Now some researchers are predicting that Moore's Law may run out of gas by around the year 2005. Unless the industry adopts new ways to shrink the size of circuits on a chip, they say, it may not be possible to keep boosting the power of chips. This could mean that the industry will be forced to say goodbye to continuous growth and profits that have made the chip business an engine of the world's economy.
Wednesday, a consortium that includes Intel, IBM and Motorola and three national laboratories plan to unveil a milestone in the quest to extend Moore's Law for at least another decade.
The consortium plans to demonstrate a prototype of a chip-making machine that it says poses the best hope for driving the next generation of growth.
The device, the fruit of more than $250 million in contributions from the companies, uses a technology called extreme ultraviolet lithography to trace ultrasmall lines of circuitry on silicon wafers.
Lasers and mirrors
Where traditional lithography uses lenses to focus beams of light into precise patterns, EUV bounces high-powered lasers -- developed initially by the military -- off a series of mirrors, which were originally designed by astronomers to reduce the image to the tiny sizes required by new chips.
Backers believe the new technique will ultimately be able to create circuits that are as small as 20% to 7% the size of the smallest possible with conventional lithography.
"It looks like the right technology," says Kevin Krewell, a senior analyst at the chip-research firm MicroDesign Resources. "You've got all the major players supporting it. It looks real solid."
Of course, not everyone agrees on the severity of the problem, let alone the solution. Limits to Moore's Law have been repeatedly predicted in the past, only to fade as makers of chip-making gear found ways to tweak their technology.
In the early 1990s, for example, naysayers said it would be impossible to assemble chips used in cellphones and products because they required circuitry as small as 0.25 micron, about 400 times thinner than a human hair. Last week Intel announced plans to build chips with 0.13 micron lines, using largely conventional processes.
But once again, researchers and analysts say they are running out of ways to extend today's lithography, which uses techniques akin to photography.
"It's like trying to draw an intricate pattern with a blunt pencil," says Linley Gwennap, principal analyst at the Linley Group, a technology-research firm in Mountain View, Calif.
Starts at 0.07 microns
The prototype "stepper," as chip lithography machines are called, is being formally unveiled at a ceremony at Sandia National Laboratories in Livermore, Calif. The system, not slated to go into commercial production for at least four years, should start at 0.07-micron circuitry. But some experts think that EUV dimensions could shrink to at least 0.03 micron, and perhaps all the way to 0.007 micron.
EUV hasn't been free of controversy, in part because of U.S. government involvement. Much of the laser technology used by the EUV consortium came from research done to design President Reagan's strategic defense initiative, by organizations such as Sandia and Lawrence Livermore National Laboratory. Mirrors used in the process stem from work on the Hubble space telescope.
Art Zafiropoulo, chief executive of Ultratech Stepper Inc., a San Jose, Calif., maker of chip equipment, has long opposed the technology transfer. He predicts the fruits of that government research will inevitably reach the three foreign companies -- Japan's Nikon Corp. and Canon Inc., and ASM Lithography Holding NV of the Netherlands -- that dominate a lithography market estimated to be worth $10 billion by the end of the decade. Handing over research to such companies means "the complete balance of trade for that $10 billion will have evaporated from this country," Zafiropoulo argues.
Others raise technical concerns. "There's still some scientific and engineering risk that it will not be feasible at all" to build chips using EUV, says Mark Pinto, chief technology officer at Agere Systems Inc., a company that was spun off from Lucent Technologies Inc. and has ties to its Bell Labs unit, which pioneered research in EUV.
"If it was a no-brainer that it was going to be good for everybody, you'd find more people willing to stand up and invest in it," says Pinto.