IDT's Perham quoted below.....SemiBull
================
Gun-shy chip makers hesitate to predict bona fide recovery
Cautious industry now expects growth to strengthen
in second half of the year
By J. Robert Lineback
After three lousy years and then a surprise chip recession
in 1998, the semiconductor industry seems determined not
to be fooled again by any false signs of recovery. Most
chip managers were encouraged by the uptick in chip sales
that showed up in the final months of 1998, but few of
them are willing to forecast a bona fide recovery in 1999
until they see more proof over the next several months.
And after last year's disaster, who could blame them?
A year ago, industry forecasts were generally upbeat
after a weak 1997 and called for increases in the
12-to-17% range for worldwide chip revenues. But the
bottom soon fell out and by midyear, chip sales had
plunged 14%. Then they flattened out, and started
improving a bit in the fourth quarter.
The 1998 forecasts had been blindsided by too much
production capacity, a glut of chip inventories at the
start of the year, Japan's recession, and
greater-than-expected fallout from the Asian financial
crisis.
This time around, the forecasts are cloaked with caution.
They predict that dollar-based revenues will increase
from 5% to as high as 12% in 1999 over 1998 chip sales,
which ended up in the $125 billion range. Most forecasts
see a sluggish first quarter, followed by strengthening
sales in the second half. This recovery will come from
rising average selling prices on ICs, particularly DRAMs,
and from much stronger growth in units and revenue.
Once again, though, it's feast or famine because the
capital spending flow for new fabs has been nearly shut
off. Unneeded plants are being shuttered. Forecasters are
now expecting shortages to show up by 2000.
To them, the only question is when do these shortages hit
-- late 1999 or in 2000?
Believe it or not, some industry analysts and managers are
even suggesting now that the threat of "millennium bugs"
in software could help push up semiconductor demand in
1999. They figure that cautious end-equipment users will
opt to replace gear rather than risk "Year 2000"
programming glitches, and that some chip buyers might
stock up on parts in case the supply chain is disrupted
temporarily by the Y2K bug.
But forecasts of growth are ignored now by most chip
makers, who are hunkered down for a slow start in 1999.
"No one in the semiconductor industry is popping the cork
and celebrating the outlook for the new year," maintained
Jeff McCreary, senior vice president of worldwide sales
and marketing at Texas Instruments Inc. in Dallas. "We
are all being somewhat conservative, and that's why we
feel better about the year."
For most of the chip industry, 1999 is just a case of
guarded optimism. "We do see growth [coming], but we're
definitely not counting on it," said Leonard (Len) Perham,
president and CEO of Integrated Device Technology Inc.
in Santa Clara, Calif. "Right now, we're taking a
speculative approach [to 1999]. By the end of March," he
said, "I believe we'll have a better idea of what's going
on."
The picture is brighter in some respects today that it was
a year ago, even though current forecasts don't reflect it.
"We see more orderly markets today than we did a year
ago and customers are telling us they have very little
inventory," said Don Macleod, chief financial officer at
National Semiconductor Corp. in Santa Clara, Calif.
"But our attitudes are clouded by concerns about the
seasonal nature of our business," he said, referring to
chip consumption by the producers of personal computers
and cellular phones. As a result, "we are looking very
anxiously for [favorable] signs in the January-February
time frame from these two major business segments," said
the National CFO.
The business is not all that promising in Japan. IC makers
there have all but stopped capital spending until their
next fiscal year begins in April. "From feedback at
Semicon Japan [in December], it looks like their capital
spending [in calendar 1999] will be either flat or down
from 1998," said John Scruggs, vice president and
general manager of the Automated Test Group at
Hewlett-Packard Co. in Loveland, Colo. "Money will be
spent more selectively," he said. "My impression is that
there will be fewer new fabs and potentially more
[money] going into the backend and test for new
technology requirements such as high-speed Rambus
[DRAM] devices."
But there could be an upside to the industry's present
cautious state of mind. This could be setting the stage for
the next boom period, according to market observers.
"Our opinion now is that there's a real chance that 1999
will end up stronger than everyone is expecting," said
analyst Bill McClean, president of IC Insights Inc. in
Scottsdale, Ariz.
"In 1993," he said, "it was not until the third quarter that
people realized that the upturn was for real. Japanese
manufacturers didn't increase their investments
throughout 1993 because they weren't sure it was going
to last. That was a key factor in the strong upturn that
continued into 1995."
So, he said, the question everyone is asking now is, "is
1999 more like 1993, or 1992, which set the stage for the
next boom period?" He figures there is "a very good
possibility for substantial growth by the end of the
year."
IC Insights "officially is calling for an 11% increase [in
global chip sales] to $140.4 billion in 1999 vs. $126
billion in 1998, which was 8% lower than chip sales in
1997," McClean said. But the market researcher also
thinks there is a 35% probability that global chip sales
could rise by more than 15% in 1999 if four things
happen: worldwide economic growth remains strong, Asian
economies recover by midyear, total capital spending
declines, and an inventory buildup begins over concerns
about supplies being disrupted by Y2K computer glitches.
Few chip managers believe that concerns over Year 2000
software glitches will trigger significant chip buying, but
they still have heard customers discussing the issue. "It's
impossible to call at this time and it could be a little bit
of the hype," said TI's McCreary. "In the area of
[controlling] double ordering, we probably fool ourselves
on how much things are under control."
But TI, like other major chip makers, is still confident it
can respond to an unexpected upturn in demand without
taking any action now. "We have enough brick and mortar
in place, and while it would put strains on us, we are in a
position to take quick action," McCreary said.
With chip makers still postponing many of their major
investment plans for building new production lines, the
stage could also be set for severe shortages in fab
equipment for cutting-edge process technologies once the
recovery is in full swing. Many suppliers of manufacturing
systems and materials have scaled back production
drastically in an attempt to survive the severe recession
in capital spending, warned analyst Risto Puhakka at VLSI
Research Inc. in San Jose.
Fab investments for most of 1999 will be limited to
"technology buys" for new 0.18-micron processes, such as
copper interconnect. In fact, copper deposition will be
the hottest capital equipment segment, with revenues
growing from $64.5 million in 1998 to $199.2 million in
1999. But total capital spending in 1999 will grow only
3% to $26.4 billion vs. $25.6 billion in 1998, which was
down 28% from the 1997 total, according to the San Jose
research firm.
One cure for the chip industry's slump in 1998 was
painful, but successful: cutting back production. The
closing of about 26 fabs eliminated "close to 10% of the
existing capacity," estimated Jean-Philippe Dauvin, vice
president and chief economist at ST Microelectronics,
based in Paris.
"This starts to be a significant [cut in chip capacity]," he
said. Also in 1998, "there were [capital] investment cuts
of 13%, and in 1999 there could be another 5%
reduction," said Dauvin, who also is chairman of the World
Semiconductor Trade Statistics (WSTS) organization.
All told, he said, "about $38 billion in fabs have been
postponed. It takes between 15 and 20 months to start a
new fab and bring it up to 5,000 wafer starts a week."
Dauvin doesn't see the industry starting to reinvest
before September 1999, which would mean that these new
fabs wouldn't reach volume production until around the
end of 2000.
The downside here is that if chip shortages develop,
higher ASPs could have a negative impact on high-volume
products. "The consumers won't be happy and they might
stop buying equipment," Dauvin added.
A handful of new fabs have been announced by major chip
makers with most of them starting construction within the
next year. A joint-venture of Philips Semiconductors and
Taiwan Semiconductor Manufacturing Co. (TSMC) plans
to start up a $1.2 billion fab in Singapore in late 2000
based on the current outlook.
"About a third of the world's capacity was in excess of
demand a year ago and the industry has been working to
reduce it," said Walter Conrads, director of international
marketing and sales for Philips Semiconductors in
Eindhoven, The Netherlands. But around mid-1999,
"capacity and demand will mesh again and we'll see a
firming of prices."
The average selling price of an IC fell by 9% in 1998 to
$1.87, after dropping 15% in 1997 and 10% in 1996,
according to IC Insights. But a reduction in overcapacity
and a firming of DRAM prices in the second half of 1999
will help push up ASPs by 7%, predicted analyst McClean.
Global unit consumption also is expected to rise in 1999
by 7.7% to 278 billion chips after edging up by only 1.1%
in 1998. "The 1998 price declines did not stimulate a
growth in unit demand like it did in 1997 [when total
revenues grew 4% and unit shipments rose 19.7%]," said
analyst Jim Feldhan, president of Semico Research Corp.
in Phoenix.
Flat unit shipments were due to slower growth rates in
end-equipment markets such as PCs and to the integration
of additional functions on chips used in portable devices,
Feldhan said. "Economic problems in Asia, Japan, and
Latin America also played a factor," he added. "If you're
not doing well [economically], you don't need to buy a
mobile phone."
IC makers aren't just waiting for old markets to recover;
they're also pushing hard to use advanced technology to
build more powerful chips for new applications and
markets.
"Whatever products we had when we went into the
downturn will not be the ones that will take us out," noted
Rich Forte, CEO of Vantis Corp., the programmable logic
device subsidiary of Advanced Micro Devices Inc. The
new sales growth will be driven by "higher-performance
products," whether they are programmable logic or
microprocessors or communication ICs," he said. "In
programmable logic, higher performance is needed for new
telecom applications such as the terabit-per-second
routers."
Research by Semico backs up such new development
activity. PLDs will snap out of their sluggish, 3% growth
rates in 1998 and will shoot up 11% to $2.36 billion in
1999, the market researcher predicted, because new
telecom applications will require higher-performance
parts.
IBM Corp.'s Microelectronics Division is pushing hard to
implement copper-based 0.18-micron technology for
high-growth, high-performance communications
applications. It also is expanding its use of
silicon-germanium (SiGe) and silicon-on-insulator (SOI)
technologies for communications gear.
"These systems need increased speed and bandwidth for
network infrastructure," said Pulin Shah, manager of
ASIC product marketing at IBM Microelectronics in East
Fishkill, N.Y. He estimated that total ASIC revenues in
1999 will grow 15%, while communications applications
will shoot up 25% and data-processing chips will rise by
10%.
While the low-cost PC is doing a good job in driving
high-unit volumes, the trick for IC vendors will be to
make an acceptable profit on such business. "We have
instigated a number of moves to get our costs in line by
taking pretty large writedowns in our Oregon fab
operations and closing down one semiconductor factory [in
Silicon Valley]," said IDT's Perham.
The company is pursuing low-cost PCs with its
x86-compatible WinChip. IDT plans partnerships and
alliances to expand its efforts to capture more of the
sub-$1,000 systems.
"That business, in terms of units, will do very well in
1999," Perham said, but "whether the revenues go up or
stay flat is still a little uncertain."
Equally cautious but just as determined to play in the
low-cost PC arena is National. "We see a lot of
opportunities in 1999," enthused National CFO Macleod.
"Today we have creditable PCs selling for $599 from big
names [in the PC business], including 15-inch monitors. And
$399 boxes are making an impact on the consumer market.
This is creating greater opportunities," he exclaimed.
While there might be strong demand for low-cost PCs,
some companies have decided to forgo this high-growth
market because margins are too low, said analyst Tony
Massimini, who tracks microprocessors at Semico
Research. "How low can you go on ASPs and still make
money?" he wondered.
Many PC houses also are finding it hard to make
acceptable profit margins with the "bare-bones" PCs,
Massimini said. "If [the cost of] one component changes
ever so slightly -- if DRAMs go up, or if you can't find
any cheap hard drives -- your profit margin could erode
by several points overnight."
But customers and corporations are hooked on the lower
prices, indicated IDT's Perham. "Let me tell you that
IDT, as a computer user, will not be paying $1,500 for
PCs," he said. "I think that sensible corporations
worldwide also are going to be paying a lot less than
$1,000 for their computers in the future." |
