The entrance to the headquarters of Sema-tech, the research consortium of chip manufacturers based in Austin, Tex., exudes a remotely military air. Vehicles entering the area must stop at a security gate, and once inside the main structure visitors must attach ID tags to their lapel while awaiting an escort to the building’s inner sanctums. Such precautions reflect the seriousness of an enterprise touted as essential to the nation’s security; it is chips, after all, that power the information economy. Incorporated in 1987 as a U.S. experiment in industrial policy, the venture was conceived to help stem semiconductor makers’ precipitous loss of market share to the Japanese. The plan was to pool the industry’s resources to refine the complex chip-manufacturing processes-an area of traditional American weakness. A clear bid to save a declining industry, the notion seemed radical partly because it arose amid the free-market rhetoric of the Reagan administration-and especially because it called for matching funds from the federal government. But the idea gained high ground when a report from the Department of Defense outlined an alarming future of relying on foreign sources for the brains of its high-tech weapons. President Reagan signed the bill authorizing $100 million in annual funding for Sematech that year.
Nearly a decade later, many have judged the experiment a success: the U.S. semiconductor industry now holds a slightly larger share than its Japanese rivals and, more importantly, is seen as a long-term world-class player. While even Sematech boosters do not claim the consortium deserves full credit for this turnaround, observers such as Michael Borrus, codirector of the University of California’s Berkeley Roundtable on the International Economy, say it has clearly played a role by “convincing traditional rivals to cooperate.” The industry’s change in fortune recently prompted Sematech members to decline further federal funding.
William J. Spencer has led the organization since 1988, having arrived with the credentials essential to such a high-stakes position. Formerly head of research at Xerox’s Palo Alto Research Center, one of the nation’s premier industrial R&D labs, he began his career at Bell Labs in 1959. Trained as a physicist, he says his career took a decisive turn when he developed a device for the first communications satellite in 1961. Inspired when the satellite actually functioned against great odds, he abandoned pure physics to pursue the practical challenges of engineering, working on microelectronics and systems development at Sandia National Laboratory before moving to Xerox.
Professing, in typically modest fashion, that he’s “made enough mistakes,” Spencer has announced that he will relinquish responsibility for Sematech’s daily operations as soon as a successor can be named (he will stay on as chairman of the board). In an interview with managing editor Sandra Hackman, he reflects on what Sematech has accomplished and where it has fallen short, and what lessons its experience holds for other cooperative ventures as well as for the overall U.S. research effort.
TR: How did this government-aided consortium manage to get off the ground? And how did the two vastly different cultures avoid insuperable conflicts?
SPENCER: Several factors were critical. First, the U.S. semiconductor industry was going in the same direction as the U.S. consumer electronics industry-into the tank. The Defense Science Board did a study indicating that by 1995 the United States would have less than 20 percent of the semiconductor market, and that eventually the domestic industry wouldn’t exist. Most people would argue, as the Defense Science Board did, that you should not have to depend on foreign sources for something as crucial as semiconductors. In fact, every nation has said, if we’re going to be a strong industrial player, we’ve got to have a position in this basic industry (which will probably grow from $200 billion now to $2 trillion in 15 years). Europe is spending a lot of money on maintaining a semiconductor industry, and so are other economic regions.
Second, industry leaders came together, realizing that even though they didn’t know what ought to be done they had to do something. Fortunately, Charlie Sporck, then CEO of National Semiconductor, took a year’s leave from his job to convince the country that launching Sematech was critical. The venture gained even more credibility when Bob Noyce, co-inventor of the integrated circuit, eventually agreed to run Sematech for its first two years.
Finally, the government not only matched industry’s funding but did so in a very hands-off way: it sent the funds and put none of the usual restrictions on them. The government sponsor, the Defense Advanced Projects Agency, did work with the consortium to establish annual goals and asked for a relatively brief report at the end of each year. But the agency did not micromanage-it essentially asked industry to manage a federal grant. The General Accounting Office did send representatives here for the first five years, but they left after seeing that the program was well run.
Proof of its success is that in 1995 semiconductor manufacturers told the government that we wanted to end federal support in 1996. Although we couldn’t have made the same progress without it, we felt we could no longer justify that role given the industry’s turnaround, and that we could continue progress on our own. And indeed the government has just concluded its financial investment in Sematech.
TR: Why did you decide to work intensively with the companies that produce chipmaking equipment, even though they are not formally members of the consortium?
SPENCER: The thinking was that if the United States is going to have a strong semiconductor industry, it had to have access to the very best equipment. It’s the ever more complex machinery that keeps the semiconductor productivity engine turning. You certainly don’t want to depend entirely on foreign sources for such critical equipment.
And the same thing that had happened to the U.S. semiconductor industry was happening in the equipment industry: the U.S. share had declined from nearly 100 percent in the 1970s to 40 percent in the late ’80s. That convinced us to try to foster cooperation among the equipment firms, the materials industry, and the semiconductor companies to ensure that we get new manufacturing technology as cost-effectively as possible. We devoted some $900 million in R&D support-more than half of Sema-tech’s funds-to that effort.
TR: And are U.S. manufacturers now buying more domestic equipment?
SPENCER: They are. Our goal was that semiconductor manufacturers would use at least 50 percent domestic equipment, and today it’s well over 60 percent.
Overall, the turnaround in the equipment industry has been even more dramatic than among semiconductor manufacturers: U.S. equipment makers now account for about 55 percent of the worldwide market, the Japanese hold 35 or 40 percent, and the Europeans control a little less than 10 percent.
TR: Are sales largely domestic as well?
SPENCER: No. If you look at any U.S. high-tech company today, more than 50 percent of its sales occur outside the United States. My guess is that by 2050, those companies will be selling 90 percent of their product outside the United States. Much of the present foreign market is in Europe, but future growth will occur largely in the Pacific Rim, Africa, and South America.
TR: Do those companies establish new factories here or abroad in order to supply those expanding markets?
SPENCER: Both. Every manufacturer worldwide recognizes that if you’re going to sell a chip-etching or lithography tool in another country, you’re probably going to have to have a factory there, and this industry is no exception.
TR: If Sematech aids manufacturers who expand in other countries, how does that benefit the United States?
SPENCER: A large percentage of profits from those ventures return to the United States, and a lot of high-paying service jobs are created as a result, such as in R&D. That’s why maintaining a strong manufacturing base is so important, and why every major economic region has a program designed to attain or retain one. If U.S. boards of directors are in the driver’s seat, they decide where to build factories and do that R&D.
TR: So is the United States seeing a net gain in jobs in this industry?
SPENCER: Yes, but the situation is complicated. One of our young engineers came to see me the other day because he was very excited to report that he was taking a job in Malaysia-he and his wife think it’s going to be a great lark to live there for the next four or five years. If that company were a Malaysian company rather than a U.S. company, my guess is that he wouldn’t have gotten the job.
A company gets capital from wherever it’s cheapest-maybe in Tokyo, maybe in London, maybe in New York. Similarly, a firm will get qualified people wherever it can to run a multibillion-dollar manufacturing facility. And since U.S. universities put out the most qualified graduates, more and more of those people are taking jobs worldwide. When I go to Taiwan or Korea today, it’s almost like having a reunion with people I’ve worked with at Xerox and AT&T. In the future, instead of moving from Detroit to Dallas, more and more people will move from San Francisco to Kuala Lumpur.
TR: And has the government’s total investment of $800 million in Sematech paid off?
SPENCER: The U.S. companies that have become stronger as a result have more than reimbursed the government in the form of taxes. In fact, one member company reported that it paid four or five times the government’s share of Sematech’s funding in taxes last year.
And of course the U.S. government, particularly the Department of Defense, is now assured of a U.S. source for key electronic components, and U.S. computer manufacturers can more easily decide when to introduce the next generation of microprocessors and memory chips.
TR: Do chip makers think they are getting a good return on their investment in Sematech?
SPENCER: Last year our members reported more than a 400 percent return-a record. The chipmaker that gained the most from Sematech reported a sixfold return on its dues. Because these returns are so strong, member companies recently voted to increase their dues by 30 percent.
TR: How does a company measure its gain?
SPENCER: Sometimes it’s easy. Suppose the consortium works to improve a piece of equipment, and that equipment reduces the cost of a chip manufacturer’s $2 billion product flow by 1 percent. That represents a $20 million benefit right there. And the company avoids having to set up its own improvement procedures, which can also cost millions.
Sometimes the benefits to our members are less tangible but still invaluable. For example, the employees they send to Sematech for a two-year stint work on the most advanced equipment and carry that information back home. Those employees also make contacts that might otherwise take 10 or 20 years to establish. Finally, our members gain access to the results of Sematech’s frequent meetings with European and Japanese consortia and share information from technical conferences.
TR: What does it cost a company to join?
SPENCER: We put a floor on joining: you couldn’t get in for less than a million dollars. That caused a lot of complaint among smaller firms. But if you are a $200 million company and pay $1 million to Sematech, you get access to the same experiments and information that a larger company does by contributing $20 million.
TR: It sounds as though not too many small companies have joined Sematech.
SPENCER: Unfortunately not, although big companies with small semiconductor divisions-such as Hewlett- Packard, Digital Equipment Corp., and Rockwell-are members. That experience is typical: recently established U.S. efforts to develop advanced batteries and the next generation of automobiles, not to mention Japanese and European consortia, have seen a similar reluctance on the part of smaller companies, even though the benefits to them could be great.
TR: Might smaller firms fear divulging information about their processes, not to mention their problems?
SPENCER: That’s entirely possible: we find that the employees companies send to work at Sematech are initially reluctant to talk about their problems and capabilities because they think they are unique. But after a short while they find that other companies are facing the same challenges and they can solve them and become more efficient and effective if they work together-it’s an amazing revelation. In fact, if you ask our member firms about their major benefit from Sematech, they won’t say the improved equipment or the advances in employee training but the better communication-both horizontally among member companies and vertically between member companies and suppliers.
TR: Does that kind of cooperation make Sematech a model for other industrial consortia?
SPENCER: We think it does-the U.S. car and battery consortiums are following our lead. And the imperiled textile industry, which is confronting many of the same problems we faced, recently created a consortium in collaboration with universities and national labs.
What’s fascinating is that the Japanese have formed about four or five consortia in the past year that are almost identical to Sematech. And Europe has begun to move its electronics consortia in a similar direction.
TR: How so?
SPENCER: Cooperative European R&D at first focused largely on developing products such as VCRs or memory chips, but those efforts didn’t prove very successful. So the Europeans are now concentrating on infrastructure-developing new equipment and funding research on materials. They have some problems we don’t: it’s hard enough to get various U.S. companies to work together on any kind of project, but imagine trying to convince 15 nations in Europe to do so. I don’t envy them.
TR: Should foreign companies be allowed to join Sematech?
SPENCER: That’s a major topic of discussion for the board right now: should we remain a national consortium? I think the issues we face today, whether related to economics, the environment, or health, are planetary issues. You can no longer do anything in one country that doesn’t have an impact on the rest of the world. What’s more, if Samsung and Hitachi build factories in the United States, the Environmental Protection Agency and the Occupational Safety and Health Agency are going to be investigating them, so companies might as well develop a common set of standards. And Sematech is a model for doing that because we’ve been able to find a way to cooperate yet compete very strongly in the market.
And chip makers from different countries are in fact starting to cooperate on setting environmental and health goals and complying with regulations. For example, we’re working on ways to reduce the use of water and chlorofluorocarbons in the chip manufacturing process-there’s no reason for each company or country to figure out how to do that individually.
We are also cooperating internationally on the move to larger wafers. One of the reasons this industry has been so productive is that the silicon wafers from which the chips are made get bigger: today we use 8-inch wafers but the industry wants to move to 12 inches. So we’ve started a program in which the wafer suppliers and the international semiconductor companies, including Korean, Taiwanese, and European firms, are all working together to set the standards for the equipment needed to process these larger wafers. We think we can save several billion dollars in what will probably be a $10$15 billion conversion by having a single set of standards, and jointly measuring how the new equipment performs.
Setting Priorities for R&D
TR: What’s your vision for manufacturing?
SPENCER: I think the United States could make a significant gain if we could apply our strengths in managing complex technological systems to streamlining the manufacturing process. By inventing the production line, Henry Ford set the tone for manufacturing throughout the industrial world for 75 years. And after the Second World War the Japanese improved the system by recognizing that if you bring parts to the factory “just in time” and you give a good deal more authority to the people working on the production line, you’ll make higher-quality products at lower cost. So companies around the world adopted those practices.
But think about taking manufacturing to an entirely new level in which you integrate it with design and the customer. For example, if an engineer alters the specifications for a car fender, the factory immediately reprograms all the stamping tools on the production line and changes all the corresponding information available to every worker, shift manager, plant manager, and even customer. Doing that requires networking software-a field in which the United States leads: not a single network protocol or computer architecture in use anywhere in the world, except for maybe Nintendo, was developed anywhere else. And all programming languages are based on English-we own this business. Why can’t we apply it to manufacturing? And what better place to start than semiconductors, because you can measure every step in the chip-making process using techniques already integral to production.
We should be able to set up recipes so that if the next batch of wafers or circuits coming through the factory requires a slightly different manufacturing approach, you just type in b instead of a and the system downloads the right specifications into the furnace, etcher, or other tool. The shift manager could instantly find out which tools are being used and which machine has got to have preventive maintenance tomorrow. The plant manager would know the status of every job in the plant. The customer could find out that custom microprocessors are at “final metal” and are going to “test” tomorrow.
TR: Has Sematech been aiming in that direction?
SPENCER: Aiming, yes. Hitting, no. When I came to Sematech, I thought, no sweat: I know a lot about systems and software, and we’re going to apply that technology to semiconductor manufacturing. But although the consortium has spent a lot of money trying to do that, we have not yet made significant progress.
TR: Why do you think integrating design and manufacturing is so problematic?
SPENCER: It’s very difficult to go into, say, a Motorola facility that cost $1.5 billion and has to yield $2 billion of revenue each year and say, if you shut your plant down for a week, I’m going to put in this software system that may raise your productivity 20 percent. The people who run the place are going to say, Not in my factory you’re not. Not only is such an experiment risky but careers often depend on how well the factory does that quarter or year. What’s more, the bosses are often physicists or chemists or materials scientists who don’t know much about software systems except that the last time they tried to bring Windows 95 up on their home PC they had to call in some computer expert, and an entire factory is a thousand or maybe a million times as complicated as a home PC. Finally, computer scientists generally don’t talk to other people and other people don’t talk to computer scientists, so it’s difficult to develop the right software.
TR: How can we solve those problems?
SPENCER: We have to get industry leaders more comfortable with software systems, and we’ve got to make the software not only more reliable but also insertable in smaller pieces so operators can link various components rather than risking their entire manufacturing facility.
TR: How do we make sure that industry and government move in that direction?
SPENCER: I would like to see the engineering profession step up and play a major role in setting industrial goals-and indeed, in establishing R&D policy generally, the way physicists did at the end of World War II. The community could do that through the national academies. Those institutions aren’t perfect, but they represent the common ground among universities, industry, and government.
To promote any kind of R&D venture effectively, engineers also need to communicate better with the public, which we are not always willing to do. In my view the arrogance of the scientific and engineering community cost it the superconducting supercollider (which was canceled in 1993). Scientists came in and said we’re going to build this big machine and it’s going to cost a mere $2 billion. Every year that figure went up. And when it got to $8 billion the research community asked President Bush to approach Japan and say, guess what, we’ve got this big hole dug in Texas, the experiments are set, and we’re sorry your scientists didn’t participate in deciding what they will be, but we need $2 billion.
We know where we want to go in basic science, we know what we need to do in semiconductors, we know what we need to do in software, we know what we need to do in biotechnology; we have clear road maps. But if citizens don’t understand or believe that their lives will be enhanced, their kids smarter, and their air cleaner, then it’s incumbent on the technology community to convince people that what they are doing will indeed pay off.
TR: Shouldn’t social scientists, not to mention representatives of public-interest groups, also help set the goals for R&D?
SPENCER: I believe that getting social scientists to understand technology is harder than getting engineers to understand social issues, especially if engineers are broadly educated. But the notion of including non-engineers in setting R&D policy has great merit. I ran the national defense section of the Galvin Commission, which evaluated the future of the national labs in 1995, and at the last minute an individual from a public-interest organization was assigned to our group. He was a really interesting young man: he arrived with a chip on his shoulder-a feeling that we wouldn’t listen to him-but he also brought a different perspective. He definitely changed the committee, and he underwent a change in his outlook toward us as well.
TR: Aside from working through the national academies, how can engineers become more involved in influencing the direction of R&D?
SPENCER: We’ve got to find a way to get more and stronger technical people into government-I think our community owes the rest of the world that kind of contribution. If we don’t do everything we can to get more technologically sophisticated people into the administration and Congress, as well as onto congressional staffs, I think the nation’s going to have a real problem, given the profound technological component of so many major issues. I would start by trying to pinpoint the best people in the country and getting them involved, as we did here at Sematech.
Japan does this wonderfully. Some of the top graduates from the best schools in Japan feel it’s an honor to go into government. Yet very few U.S. engineers work on public policy in Washington after obtaining an advanced degree. The dilemma is that if I graduate from Berkeley with a PhD in electrical engineering, I’m tempted to take a job at my thesis adviser’s new company because five years later I may be a multimillionaire. And in this country that’s much more important than going to Washington to try to pass a significant piece of legislation, even though doing that would benefit a great many more people.
TR: Do the nation’s R&D priorities need to change?
SPENCER: When the Cold War ended, the United States was spending most of its research funds in two areas of paranoia: fear of the Soviets, and fear of death. We were putting $15 billion a year into the National Institutes of Health and about a third of a trillion dollars into defense. We don’t have to spend such sums on defense any longer; now we need to learn how to turn our fear of an enemy into a more positive drive to improve the quality of life.
In my younger years, I once proposed that if we’re going to spend $15 billion a year on getting ourselves healthy, we ought to spend at least that much on having something to eat. Rather than hoping for civilian fallout from military spending, we’ve now got an opportunity to focus directly on economic strength, which today is as important as military strength. The National Institutes of Health provide a strong model for creating a science and technology institute that would focus on our national economic health. Alternatively, if you cut the nuclear weapons program and defense R&D by a factor of two, you could add that money to the National Science Foundation.
TR: Congress now often seems hostile to funding anything that’s not basic research.
SPENCER: The people I’ve met in Washington are smart, hardworking, and dedicated. The problem is that we as a technical profession have not been able to get our story across as well as we should. We simply have to work harder at establishing a better relationship with members of the government at all levels.
The most important task is to create a mechanism for learning from past mistakes and successes and for applying the lessons to the next effort. That way we’re always moving in the right direction, and Congress and the pub-lic can be confident that R&D funds are being well invested.
DeepMind’s cofounder: Generative AI is just a phase. What’s next is interactive AI.
“This is a profound moment in the history of technology,” says Mustafa Suleyman.
What to know about this autumn’s covid vaccines
New variants will pose a challenge, but early signs suggest the shots will still boost antibody responses.
Human-plus-AI solutions mitigate security threats
With the right human oversight, emerging technologies like artificial intelligence can help keep business and customer data secure
Next slide, please: A brief history of the corporate presentation
From million-dollar slide shows to Steve Jobs’s introduction of the iPhone, a bit of show business never hurt plain old business.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.