Lives of the Silicon Saints


Microcosm: The Quantum Revolution in Economics and Technology, by George Gilder, New York: Simon & Schuster, 426 pages, $19.95

Reading George Gilder's Microcosm is sometimes disorienting. Fascinating accounts of the development of semiconductor technology and the chip industry are punctuated by passionate, if irrelevant, assertions about free will, the nature of the universe, the brain/mind dichotomy, and the role of faith. Given the tacked-on feel of these philosophical ruminations, I wondered why they had been included; Gilder's acknowledgments at the end of the book finger a novelist friend of his as the culprit who "persuaded me that the heart of the book must be the physics and philosophy, as well as the technology, of the microcosm."

It would be a shame if this decision chased away readers interested in understanding the dynamics of the chip business, because Gilder does a truly superb job of describing both the underlying technological imperatives and the personalities and disputes that have created the world semiconductor industry. Unfortunately, most of the book's reviewers seem to have fixated on the "big-picture" material at the beginning and end, where the book is weakest; Robert Wright in The New Republic, for example, performed a masterly dissection of Microcosm's philosophical flaws, while practically ignoring the middle 80 percent of the book.

The central figure in Microcosm is Carver Mead, a Caltech professor who intellectually pioneered the ever-greater miniaturization of components on silicon chips. In 1968, the conventional wisdom said that further increases in the density of logic devices on a chip would cause runaway waste heat and crippling material defects. Mead, by "listening to the technology and finding out what it is telling you," showed that shrinking the size of switches and packing them in more closely would actually reduce the amount of power needed to run a chip, thereby reducing waste heat. Production costs would be slashed and reliability would increase.

Given this technological premise, the future trajectory of the semiconductor industry could be envisioned. First, as more and more logic gates were etched on a chip, yielding awesome increases in computing speed, products that put as many functions as possible on a single chip would drive out products that relied on macrocosmic connections between separate chips. Second, computing power would diffuse widely, ending up inside ovens, cameras, and toys.

Third, the bottleneck of innovation would not be production techniques but rather designing the layout of the millions and millions of switches and data pathways etched onto the silicon. As a result, the distinction between hardware and software would blur, and most of the value of a chip would reside in its design, just as most of the value of a book is in its information content rather than the physical substance of its paper and binding.

Gilder interprets the history of the chip industry as the working out of "Mead's Laws," as he calls them, much as orthodox Marxists interpret the history of mankind as the relentless unfolding of the laws of class struggle. One difference is that technological determinism is a lot more plausible for explaining 20 years of the semiconductor industry than it is for explaining hundreds of thousands of years of economics, politics, art, social mores, and religion. Another is that Gilder romanticizes the entrepreneurial and scientific heroes who have struggled to bring forth the fruits of Mead's vision; he vividly describes the self-doubt, the stubborn persistence, and the heady rush of exhilaration that come from pioneering new technologies in the face of industry skepticism and financial peril.

The religious quality of this account practically hits the reader in the head. The chapter introducing Mead is entitled "The Prophet," and the various characters are described in a manner reminiscent of the lives of Christian saints and heroes, their frequently difficult private lives and inner motivations unearthed along with their scientific and business achievements. While Gilder maintains a sympathetic tone even toward those whose views place them in opposition to the One True Path, he vents his rhetorical spleen against the inferior technologies they champion, in the great tradition of hating the sin rather than the sinner. This approach makes the narrative much more engaging by giving the reader a rooting interest in what might otherwise seem dry technical disputes, even if it gets a little tiring to see big computers constantly disparaged as "lordly mainframes."

As for the technical accuracy of Gilder's discussion, it is pretty easy for even a nonspecialist to tell which parts are controversial and which relatively indisputable. Many of the latter concern disputes in which the Mead view was vindicated by market outcomes. In any event, Gilder includes a fairly extensive annotated bibliography that can help the interested reader do his own exploration of the issues.

Where Microcosm is somewhat disappointing is in its discussion of business strategy, economics, and public policy. One problem is that Gilder's preference for dogmatic, strongly worded statements undermines his case, and I say that as one who is in basic agreement with most of his views. For example, Gilder flatly asserts that "efficiency in manufacturing any product increases some 30 percent with each doubling of accumulated volume." Aside from missing the crucial distinction between industry volume and the volume of a single firm, this assertion is false. Such "learning curves" for some products lead to even steeper efficiency gains, while other products display no volume sensitivity to speak of. The learning curve is indeed important in the semiconductor industry, but Gilder feels compelled to make the bolder, more general statement at the expense of his credibility.

He complains that business schools have been discouraging executives from selling below initial cost to build up sales volume and learning efficiencies. At least in my neck of the woods, we teach our students that this strategy, which was popularized by consultants in the 1970s, is sometimes essential, as in the aircraft industry, and sometimes useless. In particular, the experience of Texas Instruments in the watch and calculator businesses, which Gilder has the nerve to cite in favor of the "race down the learning curve" strategy, shows that if learning effects "spill over" to competitors then there is little to be gained by selling below initial cost; all you do is subsidize the learning of your rivals.

The realm where Gilder's message is most important is that of public policy, where techno-nationalist hysteria threatens to overwhelm the fundamental truth that government-backed schemes to improve the commercial performance of particular industries have a disastrous record and no clear rationale. In pursuit of "competitiveness," our policymakers negotiated a cartel agreement with Japanese memory-chip producers to restrict output and raise prices; this deal bailed the Japanese out after they had lost $3 billion through excess investment and production while badly damaging domestic producers of computer systems who use memory chips in their products.

Now we have Sematech, a government-sponsored (and partly funded) consortium of U.S. chip firms aimed at developing new fabrication techniques. Waiting in the wings are expensive proposals to pump tax dollars into a domestic high-definition television (HDTV) industry, the alleged next wave in consumer electronics.

It is here where Microcosm is most frustrating. The problem is partly expository, as Gilder scatters arguments in different places instead of systematically addressing the issues. Some of his points are quite telling, but he seems throughout to be preaching to the choir, not really trying to influence the potentially persuadable citizen or policymaker who is afraid of Japanese encroachments on America's technology base. Furthermore, he sometimes doesn't play fair, as when he constantly lauds Micron, a firm that he profiled in a previous book, without mentioning that this company was one of the prime movers behind—and chief beneficiaries of—the U.S.-Japan chip agreement.

By contrast, Gilder does systematically present the case, advanced by Charles Ferguson, for an industrial policy in semiconductors. This case is based on the notion that the American chip industry is plagued by excessive entrepreneurship, with employees quitting to start new firms based on technology developed at their old companies. These startup firms allegedly get in trouble because they are not big enough to handle the multiple demands of development, manufacturing, and marketing, and because they too are plagued by a horde of imitators, some formed by defectors from their own ranks. They end up selling their technology to foreign companies just to stay afloat. According to this view, the Japanese, with large, integrated producers and close government supervision, end up profiting from the research and development expenditures of the United States.

Part of the answer to such arguments is to be found in the industry experience since 1985, which Gilder cites in another context. An explosion of startups, many focusing on the design of "application-specific" chips, sometimes relying on outside firms to actually manufacture their products, have proven to be highly profitable. Such companies as Cypress, Chips & Technologies, and Performance have thrived even as pundits have wrung their hands over the purported demise of the United States semiconductor industry. These startups have had a higher rate of success than did their predecessors during the industry's heroic period of the 1970s.

Gilder's cleverest argument is that Japan has been less successful in semiconductors than it has in such industries as machine tools, automobiles, motorcycles, and televisions. In these other industries, the Japanese were more entrepreneurial than their U.S. competitors, starting up many more firms and running rings around their large American rivals. Only in semiconductors, where U.S. entrepreneurs have started more new firms than the Japanese have, has America held its own. Efforts to force consolidation and inhibit entrepreneurship are thus exactly the wrong way to enhance "competitiveness," even assuming that this goal is appropriate and meaningful for government policy.

We are going to be facing more and more pressure for various interventionist policies from the "industrial policy elite" in the academy and in industry. The semiconductor industry, as an important and symbolically potent part of the economy, is likely to be a key battleground. Anyone interested in the debate or the industry ought to find Microcosm an enjoyable source of information and inspiration.

Steven R. Postrel is an assistant professor at the Anderson Graduate School of Management at UCLA.