prices, lo and behold, a few months later the market expands." For instance, in a decade it will be possible to cram the power of today's supercomputers onto a chip costing a few hundred bucks.
HIDDEN IN THE QUOTIDIAN. But tomorrow's ultra-tiny transistors will also be harnessed to make smaller and, thus, cheaper microprocessors--because most applications won't need a billion-transistor chip. Historically, the biggest, latest-generation chip that Intel produces accounts for less than 2% of total microprocessor-unit sales. The others hide in everyday products. For example, Moore notes, videocassette recorders have a half-dozen microprocessors, most cars now have at least a couple of dozen, and some luxury cars have triple that number.
Consider what might happen when $1 chips have enough zip to leave today's workstations choking in their dust. "A robot that cleans your house seems a reasonable thing to expect," says Moore. "And I'm sure that silicon intelligence is going to evolve eventually to the point where it'll get harder and harder to tell intelligent systems from human beings."
Pressed to cite more possibilities, Moore demurs. "In the late 1970s," he explains, "I looked around the house for all the potential applications for microprocessors I could find. I came up with 85 potential new uses, like light switches," he recalls. "A couple years later, I took another look, and more than half of the applications then on the market--just two years later--were things I hadn't even envisioned. So my batting average at making predictions is not especially good."
No apologies necessary, Gordon. You came up with a beaut 32 years ago.