What the Analysts Are Saying
What's the outlook for the semiconductor industry? Channel recently asked three noted industry analysts for their opinions about what will be happening over the next year or two. Their forecasts follow.
Outlook 1999: Past Bottom, But Slower Recovery This Time
Clark J. Fuhs
A year from now, the industry will view 1999 as a recovery year, but on paper it will probably not look very good. On a quarterly run rate basis, the wafer fab equipment market hit a low point in the third quarter of 1998, but the recovery will not be very steep, leaving calendar 1999 with a tough comparison, essentially flat. However, fourth quarter 1999 will see run rates for the industry 40 percentt or so higher than the depressed rates of the fourth quarter 1998. Year-end 1999 run rates will still be significantly below peak levels experienced in the second quarter of 1996 and fourth quarter of 1997. We do not expect a record quarter for the industry to occur until early 2001.
For the past two years, we have been calling for a "W" recovery pattern, with the second-phase downturn being caused by the fundamentals of overcapacity and financial health eventually winning over the desire for technology. The financial crisis that caused capital restraint was worse than we projected, as semiconductor demand was negatively affected in 1998 by slowing or depressed economies.
DRAM Overcapacity Continues Near-Term, But There is Light
There is a bright spot, as we see the first fundamental signs aligning to create a spending boom during late 2000 through early 2002, with a shortage in the DRAM market emerging during 2000. Presently, we are estimating the overcapacity in DRAMs to be between 15 and 20 percent. Our silicon area demand model shows that the movement to the more silicon-efficient 64 megabit density will sustain the oversupply throughout 1999. However, net supply will be trending down as well over the next year as capital spending cuts make their mark on reducing the rate of supply increase. The situation should set up nicely for a shortage of capacity in the year 2000.
Foundry: Oversupply Became "Acute" in 1998, Spending Down in 1999
Dataquest's analysis of supply and demand in the foundry industry has shown that during 1998 an oversupply condition would emerge in the foundry industry. The slowdown in semiconductor demand actually drove the oversupply to the acute level of 30-35 percent by mid-year. The foundry suppliers reacted quickly, driven by profitability in their business model. Many suppliers cut back spending rates by 50 percent or more from the first half to the second half. Our silicon supply/demand model shows that the industry will be in oversupply through at least mid-2000, so we expect the foundry suppliers to remain at current spending levels or lower throughout 1999, representing perhaps a 20 to 30 percent cut from 1998 levels.
Capital Spending Forecast, 1997 to 2003 (Millions of Dollars)
1997 1998 1999 2000 2001 2002 2003 1997-2003
CAGR (%)
Total Capital Spending Growth (%) 40,505
-9.9 30,259
-25.3 29,159
-3.6 37,143
27.4 59,378
59.9 72,898
22.8 70,702
-3.0 9.7
Percent of Semiconductors (excluding 300mm pilot) 27.2 21.3 18.0 17.9 23.6 28.1 24.9
Americas Growth (%) 14,178
0.5 11,096
-21.7 11,594
4.5 13,846
19.4 19,411
40.2 23,670
21.9 25,920
9.5 10.6
Japan Growth (%) 7,986
-17.3 5,765
-27.8 4,754
-17.5 6,248
31.4 10,14
62.3 11,498
13.4 10,192
-11.4 4.1
Europe Growth (%) 4,089
-18.8 3,978
-2.7 3,909
-1.7 4,807
23.0 7,352
52.9 9,122
24.1 8,391
-8.0 12.7
Asia/Pacific Growth (%) 14,253
-11.7 9,420
-33.9 8,902
-5.5 12,242
37.5 22,472
83.6 28,608
27.3 26,199
-8.4 10.7
Source: Dataquest (January 1999)
Table 1
The Wafer Fab Equipment Forecast
The movement to a "minimum investment" pattern in 1998-1999 has meant a severe cut in spending levels in 1998 compared to 1997 (see Tables 1 and 2). The stall in demand for semiconductors has pushed the sustained recovery into early 2000, and therefore the growth forecast for 1999 is essentially flat overall.
Wafer Fab Equipment Revenue Forecast, 1997 to 2003 (Millions of Dollars)
1997 1998 1999 2000 2001 2002 2003 1997-2003
CAGR (%)
Total Wafer Fab Equipment
Growth (%) 20,171
-7.0 15,423
-23.5 15,155
-1.7 19,453
28.4 30,461
56.6 36,755
20.7 36,014
-2.0 10.1
-
Americas
Growth (%) 6,720
15.3 5,115
-23.9 5,637
10.2 6,924
22.8 9,443
36.4 11,393
20.7 12,610
10.7 11.1
-
Japan
Growth (%) 5,047
-22.9 3,626
-28.2 3,153
-13.0 4,186
32.8 6,835
63.3 7,564
10.7 6,821
-9.8 5.1
-
Europe
Growth (%) 2,380
-15.4 2,246
-5.6 2,320
3.3 2,829
22.0 4,109
45.2 4,924
19.8 4,677
-5.0 11.9
-
Asia/Pacific
Growth (%) 6,024
-7.2 4,436
-26.4 4,045
-8.8 5,514
36.3 10,074
82.7 12,874
27.8 11,906
-7.5 12.0
-
Source: Dataquest (January 1999)
Table 2
Semiconductor Demand Returning to Health
Since a sustainable recovery in spending and the equipment market must come from capacity buying, we must have a healthy chip market in 1999 to have a growing equipment market in 2000. Dataquest believes the industry has seen a major inflection point in the last several months. Pricing concerns in the logic and DRAM areas should subdue the growth rate for the chip industry to the 11-12 percent level, but unit demand will be good. As we enter 2000, several issues will align to create the fuel for a semiconductor boom: the improving capacity situation; the economies of the world will be on a recovery path; several electronic equipment markets such as digital TV set-top boxes, digital cameras and DVD systems will be in a position to ramp semiconductor demand; and the Y2K "millennium bug" issue will be behind us.
About the Author . . .
Clark J. Fuhs is vice president and director, Semiconductor Manufac-turing Analysis, for GartnerGroup/ Dataquest, San Jose. He may be reached at 408.468.8375 or via e-mail at clark.fuhs@gartner.com.
Semiconductor Capital Equipment Industry Update
Brett Hodess
We continue to believe that our fundamental analysis is correct that orders will bottom by year-end 1998 with several relatively flat quarters to follow before investment in 0.18 micron starts to drive growth in the second half of 1999 for a recovery beginning during 2000. There is still a fair amount of uncertainty associated with the timing and breadth of the recovery. However, the following observations put these uncertainties into context.
We estimate that orders will complete a "saucer-bottom" over the next two-three quarters, bringing 1998 to a 25-30 percent decline for equipment sales. Overall, semiconductor equipment orders are down 69 percent year-over-year and 71 percent from the peak in November 1997. Declines at this level make the 1998 downturn the worst in the history of the modern semiconductor equipment industry. However, we note that 15-20 percent of orders are normally for spare parts, service and upgrades; thus, it is apparent that new equipment orders have very little further decline to go.
Full year 1999 semiconductor equipment capital spending to decline 10 percent. Current indications are that orders will remain at the low level of the second half of 1998 into the first half of 1999. Given the weak first-half outlook, the second half of 1999 would have to show a 100 percent year-over-year growth rate to get 1999 flat with 1998. Our analysis shows that leading-edge investment in 0.18 micron is likely to show enough strength to drive growth by the second half of 1999. This would provide about a 50 percent year-over-year increase in the second half of 1999, resulting in a 6-10 percent down year overall. Figure 1 shows the order pattern we forecast for equipment over the next several quarters.
Figure 1
The improvement in the second half of 1999 will not be broad based. At this time, the major logic companies are planning to begin the ramp-up of 0.18 micron during 1999. Intel has announced that both Coppermine and Cascade will ship next year on 0.18 micron. IBM has announced a shrink of its copper dual damascene process to 0.18 micron. Texas Instruments has announced aggressive plans to push forward the technology on both its DSPs and its SPARC microprocessors for Sun Microsystems. Both TSMC and UMC, the two largest foundries, have announced plans to prototype 0.18 micron for key foundry customers in 1999. Samsung has announced 0.18 micron for a 1 gigahertz Alpha microprocessor. Micron has announced the shift to 0.18 micron for the 64 megabit DRAM. Spending by these major players will be focused mainly on upgrades to existing facilities for the specific equipment needed to get to 0.18 micron.
The industry is absorbing excess capacity-finally. 1998 is turning out to be a record year for fab closures. To date, 23 fabs have been shut, including three large, advanced fabs started up in the last two years. Capacity utilization is reaching its lowest point since the Semicon-ductor Industry Association began keeping data. We estimate that the cutbacks by major DRAM producers could reduce the supply growth rate of DRAMs to 50 percent for 1999. With demand expected to grow by 70-80 percent, the long-term average for the DRAM market could achieve balance going into 2000. As a result of the cutbacks, capital spending will drop to 17-18 percent of semiconductor sales in 1998 and 1999, below the 22 percent long-term average that has marked balanced supply and demand. With the excess capacity wrung out of the system by late 1999, the industry will move toward increased new fab activity in 2000 following the 0.18-micron upgrade activity in 1999.
About the Author . . .
Brett Hodess is managing director of semiconductor research at NationsBanc Montgomery Securities, San Francisco. He may be reached at 415.627.2355 or by e-mail at bhodess@montgomery.com.
1998: Is There a Road Ahead, or Is It Still 4-Wheel Drive Only?
G. Dan. Hutcheson
As everyone now knows, the chip equipment industry is coming off a very rough year. Overall sales were off by a third. Next year looks flat by all reckoning. There are few drivers to spur growth, and infrastructure issues remain firmly in place. The problem continues to be growth in excess capacity even though spending is down. This is a result of the global fire sale in wafer fab equipment. Typical prices are down 30 percent, making it possible to equip 0.18-micron fabs for around $800 million. The real problem with discounting is that it pushes up capacity, thereby pushing out the upturn. Manufacturing capacity at the end of last year was almost twice 1995's level.
Even still, conditions can change so fast in the current environment that it is almost impossible to tell what the road ahead holds. Forecasting has become a dangerous business in recent years. The average error for all forecasters of semiconductors was 9 percent prior to 1995. This range is now half again as large. Moreover, what used to be an annual event has become a monthly, weekly, even daily affair. The historical lags have not held constant in recent periods, and random events have swamped traditional market relationships and ratios.
Back in 1994, we were convinced that the silicon cycle still existed and that the market would decline in the latter part of 1995. I earned the nickname Dr. Doom that year. As it turned out, we had been right about the cycle, but missed the timing of the downturn by six months. The actual decline came in the first half of 1996. But 1995's error then led us to throw out all the historical data and be optimistic about 1996. Wrong again. We got 1997 reasonably right, correctly predicting a decline. By September of 1997, it looked like the industry was finally turning up. Capacity utilization was back in the 90-plus percent range, and equipment bookings had risen all year. Moreover, back-to-back recessions are extremely rare, and global financial crises have never affected the equipment industry. Then disaster hit in Korea, the PC market went south and shrinks more than doubled capacity.
The Asian financial crisis may have been the trigger for this year's disaster, but it was not the cause. Korea is the only major chip-producing nation that has had a financing problem. Taiwan's chip companies are mostly equity funded, so their problems are not related to the Asian financial crisis. Japan has been down since the early 1990s, so there was nothing new here. Moreover, as a nation, Japan still has the largest savings base in the world. So, while her banks may be weak, she is financially sound - with lots of cash. Japan's problem is a lack of decisiveness. So, little of this savings base is being applied to good use. But today's industry growth problems are more systemic than they are external.
So what is the source of the current industry problems? It's the PC market. Few were willing, last year and even early this year, to face up to the fact the PC market was maturing. At the time, even the most negative forecasters were basing their analysis solely on the effects of the Asian financial crisis. Everyone believed the PC market would hold up, many stating that the $1,000 PC would lead to a new phase of growth. That is, with the exception of many CEOs and IS directors, who were beginning to question the need for faster, high-priced PCs. Now, many are blaming the $1,000 PC. A trend, which we have regularly warned, that would lead to an industry disaster. It cut revenues for the most lucrative part of the chip market in half. Worse, it has done little to spur demand. PC unit volume growth has gone from the mid-30 percent range a few years ago to the low-10 percent range this year. Combined, these two trends have cut the heart out of the chip market. But the industry of technology is also paying the price for having violated many important business maxims.
Maxims Violated
Watch, don't listen to the customer. Everyone in the chip equipment industry should have known that the industry was building up too much capacity in 1994 and 1995. All you had to do was look at the obvious: too many fab announcements and too much spending. Worse, it continued to add capacity in 1996 and 1997, only serving to degrade the revenue-producing potential of the equipment (another way of saying chip prices fell). If, in 1995, you questioned this insanity, the answer was, "This is what customers are saying, and all these customers must know more than you do." Another good example is that in 1997, most of the world's lithography experts were proclaiming that "i-line lithography would never go below 0.3 micron" just as Micron Technology went way off the road and brought i-line into production at 0.2 micron. Everyone soon followed, which had a huge, totally unpredicted and unprecedented impact on capacity in 1998.
If you find yourself in a hole, stop digging. Most technology markets are inelastic in the short term, so dropping prices only prolongs a downturn. As mentioned above, the IC price drops that made $1,000 PCs possible did little to spur demand. But worse, the current price drops in chip equipment are allowing IC makers to continue to add capacity. Current manufacturing capacity in millions of square inches is almost double 1995's level, which, in turn, puts more pressure on chip prices and forces chipmakers to negotiate even harder deals on equipment. The cycle repeats itself until equipment companies break under the load. The solution? Don't take unprofitable business.
Monopolies drive growth in their formative years, but hinder growth once they are in place. Moore's Law continues to go on, but the software industry has come to a grinding halt as Microsoft has gained dominance. There are few new ideas coming from the software industry today, just more features you don't need or can't figure out how to use, none of which require significantly greater system performance. So the IC market is stagnant as a result. Microsoft has never been known for coming up with new ideas. It has been known for developing other companies' ideas into markets. But as its monopoly has developed, there have been fewer "others" to come up with new ideas. The chip industry loses as a result. The best way to solve this problem is to break up Microsoft. Look what happened after AT&T was broken up, or the European telecom monopolies. Also, once IBM was stopped, the computer industry blossomed again with the PC, the very trend that got Microsoft started. Likewise, IBM was made possible in part by stopping the monopolies GE and Westinghouse had built. Busting trusts ultimately reinvigorates an industry.
Confusing names confuse customers. Once confused, they stop buying. I am convinced that part of the PC industry's problem in 1998 can be traced back to Intel's abandoning of the X86 brand series and shifting to the Pentium. S
|
