Chip Makers Brace For Shake-Up Soaring Demand, Cost Trigger Consolidation

The computer chip industry, since it began in 1959, has amazed and reshaped the world.

But in economic terms, it’s where the automobile industry was in 1910.

Demand continues to rise sharply for computer chips as improved performance inspires new uses for them, just as better engines and roads helped sell cars during the second and third decades of the century.

But production costs are rising even faster. New chip fabrication plants cost $1 billion or more.

The industry is nearing the point that automakers reached when mass production forced huge capital expenditures that most couldn’t afford. The number of manufacturers dropped from approximately 250 in 1908 to 44 in 1929.

“It’s consolidation,” said Michael Detourzios, director of the MIT Laboratory for Computer Science. “Look at the automobile manufacturers … We have 20 automobile manufacturers worldwide and that’s it.”

Some chip industry leaders say their business won’t shrink the same way. Supply can’t keep up with demand now, they point out, and while that condition persists, companies will have no trouble finding capital for new plants and development.

“If you’re a chip company, money is hanging on trees,” said T.J. Rodgers, chief executive officer of Cypress Semiconductor Corp.

But many feel major change is near for the industry. And change in semiconductor economics will eventually influence how computers are made and software is written.

From the start, the chip industry has possessed a wonderful trait that Gordon Moore, a physicist who co-founded and is now chairman of Intel Corp., identified in a 1965 magazine article.

He noticed that manufacturers had been able to double the number of circuits on a chip every year, causing an exponential leap in power each time. That leap in power meant the cost per circuit was cut in half each time.

The observation became known as “Moore’s Law” and has been a goal for chip engineers to sustain. Around 1976 or 1977, the doubling slowed to every 18 months and some experts say it has recently slowed again.

“When Gordon wrote his paper in 1965, it was like being in the diamond fields of South Africa,” said Dan Hutcheson, president of VLSI Research Inc., a research firm that specializes in semiconductors. “But now you’ve got to dig these huge mines to get the diamond. We’re running into technology barriers that slow us down.”

It remains, of course, a spectacular achievement that the number of circuits on a thumbnail-sized chip has increased from 1 to 64 million. They are a maze of submicron dimension, thousands of times less than the width of a human hair and invisible except under special microscopes.

With current design, testing and manufacturing techniques, most engineers believe the doubling of circuitry can continue at least another decade.

But can all the chip makers stay in the game?

The equipment to make circuits that small is also doubling with each chip generation and research and testing complexities add up. Huge volumes of chips must be sold to provide a return on such investments.

“If you have the same amount of time to amortize the costs, but the costs are going up, the only variable is to increase volume,” said Mark Ellsberry, vice president of marketing for the U.S. chip subsidiary of Samsung, the leading maker of memory chips. “The only way we can increase it drastically is to increase market share.”

High costs have driven many companies to band together for new development. In recent weeks, Motorola joined three other large companies - IBM, Toshiba and Siemens on memory chip research and Advanced Micro Devices acquired rival Nexgen to take advantage of future microprocessor designs.

Speaking at an engineering conference in July, Gordon Moore warned financial, rather than technological, difficulties will now control the rate of the industry’s progress.

“We just will not be able to go as fast as we would like because we cannot afford it, in spite of your best technical contributions,” Moore said.

This realization has started to permeate the industry. Executives in recent months have stopped talking about improvements in chip prices and focused on gains in productivity.

“It’s a change in the sense that we would like to be able to talk to our customers in terms of the value that we provide, not in terms of what the cost per chip is,” said Gilbert Amelio, chief executive officer of National Semiconductor Corp.

With so many transistors available on a chip, it becomes increasingly possible for one chip to perform functions that previously took two or more.

It is even possible to blur the distinction between memory chips and processing chips, changing the way performance has traditionally been measured, said Jim Glaze, vice president of technology for the Semiconductor Industry Association.

“Historically, we tended to use the cost per transistor,” Glaze said. “But when you start talking about higher levels of integration, that creates a lot of productivity. So maybe price will be higher, but the productivity will be higher yet.”

The construction of the personal computer illustrates the change. PCs have a microprocessor chip that performs the calculations or changes on a piece of software. It works with memory chips, which store portions of a program while moving in and out of the microprocessor.

Future microprocessors, including the Pentium Pro that Intel recently began selling, achieve a performance gain in part by devoting a portion of the available transistors to memory functions. Samsung and other memory chip makers are starting to talk about adding processing capabilities to their products.

“We’re now getting to the point where an integrated circuit is a complete system,” said Gene Frantz, a Texas Instruments Fellow and leader of digital signal processing chip design.

That will affect the work done by the customers of chip makers, including manufacturers of computers and cellular phones.

For instance, the cellular phone is already near the point where all of its functions can be done on a single chip. That leaves little for a cellular phone maker to do besides pick a chip to use, write software and decide what the phone is going to look like.

“We have to find a way to combine all this together on one chip and still give our customers flexibility to personalize the circuits,” Samsung’s Ellsberry said.