Since 1970, the semiconductor industry has experienced seven cycles (’71-73 – 3 years), (’78-’80 – 3 years), (’83-’84 – 3 years), (’86-’88 – 3 years), (‘93-’95 – 3 years), (’99-’00 – 3 years), (’02-’04 – 3 years). All cycles ended due to overcapacity and in all but two cycles (’83-’84) and (’02-’04), a worldwide recession contributed.
The semiconductor equipment industry cycles have mirrored the semiconductor cycle until the last cycle of (’02-’04), with peaks and valleys falling within a few months of each other. But the last cycle was different. While semiconductor sales continued on an upward slope in the last half of 2002, equipment sales "tanked" -- a real divergence that had not happened in the prior years. A low level of equipment billings continued to increase for eight months until July ’02 and then dropped for the next 12 months until an uptick in Aug.‘03. The current equipment cycle ended in June ’04 and lasted only 11 months! Meanwhile semiconductor sales growth has continued practically unabated from Mar. ’02, but that too is about to end.
We warned in our Press Release of August 3 that the pendulum in equipment purchased in late 2003 and 2004 may have swung too far; that semiconductor manufacturers didn’t need a 50+% increase in equipment purchases for only a 27% growth in semiconductor sales. We stated at that time that as the pace of semiconductor sales slow, we would see equipment pushouts and eventual cancellations, as semiconductor manufacturers reined in expenses and recognized that they overshot their needs.
It is apparent and obvious that excess semiconductor inventory were brought about by excessive equipment purchases in ‘00 and ’04. But many things happened in the in-between years of the two cycles that will have ramifications in future cycles.
Semiconductor manufacturers got smarter in 2002 and began a series of technological advances that would increase revenues without the need to increase equipment purchases – 300mm wafer production would increase the number of chips per wafer by a factor of 2.4; aggressive die shrinks would further increase the number of chips per wafer; the transition to copper by logic and some flash memory makers would increase speed and performance of devices 20% without the need to go to smaller linewidths; and extending the lifetime of equipment, particularly 248nm lithography tools, would reduce needs for new state-of-the-art tools.
So why did the semiconductor industry apparently get dumber all of a sudden and buy all that unnecessary equipment in 2004? Sadly, it has to do with forecasts and the fear, uncertainty, and doubt (FUD factor) that prevail in the industry.
The ramp in semiconductor production and equipment purchases in 2000 that resulted in nearly $10 billion in excess chip capacity in 2001 was a result of erroneous forecasts of huge growths in electronic system end markets that never materialized, which led end-users to double-up on chip purchases. Forecasts of increasingly high demand and prices for DRAMs and flash memory also forced end-users to buy more chips earlier than they needed to avoid price hikes and allocations.
The ramp in equipment purchases in 2004 was for a different reason and a result of forecasts of hyper growth in the equipment market that were in excess of 50% in 2004. Surprisingly these forecasts were correct. While the equipment market has grown 68% in the past three quarters, we project a 22% quarterly sequential drop to 53% growth for all of ’04. But it is now apparent that the semiconductor manufacturers did not need 50+% growth in equipment in 2004 for only 26% growth in chip sales. Excess chip inventory in now approaching $2 billion, and even foundry giants such as TSMC and UMC are seeing revenues drop for the past two months. Warnings and layoffs abound in the industry.
So the burning question is why did the semiconductor manufacturers purchase all that needless equipment in the first place? First of all they listened to the forecasts, not only by market researchers, but by SEMI itself. (Obviously they didn’t heed our advice and forecasts of only 32% growth needed). Fear set in of the likelihood of allocation of equipment in 2004 for meeting even modest semiconductor growth of only 17% (WSTS initial forecasts for 2004 made in late 2003). Executives at the semiconductor houses got to be where they are because they are smart enough to know when a forecast is hogwash. But they were forced to succumb to the FUD factor.
And therein lies the problem with the industry itself. Consolidation has limited the number of suppliers of equipment and limited supply means long delivery cycles. Business plan depends upon a company’s ability to obtain equipment in a timely manner and at acceptable prices. During times of significant demand leadtimes for delivery can be as long as one year, and shortages of equipment could result in an increase in equipment prices as well as chip production delays. If semiconductor manufacturers are not able to obtain equipment in a timely manner and at a reasonable cost, they may not be able to meet customers’ orders or achieve expansion plans. These could negatively impact competitiveness, financial conditions and results of operations.
As large equipment companies such as Applied Materials, Novellus, TEL, Lam Research, and KLA-Tencor gobble up small competitors, semiconductor manufacturers have a choice of placing large, knowingly unneeded equipment from the “big boys” or buying from smaller companies. In fact, smaller companies have taken market share in recent years as they have focused on advanced technology and exemplary service to their customers. But these companies often have limited manufacturing capabilities, and some are small enough that semiconductor manufacturers question their longevity in meeting future sales demands and service.
These problems will remain with the industry, and unless a standards body or organization addresses them, not stock market prices and stock options, equipment cycles will fly high and remain short lived.
For more information on semiconductors, click here.