Intel Is Investing Billions Of Dollars Into This Unproven Technology
Moore's Law, named after Intel co-founder Gordon Moore, predicts that the number of transistors that can be cost-effectively packed into a chip will double every two years.
The more transistors on a chip, the more powerful the chip becomes without using more space or energy. That's why computers have become exponentially faster and smaller over the past few decades.
But unless there's some kind of technological breakthrough, some physicists say we're reaching the scientific limit to Moore's Law. Intel is on the verge of making transistors so microscopic that their performance would be affected by quantum mechanics.
Intel CEO Brian Krzanich thinks a new, very unproven technology called "extreme ultraviolet lithography" (EUV) is the answer. EUV should allow Intel to continue to make its chips smaller, faster, and more efficient. But first Intel has to figure out how to get it to work on a commercial scale.
Unfortunately that's easier said than done.
Ultraviolet lithography is the technology that allows companies like Intel to use wavelengths of light to "etch" their chip designs onto silicon wafers at a scale far smaller than the human eye can see. The smaller the wavelength of light used, the smaller and more exact these etches can be.
Extreme ultraviolet lithography has been on Intel's radar since 1997, seen by many in the semiconductor industry as the next step needed to keep Moore's Law alive and well.
Hope grew high for EUV in 2009 when Dutch firm ASML unveiled a light source that could reliably produce wavelengths of light appropriate for it. Yet there has been relatively little (public) progress since then, which is partly why Intel decided to invest $4.1 billion in ASML last summer.
There have been numerous challenges and roadblocks to making its transistors smaller before, but Intel hasn't been stopped yet. At IDF it showed off a PC running on a next-generation "Broadwell" chip using 14 nanometer (nm) transistors, twice the density of its closest competitors.
Intel promises that Broadwell chip will be released early next year, even while Krzanich admitted on a recent earnings call that Intel hasn't completely mastered the tech needed to build it.
Even so, Intel is working on 10 nm transistors by 2015 and 7 nm in 2017. These are insanely small. In comparison, a human blood cell is about 7,000 nm while a strand of DNA is 2.5 nm.
Intel won't confirm that it will be using EUV to get there, but then it's famous for not discussing its manufacturing technologies, for obvious competitive reasons.
Still, if EUV makes it to market, we could see chips that make 7 nm seem unimpressive by the end of the decade.
If it doesn't, we could see the first break in the trend of transistor density (and the performance boosts that came with it) in four decades.
If that's the case, companies will either have to find other ways to make devices feel faster or consumers will have to get used to a new normal where each new generation isn't a leap forward in terms of raw computing power.
