Gary Baldwin gets revved up when he shows off the brilliantly colored tubes of light that line the corridor of Cory Hall on the UC/Berkeley campus, where the Gigascale Silicon Research Center is housed. Though they look like neon lights, they're actually diodes, like those in your watch or calculator. What excites Baldwin is that, while they're as bright as ordinary incandescent tights, they are nearly 10 times more energy efficient.
It was the kind of visionary research that produced these power-saving colored tights that prompted Baldwin to leave Hewlett-Packard after a quarter-century and take a major pay cut to administer the Gigascale Center. At Berkeley, he says, people have a "fire in the belly" about doing this sort of work. The Gigascale Center aspires to revolutionize the science behind the silicon chip.
Yet, since its inception, immodesty has been the hallmark of the semiconductor field. Moore's law--the axiom proposed by Intel's founder, Gordon Moore, that the number of transistors on a chip will double every 18 months--has held up for more than a third of a century. But as transistors shrink in size below 50 nanometers, 1/20,000th the width of a pinhead, electron behavior becomes much harder to regulate. Still, gigascale proportions are essential if engineers are to build the smaller, cheaper, and more powerful semiconductors of the next generation. The goal is to reach gigascale--a billion transistors on a single chip, nearly 25 times more than is contained on Intel's Pentium 4 processor. It's an immodest goal, and nothing less than revolutionary research is needed to achieve it--"a moon shot," in the words of the 2000 Annual Report issued by MARCO--the Microelectronics Advanced Research Corporation.
But what is MARCO? The Gigascale Center, together with similar centers at three other campuses, represents a path-breaking partnership among the federal government, higher education, and Silicon Valley that is MARCO. MARCO links the research branch of the Department of Defense, 22 universities, and some two dozen firms in the semiconductor industry, including such Fortune 500 names as Intel, Motorola, and AMD.
Such collaborations between higher education and industry have gotten a good deal of press in recent years, much of it negative. "The Kept University," the cover story in Atlantic Monthly's March 2000 issue, paints an alarming picture of multimillion-dollar deals that put higher education in thrall to big business. Like scores of stories in the national and international press, "The Kept University" homes in on the $25-million, five year contract between Berkeley's College of Natural Resources and the biotechnology firm Novartis, which includes support for research on genetically engineered crops.
Nevertheless, such collaborations are necessary. Not even the richest university can afford to underwrite major science projects all on its own. Though government still funds the lion's share of research, industry pays an increasing portion of the bill--$2 billion of research money annually to colleges and universities nationwide, including $27.8 million to Berkeley in 2000.
BEHIND MARCO'S CONCEPTION
In the past, the kind of exploratory research now being done at the Gigascale Center was carried out at such storied places as AT&T's Bell Labs and Xerox PARC (Palo Alto Research Center). Yet while these think tanks made great advances--the transistor, the Xerox photocopier, the laser--the companies that fronted the costs often didn't reap the financial benefits. In the 1970s and '80s, cost-conscious managers radically shrunk the labs' budgets, and there was nothing to replace them.
In 1994, Peter Verhofstadt, then chief scientist at the Semiconductor Research Corporation, an industry-wide research group, became concerned that the knowledge needed to create the next generation of semiconductors simply wouldn't be available. …