Posted on 09/28/2003 10:34:52 AM PDT by nwrep
|
SAN JOSE, Calif. — A panel session at the Custom Integrated Circuits Conference here Tuesday (Sept. 23) debated the implications for U.S. design engineers of IC design outsourcing.
Panelists offered free-market platitudes, candid warnings, reassuring economic generalizations and some incisive observations that may help individual designers find a foothold on what promises to be a slippery slope for the profession in the coming years.
Rakesh Kumar, president of operations-outsourcing venture TCX Inc. (Poway, Calif.) said the U.S. has a long history of outsourcing its key industries, including the steel and automotive businesses. He noted that the electronics industry has followed suit, outsourcing the vast majority of packaging, assembly and test.
"Every industry has done this," Kumar said. "The only difference is at what point the outsourcing curve turns over — at 20 percent, 50 percent or 100 percent of the U.S. capacity." The question is not whether the U.S. will outsource chip design, but whether it will retain some design capability or send out everything, Kumar said.
Kumar posed three questions for the panel: Is chip design outsourcing inevitable? What role will it leave for the US? and What can an individual engineer do about it?
Ann Lee Saxenian, professor of political science at the University of California Berkeley and an author on the subject of economic globalization, documented a substantial flight of the electronics industry from the U.S. She said the U.S. share of the global semiconductor market would drop to 30 percent by 2010, while the Asia-Pacific share would rise to 35 percent and Japan's declined to 20 percent.
In contrast to the overall figures, she noted, 40 percent of all fabless semiconductor revenue flowed into Silicon Valley companies in 2002.
Saxenian described the redistribution of the industry not as outsourcing but as a new global division of labor. In this new order, a once monolithic industry is disaggregating, with individual tasks migrating to locations that can perform them most productively.
Seen in this light, she said, the U.S. would likely retain dominance in IC architectural design, in investment into the semiconductor industry and in design of chip manufacturing equipment and EDA tools.
Ed Ross, president of TSMC USA, charted the competitive landscape in the Chinese, Taiwan and U.S. semiconductor markets. Contrary to most U.S. executives, Ross said Taiwan and China have collectively become a hotbed of design activity. "There are about 350 design houses in Taiwan today, and 500 in mainland China," he said. "Many of these are small, but not all of them. And some are very sophisticated."
Right mix
China in particular had the right mix of advantages to prosper rapidly, he added. "The industry receives heavy government investment in China," Ross said, "and benefits from a very strong local market. But on the minus side, China currently suffers a critical lack of experienced managers, and their continued lack of effective legal protection for intellectual property could become a serious limitation."
On balance, Ross said, China would mature as a design community more rapidly than Taiwan "for one simple reason. They are importing a lot of managers from Taiwan who have already been through the experience." If there is a dark cloud looming over the Chinese industry, Ross said, it is that the huge fab building campaign could lead to global overcapacity by 2005 or 2006.
Werner Goertz, vice president at outsourcing megastore Wipro Technologies (Bangalore, India,) added a different perspective. On a macroeconomic scale, Goertz said design outsourcing was a nonissue. He showed data indicating that the total number of jobs projected to leave the U.S. from outsourcing by 2005 — about 3 million — would be only slightly larger than the number of jobs lost through normal operations in the U.S. in the boom year of 1997.
Further, Goertz said productivity increases from outsourcing enriched U.S.-based companies that outsourced design work. Hence, in a trickle-down view of engineering economics, the job loss benefited the U.S. engineers.
Geortz counseled engineers to effectively run for high ground, or move their careers away from basic design and into architectural design or design management — tasks least likely to be outsourced.
Behrooz Abdi, a vice president heavily involved in mixed-signal design at Motorola, said outsourcing is not just about lower salaries. He said the underlying problems were that productivity growth had outpaced demand, and that companies had lost their differentiation. This has forced faster time to market and lower development costs as a substitute for successful new products.
Abdi agreed that the best path for individual engineers and for companies was to innovate at the systems level rather than trying to differentiate themselves on the basis of chip or circuit design.
Gloomier view
Brian Fitzgerald, chief executive of the small design services house ChipWrights (Boston) was pessimistic.
"I think design outsourcing is necessary to a small company in order to compete," Fitzgerald said, "but in the long term I think it is bad for the country." Fitzgerald said he is continually approached by offshore design shops offering to work "three to five times cheaper than we can do it here.'' The more outsourcing, the more global competition is lowering engineering salaries and career opportunities in the U.S.
"One of my engineers comes to me and says he has to have a 10 percent raise. I know he's good, but I also know I could get maybe five times more work done for the same money I'm paying him now. So what am I going to do?" Fitzgerlad asked. ''The more we cut away at the incentives for people in the U.S. to take up engineering careers, the more we undermine out ability to innovate."
The panel's consensus was that IC design outsourcing is inevitable, and probably irreversible. The U.S. will be left with product specification for the domestic market, architectural design and investment from venture capital firms. All individual engineers can do in the face of a flood of outsourced design work is to flee to the relative safety of system architecture, or target highly individual analog or RF design talents.
They can also cling to the hope that the water stops rising, panlists said.
Okay, what do you mean? I don't get paper or plastic.
He is suggesting that your son will become a grocery clerk if he studies electronics. I don't agree. There are a lot of things that could happen in the world and if we don't have people who understand modern technology we are in big trouble.
Intel, for one. Motorolla for another. Basically any designer of advanced semiconductors for the embedded market. Intel would love its StrongARM technology to be in every refridgerator, and Motorolla wants its PowerPC technology in every set top box. Like I said, DirectTV now has a TiVO like recorder in its settop box. That requires a CPU that has the power of a desktop PC.
Experience matters tremendously in the semiconductor industry. Nobody out there can build a Pentium 5 CPU unless they already have years of experience in the CPU industry. So if a TV set top box or a CD player with built-in MP3 jukebox requires a CPU the level of a PC CPU, only semiconductor companies with the experience in computer CPUs need apply.
And even the manufacturing of advanced semiconductors will remain in the US, and perhaps to some extent to Japan and Europe. I know one company that tried to take a small, 10-year old CPU design and get it fabricated by UMC in Taiwan, and UMC could not do it. It was too complex. If Taiwan cannot do it there is no way China could do it.
I was at a conference once, and somebody mentioned the number of CPU designers there were in the world. It is not a big number. I think it is about 6,000, and half of the 6,000 were employed by Intel. The number of CPU architects is much fewer (I seem to remember a number of 300, but I could be wrong on that one).
While the universities can instruct on very simple semiconductor designs, the real learning happens in the semiconductor companies, under the tutalage of the CPU architects. So the 6,000 number is no doubt growing, but slowly.
The problem is CPU design is a building block process. You cannot simply start from scratch, hire some electrical engineers, and have them design some complex circuits. Instead, the approach is more like going to the moon. First put a satellite into orbit. Then a monkey. Then a man. Then send a robot lander to the moon, and so on.
It takes about the same amount of time as getting to the moon as well. For example, new CPUs can take from three to five or even more years to go from idea to volume production. Some ideas are out there for years before the semiconductor technology catches up and allows the idea to be implemented. Semiconductor companies today are working on ideas that will not see the light of day until the end of the decade.
Modern CPUs projects today are on the engineering scale of the Manhattan Project or Project Apollo. I am not kidding. They are pushing the limits of physics every day. So you can't just bring somebody in as needed. I guess you could if bring in a team of 500 people for five years.
Because Americans bring in an electical engineer from a foriegn country and tell him to work on this one circuit out of 300 on the chip. And the person who tells him to do that is the engineering team leader in charge of that section of the chip, who knows how to manage a team because he has done it for the last 10 years. And that team leader is managed by the design team leader, who has done this for his entire working life. And beside him are a handful of visionaries with all of the radical ideas on how to make great things. And the design team leader decides what is practical, based on his 20 years of experience.
It is easy to hire in somebody at the bottom. It is hard to grow the experience needed to manage the complexity of the project. I know sombody that spent over 20 years at IBM working on CPU designs and leading PowerPC processor design teams. Because it takes 3-5 years to design a processor, and it takes 3-5 iterations to develop the leadership experience, it takes 10 to 15 years to develop a design team leader. IBM sent this guy to engineering managment school between each project. He took on increasing responsibility within the design teams over time.
What you suggest would be like hiring in a General, then conscripting an army, and expecting it to win a war. Such an effort would not win the war because an army needs career NCOs with 15 years of experience to make an army work.
If you could just bring somebody in, somebody would have done it. The fact that somebody has not done it is the best indication of how hard it is. It could be done, but it is a 20 year project.
Or shoe fitters at Foot Locker.
But I guess it's RACIST to suggest this...
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.