The inevitability of business cycles
12/01/1996
The inevitability of business cycles
Ed Korczynski, West Coast Editor
Gerhard Parker was one of the main speakers at this year`s SEMI awards dinner, held September 25 at the Westin Hotel in Santa Clara, CA. As an executive VP of Intel Corp., and VP and GM of the company`s Technology and Manufacturing Group, Parker oversees all capital spending at Intel, and thus has the largest checkbook in the Semiconductor Industry Association.
During a predinner press conference, Parker was asked, "What can be done to minimize the extremes of the equipment business cycles?" Parker`s surprising reply was that he expects extreme cycles to be with us for the foreseeable future, and that industry trends show they will only increase. "There`s very stable growth here that we are all guilty of putting into oscillation," he said.
"I`m less than optimistic that industry maturity is going to make anything more stable," he continued. "If anything, equipment companies have been doing more outsourcing, so there are more layers of people forecasting to each other. By doing all this outsourcing and subcontracting, hopefully to hedge themselves, they just make it worse for somebody else. So in a sense, I`m not optimistic that this thing is ever going to be stable."
The worldwide PC and electronics industries continue to grow at robust rates. All support and supplier industry layers must, of course, continue to grow to meet increasing demand. However, a downturn of a percent or two ripples through the industry and causes havoc when people lose sight of the total macroeconomic picture.
Chipmakers occupy one horizontal layer within a large vertical supply pyramid (Fig.1). More layers have been added as the industry has matured. The fierce competition within each horizontal layer, the sheer number of layers, and fallible forecasting all contribute to increasingly large oscillations in the bottom third of the supply pyramid.
The sheer number of layers increases the magnitude of the fluctuations. If there`s a slight tick at the top, and if every layer increases its forecast by 10 percent, the bottom of the supply pyramid can end up with a forecast that is literally twice as high as reality. If each layer in Fig.1 merely multiplies the forecast from above by 1.1 (in an upturn) or by 0.9 (in a downturn), then the bottom layer forecast oscillates between plus 136 percent and minus 61 percent. Interlayer forecasting is a main reason for the equipment industry`s wild mood swings.
 |
Figure 1. The effect of each layer in the supply pyramid changing its forecast by just 10 percent up or down based on the next higher layer. Outsourcing has clearly exacerbated the problem.
Inevitability of cycles
The horizontal layers all depend on each other vertically and forecast to each other. It can be shown that the system naturally tends to go into oscillations. Wild swings are primarily a result of people trying to outguess each other as they grapple for increased market share within a horizontal layer. When a company tries to increase market share, it forecasts an increased demand to the next lower layer, and that next layer starts adding capacity without determining whether the end-product demand is really there.
Parker explained, "Where you typically get into trouble, quite frankly, is where people drop market share or where there`s a technology transition. You don`t recover market share on the last generation technology; you look to who`s got the next capability because that`s how you recover. So as you go through technology transitions and there are people forecasting based on market share, it really goes out of control very quickly."
For example, while DRAM bit growth/year is relatively steady at 50-80 percent, unit shipments/year fluctuate due to product generations, and a slight under- or over-capacity results in wide price fluctuations. Market shares may shift during generation changes, and companies within a given horizontal layer tend to be overly optimistic in their forecasts during these periods. However, a company that provides truly enabling new technology will often find that its business explodes and it has underestimated its capacity requirements.
 |
Figure 2. Book-to-bill figures for IC and front-end equipment from 1990-1996.
Parker confessed, "By the way, we see this inside of a company too. If our wafer fab is producing wafers at a certain rate, it sends them into the wafer-sort area for test, and those guys don`t want to be caught short, and they know it`s not linear, so they go for a 10 percent buffer. And then, of course, the assembly guys see that they have only so much capacity, so they better have a 10 percent buffer. We`ve gotten ourselves into situations where we`d wind up with 30-40 percent more capacity at the tail end than we really need. Just because everybody is trying to outguess each other with forecasting."
It is also arguable that the consistent 20 percent range compound annual growth rate (CAGR) of the semiconductor industry, and the larger growth of the semiconductor equipment industry inherently induce oscillations in capital equipment purchases (see Fig. 2 and "Whither the equipment market? Views differ, but growth will return," Solid State Technology, October 1996, p. 50). If the industry ever matures to a point where its growth rate is 10 percent instead of 20 percent, then there would probably be less volatility. That day may come sometime in the next decade, since the semiconductor content in electronics can only increase to a certain maximum, at which point the growth rate of the IC industry will have to slow down to the approximate 10 percent of the electronics industry.
Elizabeth Schumann, SEMI senior market analyst, said, "It`s compounded by the fact that all capital goods industries over history have always been cyclic in nature and probably always will be. This is because capital is always added in chunks, not in steady progressions of capacity."
An interesting countercycle runs through the middle of this downturn for some subcontractors. Companies that provide consulting engineering services have actually seen their business increase in the last 12 months. Mega Professionals International (MPI), a software consulting company in Milpitas CA, has seen its semiconductor equipment business grow over 100 percent. Manish Mehta, marketing manager at MPI, explained, "Equipment companies have had a hiring freeze, so they count on subcontractors to get the work done."
Machine shops - an example
Established semiconductor equipment manufacturers such as Varian and Watkins-Johnson used to have major machine shops in-house, but have generally shut them down over the last ten years. Newer companies chose to outsource their machining from the beginning, and thus off-load the burden of capital and human resources investments.
Of all the subcomponent manufacturers, independent machine shops in Northern California have been hit particularly hard in the downturn. There are an estimated 4000 machine shops of all sizes in the San Francisco Bay area. Relatively large (over 100 employees), well-capitalized companies are struggling, while countless small companies have been devastated, according to Richard Wills, president of D & H Manufacturing Co. and chairman-elect of the National Tool Machining Association (representing 2700 shops within the $25 billion industry). Two developments within the last five years are directly responsible for this current calamity: increased minimum tooling investment levels, and increased reliance upon single customers.
Fully automatic, highly efficient, $2-3 million "linear" machining systems have become available within the last few years. Wills states that more linear systems have been installed in Northern California than in the rest of the United States. This enormous manufacturing capacity also serves the military, aerospace, and medical industries, but it is clear that much of the investment was driven by semiconductor equipment forecasts. Now many of these tools sit idle, while their owners struggle with the payments.
 |
Figure 3. The $2000 average selling price of a PC is based on the PC maker, integration, storage, chips/motherboard, memory, and display.
Recently, equipment manufacturers aggressively pursued exclusive "partnerships" with single large machine shops, each of which farmed out work to 10-20 small (under 20 person) shops. Many partner shops had over 90 percent of their business dedicated to just one equipment manufacturing customer. Silicon Valley press reports and trade association officials indicate that while the partner shops are definitely suffering right now, it is the smaller second-tier shops that have had their business literally turn off overnight.
Wills stated that subcontractors cannot afford to commit to the same style business relations when orders begin to pick up again. "We have to have a position that is smart for both the customer and for us, and that is that we not be locked in with a percentage of their work that could fold our company."
To grow the pie larger
The $2000 average selling price of a PC is based on established percentages from each of its major components (Fig. 3). So to increase sales for any component, companies can`t take a larger piece of the pie - they have to grow the whole pie. PC sales still drive some 60 percent of semiconductor manufacturing, according to Intel, so companies have to add value to convince people to buy more PCs to grow the whole market.
Some growth will come from emerging markets, but most will require adding sufficient value to convince end-users to replace old PCs. That`s true for the equipment industry too, as new equipment generations that are more productive allow higher layers in the supply pyramid to continue to make PCs more attractive. Merely increasing the cost of the product sold can`t double sales; what doubles the market size is twice the value in a PC and twice as many users.
Parker`s final advice for dealing with cyclic business trends: "Don`t panic and overreact in a downturn, and the next time you see a book-to-bill of 1.5 don`t get too excited either. The market is strong and the market is growing, and you`re going to see perturbations. That`s human nature, because we`re all going to try to outguess each other just a little bit in forecasting."