How Intel Puts Innovation Inside

What happens when your organization's bedrock ideas for growth and innovation just aren't working anymore? That is one scary place to be, and some businesses never find the way out. (Just ask Polaroid.) But other companies regroup in ways that are exciting and outright inspiring.

Right now, one of the most fascinating examples is taking place at Intel's research-and-development labs in Hillsboro, Oregon. There, 600 miles north of company headquarters in Santa Clara, California, Intel is devising future generations of semiconductor chips. This is the world of bunny-suited engineers in "clean rooms," using unimaginably costly machines to experiment with new ways to etch tiny circuits on to silicon wafers.

Sure, overall chip sales are down for the first time in years, largely due to the slump in the personal-computer industry. But Intel continues to pour more than $4 billion a year into R&D. It does so because it believes that recessions are a great time to gain ground on the competition. What's more, Intel's identity is inseparable from innovation -- and has been since the mid-1960s, when the company's cofounder, Gordon Moore, declared that it's possible to keep doubling a chip's computing power every 18 months without jacking up the cost.

Ask Intel's scientists in Hillsboro how they plan to keep Moore's Law working, and they will talk about obstacles never faced before. Most notably, as microcircuits keep shrinking and become ever more powerful, they run hotter and hotter. At some point, even the most advanced cooling systems aren't much use. Instead of producing microcircuits, Intel could end up making furnaces.

In a February 2001 speech, Pat Gelsinger, Intel's vice president and CTO, summed up the problem with frightening clarity. Intel's existing chips generate about as much heat as a hot plate, he observed. Extrapolate recent trends, and in three years, the chips will rival a nuclear reactor. In two more years, a rocket nozzle. And in another two years, the sun's surface. The upshot? If Intel can't figure out a radically different way to build chips, progress will come to a blazing stop.

So what's the remedy?

An ingenious answer emerged this past fall from a team of about 100 Intel researchers in Hillsboro, led by Gerald Marcyk. They came up with the road map for a future chip that will be 10 times faster than what's currently on the market -- without consuming more power or generating more heat. The TeraHertz chip is at least five years away from production, but it's already commanding attention at industry conferences. If it comes to market as planned, the new chip and its successors will ward off the fiery-inferno scenario for many years to come.

As big a story as the new chip is, there's an even bigger story: that of the innovation process at Intel -- how the company manages the search for big ideas. Long before such projects as the TeraHertz chip achieve their crucial "Eureka!" moment in the clean rooms, Intel does four things: It defines its crucial challenge correctly, it puts the right people on the problem, it knocks down the barriers between R&D and manufacturing, and it gives researchers the right mix of autonomy and guidance.

Look closely at each of those steps, and you'll see a shrewd approach that could help almost any company get back on the growth track.

In the case of the TeraHertz chip, Intel kick-started the process in 1998 with a blunt challenge to a small, elite team of scientists. Youssef El-Mansy, Intel's head of logic-technology development, recalls, "I told them, 'In a few years, I'd like you to publish a paper that says, We have designed the fastest transistor in the world.' " But he didn't tell them how to build it. He just set the performance goal. And in doing so, he made it easy -- and in fact, outright necessary -- for the scientists to rethink their basic design approaches.

Before long, team members realized that they couldn't use silicon dioxide, a traditional insulator, in a crucial part of the microcircuit. Using silicon dioxide had worked for more than three decades of Intel design, but the relentless push to make everything smaller now meant that it had to be applied in layers just three atoms thick. It couldn't be applied any more thinly -- and it was already causing electrical current to leak in ways that made it a major culprit for overheating problems.

Abandoning silicon dioxide was an emotionally wrenching decision for Intel scientists, akin to being an automaker and deciding not to use an internal-combustion engine. But new materials held much greater promise. Scientists felt liberated enough from their traditional assumptions to begin testing new alternatives, along with other major departures in chip architecture. The result: Current leakage dwindled as much as 10 thousandfold, indicating that relief was in sight for the overheating crisis.