Get ready for another forecast hike as analysts wait for DRAMs to join boom
By J. Robert Lineback
Semiconductor Business News
(08/24/00, 09:55:16 AM EDT)
SANTA CLARA, Calif. -- Don't look now, but some industry analysts are beginning to think their recently revised mid-year forecasts for much higher semiconductor growth and capital investments might still be too low.
During a panel discussion of semiconductor equipment analysts here on Wednesday, the consensus was that wafer fab investments are now 80% higher than a year ago and chip revenues could exceed 45% growth in 2000 over 1999 sales. Just six weeks ago, many market researchers hiked their projections for semiconductor capital spending growth in 2000 to the 60-70% range (see March 1 story). In June, a mid-year 2000 forecast of 31% growth in chip sales to $195 billion issued by the Semiconductor Industry Association (see June 7 story), followed by still higher projections from analysts, who pushed some growth estimates toward 40% last month.
But now new "high-side" forecasts are suggesting that the percentage of growth should be nudged up again after a strong summer of plant investments and still-solid unit demand for semiconductors.
The kicker in the current outlook this is that industry growth rates in both plant investments and chip revenues have not yet felt any significant impact from the lagging DRAM segment. Memory manufacturers are still holding back potentially huge investments in new capacity while they wait for clear signs of shortages to surface, said analysts from Dataquest Inc and VLSI Research Inc. during the panel discussion, hosted by the Semiconductor Equipment and Materials International (SEMI) trade group.
"Why spend more money on capacity when you don't need more capacity?" asked G. Dan Hutcheson, president of VLSI Research in San Jose. The analyst reasoned that DRAM manufacturers in Asia were not likely to make major new investments until they see more profits and higher prices. "I'm just happy [about the lack of DRAM fab investments in the current strong business cycle] because they have been the problem all along," quipped Hutcheson, referring to the downfall of the last cycle five years ago when too much memory capacity was brought on stream.
Dataquest analyst Klaus-Dieter Rinnen suggested that DRAM makers were moving closer to making significant investments in new wafer-processing capacity because shortages of memories were beginning to appear. He cited Samsung Electronics Co. Ltd.'s decision to sharply hike capital spending to $4.1 billion in 2000 vs. its original budget of $2.6 billion at the start of the year.
The slump in DRAM fab investments "bottomed out last year," said Rinnen, who is chief analyst and director of semiconductor manufacturing analysis at Dataquest in San Jose. Dataquest is predicting a shortage of DRAM production capacity will start in the third quarter of 2000 and last until the second quarter of 2002, eventually giving way to another memory glut in the third quarter of that year.
The current strong investment cycle in new chip plants continues to be driven by logic manufacturers and foundries. Last year, capital investments in memory plants represented just 25% of the total, Rinnen said, but in 2000 that percentage will grow to 32%, counting both flash and DRAM investments. Typically, 30-35% of semiconductor capital investments go to memory production, but in up-cycle--such as the industry has experienced for 18 months--DRAM investments have represented 45-50% of the total, said the Dataquest analyst.
VLSI Research's Hutcheson noted that lower investments in DRAM production capacity is partly a result of lower memory requirements in many of the system applications now driving semiconductor unit growth. Wireless systems and cellular phones, for example, require flash memory, but not much DRAM compared to PCs, which have been the main growth engine of the past three recover cycles in the industry, he said.
Overall, chip pricing has fallen sharper than expected in 2000, Hutcheson told the SEMI gathering on Wednesday. VLSI Research uses a model to track chip pricing trends that starts the year at a value of 100 and in a typical year the figure will drop to 70--representing the historical Moore's Law curve of 30% annual reduction in the cost of a transistor per megahertz. "We're now running in the low 60s," Hutcheson said, referring to the erosion of average selling prices for chips. "And so far this summer, we haven't seen the forecasted uptick in memory prices."
Currently, Hutcheson is placing semiconductor capital spending growth--without significant involvement from DRAM makers--at 87% over 1999. His models show 2001 will grow another 30% in 2001. Semiconductor revenues are now expected to increase 44% in 2000 and 45% in 2001, according to VLSI Research's preliminary outlook. Dataquest is also suggesting a high-side to its forecast of growth in the high 70-percent range for capital equipment spending. Semiconductor sales growth could be strong enough to push total chip revenues to the $300 billion in 2001--a year ahead of the current outlook, said Rinnen after the panel discussion.
But analysts agreed on Wednesday that the current business cycle will continue longer and stronger than previous cycles. "I think people are underestimating the strength of this cycle," said Clark J. Fuhs, vice president at J.P. Morgan Securities Inc. Fuhs left Dataquest as its chief semiconductor equipment analyst a couple of months ago to track chip supply and demand at the financial brokerage house.
Fuhs said semiconductor unit growth "has been hanging in the 30-35% range for the last 6 to 8 months" compared to last year. He added that in the last three business cycles (in the late 1980s and '90s) the industry neve had been that high in chip unit growth. "We've never been +30% growth year-over-year."
Semiconductor fab utilization, tracked by the U.S. government, continues to show American plants running at over 95% since late 1999, Fuhs said. He also noted that analog semiconductor sales continue an unprecedented run as the leader in growth. In the past 20 years, analog circuits have surged ahead of digital IC growth rates early in a semiconductor recovery, but that situation usually plays out within several months of the upturn, Fuhs said, citing market research at Dataquest during the early '80s.
"We have been in the cycle [of recovery and strong growth] for a little over a year now, and analog is still leading the charge. So this is a tremendous cycle," he told the happy group of semiconductor capital equipment and materials executives. "We have been in this cycle now for a good year-and-a-half, and we think it has at least that far to go." |
