In The News
1998, IC Insights, Inc., Fisher-Holstein, Inc.
The Semiconductor Market Upturn is Close at Hand
Undoubtedly 1998 will be known as the year of the great semiconductor bust and it will be compared to the downturn of 1985 when sales fell 17%. During downturns, industry pundits find it hard to see an end to the bust, and memories of years of torrid growth (29% in 1993, 32% in 1994, 42% in 1995) dim as companies close fabs, lay off people, and ration paper clips. However, semiconductor industry growth has averaged 17% per year for over 30 years --- through two oil crises, war, recession, the collapse of the Eastern bloc, the opening of China, and other convulsions.
At some point, industry growth must slow to more closely match the rate of growth of electronic equipment sales --- which averaged 10% per year during the past 30 years. But in the meantime, there is good reason to believe that growth will return soon. Once again it will catch the naysayers, who do not believe in this industry's tendency to let history repeat itself, by surprise. The purpose of this article is to: identify and document what we believe to be firm evidence that an industry upturn has begun and look at how this upturn is going to play out over the next three years.
Underlying our forecast of an upturn in the industry are the following assumptions for the 1998-2002 timeframe:
On the basis of these assumptions, each one of which can significantly impact the forecast, we are preparing to show evidence that the long anticipated industry upturn is around the corner.
Is The 1996-98 Bust Worst Ever?
There are as many theories as to the causes of the semiconductor bust of 1998 as there are pundits. Reasons cited focus on the great DRAM glut, the Asian flu, and saturation of the PC market. While these particular factors may account for the unusual double dip nature of the bust --- which has not happened in the past decade --- a bust is entirely characteristic of the industry. One reason for the apparently drastic nature of this downturn --- compared to other downturns, is the sky high US dollar along with the change in the value of currencies of key semiconductor producing countries. Looking at the Japanese yen as an example:
The yen, between the peak of the market in 1995 and 1998, devalued considerably. Because the industry still states its global revenues in US dollar terms, its appreciation accentuates any shrinkage in the size and growth rate of non-US manufacturers, obviously more so of those in countries with depreciating currencies. The steep plunge of the won from about the US$1=780 won range to US$1=1,550 won range within one year had a similar effect of shrinking the apparent dollar size of their output, holding other things constant. If we were to restate the industry's statistics in a basket of currencies that are more stable, such as ECUs, it would show an industry performance considerably better than the present US dollar based numbers indicate and would provide a more optimistic picture of the industry.
Other Exchange Rate Impacts
Another little-noticed effect is created when the currencies of key semiconductor producing countries fall quickly; it gives the industry based there a short term boost in profitability if the currency exposure is not covered. In particular, Korean companies which paid for equipment in Won just prior to its collapse found that they could, over the life of that equipment, realize a short term one time gain from the devaluation. Their products are primarily produced for export and sold for Yen or dollars that are now worth more Won. This allowed them to cut prices further in dollar terms and still realize a larger paper profit or a smaller loss, and generate cash flow in dollars or Yen for their hard pressed conglomerates.
While the decline of the Won makes it very difficult for the Koreans to equip and start offshore fabs, they benefit from devaluation in the short run. In the medium term, their competitors will reap the benefits of the Koreans' inability to bring their offshore fabs on stream at the right strategic time, and in the longer term by their need to invest more expensive dollars or Yen in future upgrades or expansions --- especially costly overseas expansions. In any case, the Koreans will be less able to subsidize their DRAM operations in the future, even through the traditional method, frowned upon by the IMF, of cross subsidies from other members of the chaebol. This suggests that a major source of endemic DRAM fab overcapacity will probably work itself out over the next year or so, removing a big impediment to healthy industry growth.
Taking a broader look at the fundamentals of the industry by focusing on the electronic equipment market, we see an anemic 3% growth rate in 1998 projected to return to around 9%, close to its historic growth rate, for the following four years (1999-2002). Electronics production is the driver for semiconductor production, and a return to a normal growth rate is a positive indicator of industry health.
Another view of the industry can be obtained from capital investment patterns. Semiconductor capital investment is traditionally cyclical, with a repeated pattern of over and under expansion. The industry has drastically shifted capital investment from levels as low as 16.1% of sales in 1987 to a peak of 32.7% of sales in 1996. It is instructive to note that the 1996 rate exceeded even that of 1984 (27.4% of sales) which led directly to the great bust in 1987 capital spending (16.1% of sales). Thus, the present situation is entirely consistent with historic industry patterns.
IC Insights believes that a steady 21-22% of sales is a reasonable, sustainable rate of capital investment for long-term industry growth. However, the industry has never heeded such norms and has paid the price in drastic boom and bust cycles. Equipment makers, who are most directly impacted by such cycles, took the brunt of these wild swings in demand, limiting their ability to support the semiconductor industry in an orderly fashion. The drastic cut in capital spending during the past year or so is typical of the industry's overreaction to excess capacity. Under-investment during the present bust years will, as before, set up the semiconductor and equipment industry for the strong upturn by the year 2000.
Additional Evidence of an Upturn
During the 2nd quarter of 1998, IC average selling prices (ASPs) fell to the lowest level since the peak of the market at the end of 1995. Unit shipments also reversed their decline as of June, 1998, and have been rising since. DRAM prices, a critical indicator, have been flat for the past two weeks --- a major shift from the non-stop declines seen during the past two years. Similarly, prices for microprocessors (MPUs) have risen recently. PC manufacturers who invested considerable effort in slimming down their inventories and switched to build-to-order or build-just-in-time (such as Dell, IBM and Compaq) have begun to bring these facilities on stream to meet the anticipated demand for PCs in the Fall of 1998. The tightly coupled nature of this new supply chain means that upticks in PC demand no longer have to work their way through a long inventory tail, but will be apparent much faster.
As the chart below shows, cycles in the industry are relatively consistent and capital spending tracks IC ASPs closely. Once ASPs begin to show a sustained rise, capital spending increases follow quickly.
What About the Upturn?
The upturn, as with the downturn, will as it did before catch people by surprise. Historically, IC manufacturers traditionally either over or under spend on expansion, leading to wild boom or bust cycles. The extent of this boom and bust is seen in the following chart:
For any given year, the industry rarely comes within 7% of its average compounded 20-year historical growth rate of 17% and is unlikely to do so again during this upturn. Another way of looking at this is to examine the year-to-year change in growth rates going into and coming out of a downturn:
The industry's ability to turn from feast to famine in a little less than a year is a characteristic that has been demonstrated over and over again. Accordingly, a growth rate change of from -8% in 1998 to +10% in 1999 is entirely consistent with industry history. Furthermore, the evidence indicates that a great semiconductor boom is shaping up for 2000 and 2001, with both years likely to register growth rates in excess of 24% per year as per the latest IC Insights forecast.
The industry's coming upturn is likely to be entirely in character with previous booms. The upturn will, however, be unusual in that it will likely be strong and driven by several factors, the impacts of which are at present still difficult to predict precisely. The above forecast by IC Insights did not factor in the issues below, any one of which can significantly impact the strength of the upturn:
We now know that the Y2K problem has had a depressing effect on corporate IT budgets and in diverting, to testing, fixing code, and installing new Y2K compliant systems, resources that would have gone toward PCs and other IT projects. However, the upside impact of the Y2K issue has not been sufficiently researched. On the basis of a pilot study, we identified numerous examples of electronic equipment that will have to be replaced prior to (or shortly after) the onset of the Y2K bug. Much of this equipment is in facilities like electric power plants, factory control and automation equipment, and other places that provide essential services. The equipment, in many cases, contains embedded processors with non-Y2K-compliant code.
While we cannot accurately forecast demand that will likely come from Y2K issues, it is likely that there will be many "boxes" that are presently working day and night in numerous locations that will fail. When such equipment fails, rather than trying to correct the code, the most expedient solution may be to simply order a new box. The upside impact of this on demand in 2000 and 2001 will be an issue to watch for as the Y2K drama unfolds in the next year.
During the next two years, digital wireless technology will, for the first time, become a major demand driver and open up new markets. The widespread availability of digital telephones in North America will likely cause an explosion of demand for wireless phones and data services. Similarly, the successful deployment of systems like Iridium and wireless data services will expand the market for electronic devices considerably beyond present dimensions. Again, while the boom from these new services is difficult to forecast, it is likely that the industry will once again experience an explosion of growth.
Finally, it would be wise not to count out the traditional markets for semiconductors, like PCs, autos, consumer electronics, etc. All of these will likely be back on their feet once the current industry slump is over. Indeed, a single killer PC application can dramatically turn around demand for both microprocessors and memory. Other one time events like the GM strike, the distribution snafus caused by Hong Kong's new airport, and China's crackdown on grey market PC imports, are all passing problems for the industry that are either resolved or will work itself out over the next few months. These factors will scarcely retard increased orders in the second half of 1998. All of these factors point to a sustained industry upturn in the coming months.
The industry upturn will, if it goes as before, generate a great boom that will rapidly soak up the excess capacity that now dogs the market. Pundits who believe that there are as many as four or five excess DRAM fabs that should now be closed may, within a year, be surprised to find insufficient capacity to meet demand. As demand materializes, there will be another rush to add capacity. In due course, demand will outstrip supply, and capacity shortages are likely to appear by the year 2001, sparking another rush to build. However, as during previous downturns, the manufacturers who stay the course during the present downturn will be best able to take advantage of the upturn. Companies like Intel and IBM will stand out for having continued to invest during the downturn and will reap the benefits of existing capacity on line to meet customer demands, when some competitors will be unable to do so. History will repeat itself again.
This article is compiled in part from a detailed 25 page 1998-2002 IC Industry Market Forecast Report available from IC Insights for $295. Contact to obtain a copy.
Additional data and information on industry capacity and trends was compiled by Fisher-Holstein, Inc. (FHI), a research firm that specializes in lifecycle cost-of-ownership analysis and capacity planning of semiconductor wafer fabrication facilities. Contact , for additional information of FHI products and services.
About The Authors
Bill McClean is president of IC Insights, Inc., an IC market research company that began operations in 1997. Prior to forming IC Insights, he worked at ICE Corporation for 17 years --- the last 10 as vice president of market research. Mr. McClean serves as contributor and managing editor of The McClean Report, the company's annual in-depth survey and forecast of the IC industry, and Strategic Reviews, its IC company information database. Mr. McClean received his Bachelor of Science degree in Marketing and an Associate degree in Aviation from the University of Illinois. He can be reached at .
Danny Lam is a director of Fisher-Holstein, Inc., a firm that specializes in lifecycle cost-of-ownership (LCCO) analysis and capacity planning for semiconductor wafer fabs and the co-developer of the FHI LCCO decision support system (DSS) that helps wafer fabs, understand and improve their lifecycle profitability. His recent publications have appeared in Semiconductor International (July, 1998), (www.semiconductor.net/semiconductor/archive/Jul98/docs/feature3.html), Site Selection Magazine (August, 1998), Infrastructure (Feb, March, April, 1998), and Semiconductor Fabtech. (www.fabtech.org) Dr. Lam is a fellow of the Economic Development Institute, Auburn University and a past fellow of the Center for Science and International Affairs, Harvard University. He received his Bachelor degree from University of Waterloo, an MBA from University of Western Ontario, and his PhD from Carleton University. He can be reached at: or through Fisher-Holstein, Inc.
John Kanz, CEO of Fisher-Holstein, Inc., has over thirty-five years of microelectronics and high-tech industry experience including hands-on R&D, specialized facility projects, corporate management, and entrepreneurial startups. With over fifty technical and academic papers, he also co-authored a key chapter in the 1996 McGraw-Hill Handbook of Technology Management and chaired US and international microelectronics standards groups. He holds a BS in Physics from the University of Washington, an MS in Physics from the University of Illinois, and an MBA and PhD in Management from the Peter Drucker Graduate Management Center at Claremont. He can be reached at: ¸ .
Copyright © 1998 IC Insights, Inc.