BARRON'S: Signs Of Life: Renewed Demand For Semiconductors Is Percolating Up To The Big Chip Foundries
By Leslie P. Norton
If anyone knows which way the wind is blowing in the semiconductor industry,
it's Morris Chang. As the founder of Taiwan Semiconductor Manufacturing, which
manufactures 6.5% of the world's integrated circuits, in dollar volume, he's
privy to the forecasts of a host of the industry's biggest guns, including
Broadcom, Motorola, STMicroelectonics and Qualcomm. They all rely on Taiwan
Semiconductor Manufacturing for some or all of their chip manufacturing.
So it's worth noting that in a recent interview with Barron's, Chang
pronounced upon a forthcoming recovery in one of the most beat-up areas of the
semiconductor sector, communications chips.
Morris Chang is not alone in his optimism. His fiercest rival, Peter Chang (no
relation), chief executive and vice chairman of United Microelectronics Corp.,
shared his opinion with Barron's that communications chips are recovering.
"It's not definite, the indicators are not clear, and to make a dramatic
recovery this year isn't possible," Peter Chang cautions. But he reports that
some customers have increased their forecasts for fourth-quarter orders. He
won't identify the specific kinds of chips, nor the customers. But big buyers
of UMC communications chips include Xilinx, Alcatel and Ericsson.
An upturn in communications chips, in turn, could signal a recovery in the
making for manufacturers of wireless handsets, such as Nokia, Motorola and
Ericsson, all of which have been been hitting new lows as investors have thrown
in the towel on the group. (STMicroelectronics, it should be noted,
manufactures chips for Nokia, the No. 1 handset maker, and Infineon makes chips
for Siemens, the No. 3 in the market.)
TSMC, with revenues last year of $5.3 billion, and UMC, with $3.5 billion, are
the largest semiconductor foundries in the world. They make their money making
chips for other companies. And when times are good, as they were over the past
few years, they make lots of money, with net profit margins in the 40%-plus
range. This year, however, most analysts expect margins to fall to 10% or less.
Foundry customers come in two basic types: design firms such as Broadcom,
which are known as fabless producers because they do none of their own
manufacturing, and integrated device manufacturers such as Intel, which
manufacture their own chips but farm out their excess demand. As recently as
December, they operated at full capacity amid booming demand for personal
computers and networking chips.
Today their plants are operating at less than 50% capacity. And despite the
half-hearted enthusiasm that ripples briefly through the market at every hint
of a market bottom -- most recently in Intel's earnings announcement last week
-- few industry observers see demand picking up anytime soon.
And for investors, foundries are a capacity-utilization story. The enormous
fixed costs of the fabs remain, whether they are running at full capacity or
50% capacity. Reducing workforce barely makes a dent; at TSMC, for example,
labor makes up just 6% of costs. No wonder, then, that the outlook seems bleak
to almost everyone, save for Chang and Chang.
Consider the market for DRAM memory chips. Gartner Dataquest forecasts that
industry revenues will plunge 56% this year -- the worst one-year decline in
history -- and won't show a strong upturn until 2003. Overall, Gartner sees
worldwide semiconductor sales falling 17% this year from an all-time high of
$68 billion last year, with "a fairly slow recovery" from there. IDC is even
more bearish, expecting a 21% decline this year with a bottom not coming until
mid-2002.
While neither TSMC nor UMC is a big manufacturer of DRAMs, many believe the
collapse in DRAM pricing is precursor to a similar collapse in high-end
microprocessors and logic chips. No one, for example, expected DRAMs to plunge
below $2, which is what mainstream 128 megabit chips fetched last week. That
compares with $16 a year ago, and it's below the cost of production for most
manufacturers. At the same time as DRAM prices are falling, some specialists
such as Hynix Semiconductor are diversifying into higher-end lines to avoid
shuttering capacity, creating new competition for the foundries.
New competition, too, is on its way in the form of new foundries. China, for
example, is building three new billion-dollar chip fabs, and foundries are
sprouting up in Europe, Malaysia and elsewhere. That's one reason Merrill Lynch
figures average selling prices for semiconductors have started sliding more
rapidly, with a 12% decline in May.
So when Morris Chang sticks to his tale on a recent steamy day in Hsinchu, a
sterile industrial park that's home to seven TSMC fabs, you might be forgiven a
smirk or two. It's a couple of weeks after an earthquake hit Hsinchu, after
all, and just a year and a half after a colossal quake halted production for
several days. This time, production continued smoothly. But even if it hadn't,
capacity utilization is so low that any production loss could be quickly
recovered.
TSMC's chairman has a placid air that in no way masks his keen vision. And as
elder statesman of the foundry industry -- he launched TSMC in '87, after a
long career as an executive with Texas Instruments and General Instrument --
he's worth paying attention to.
Specifically, Morris Chang sees an upturn in PC demand in the second half,
since the cleanout of excess inventories is virtually complete. "We've either
reached bottom or are close to the bottom now." Intel has slashed prices for
its Pentium 4 processor, and Microsoft will soon launch its Windows XP
operating system. Those two factors combined should begin to drive PC sales
before the end of the year.
PC chips account for a third of TSMC's sales, and communications chips are
another third. Consumer devices contribute some 20%, and there are signs of
life too. Set-top boxes, for example. "That's something we didn't have during
the first half." And don't forget the next generation of videogames, which
should be in full production in time for Christmas.
More important, Chang anticipates an upturn in communications chips before
anybody expects. "The improvement will start early next year, and if you mean a
very strong increase, then you'll see one in the second half." If true, that
would be even better news for UMC and and Singapore-based Chartered
Semiconductor Manufacturing, for which telecom chips make up nearly 50% of
sales.
In its earnings announcement last week, Nokia, too, suggested the bottom was
near in wireless handset sales. If so, that would trigger an upturn in demand
for communications chips.
Supporting the bear argument, however, Applied Micro Circuits, which produces
chips used in networking equipment, slashed its fiscal first-quarter (ended
June 30) sales forecast as order cancellations mounted.
Nor does Dan Heyler, Merrill Lynch's lead chip analyst in Asia, expect a
near-term turnaround. Heyler figures capacity-utilization figures for TSMC will
fall to 40% or below in the coming months. "That's overly pessimistic," snaps
Chang. "Utilization was 45%-46%, maybe, and I expect the second half to be
better. We will prove it by the facts."
UMC's Peter Chang, who has been notably more bearish in his forecasts than his
rival Morris Chang, also now thinks the bottom is near. "The overall economic
situation is getting close to a bottom," he says. "Then we'll recover." And
given the inventory clearout by PC makers and short life cycles of consumer
products, he adds, a rebound, when it comes, could be swift.
Other beacons have emerged in the storm of horrible news. This month Conexant
Systems, a major IDM and the biggest maker of chips for modems, said it would
undertake new restructuring initiatives, including cutting its workforce,
taking charges, and accelerating its move to a fabless model. It plans to shift
most of its semiconductor production to Taiwan and quit investing in
manufacturing capacity.
Make no mistake. There's still plenty of bad news. Visibility is befogged, and
limited to a matter of weeks. Says Frank Wen, UMC's general manager: "It used
to be that customers gave us a purchase order three months ahead of time. That
kind of luxury doesn't exist anymore. It's last minute, say one to
one-and-a-half months ahead. Then they want a rush." Pricing is clearly on the
slide. The floor for DRAMs vanished, the bears reason, so what's to prevent
prices for higher-end chips from heading south?
Yet short sellers have been burned. Even as technology shares have been
battered this year, ADRs of TSMC's are up 17.8%, while UMC's are flat.
Chartered is down 7.4%.
One reason shares of foundries are not making new lows is that many investors
view them as a bellwether on a reviving global economy. With the Federal
Reserve cutting interest rates aggressively, the Philadelphia Stock Exchange's
semiconductor index has sprung higher in recent months, though it's down 2.6%
this year. In the face of a steady diet of negativity, investors remember that,
coming out of the last downturn in 1998, the so-called SOX index rose six-fold.
In other words, investors aren't waiting to see a bottom out of fear that
they'll miss the first big leg up on these stocks.
Complicating matters is that TSMC and UMC are the first- and third-largest
components of the benchmark Taiwan Weighted index, together accounting for 20%
of the market. That means they're first to suffer when foreign investors choose
to dump Taiwanese equities during times of saber rattling between Taiwan and
mainland China. And they're the first to be snapped up when investors move into
the market, as they did when Morgan Stanley Capital International announced it
would add Taiwan to its global index.
Foreigners suspect, too, that Taiwan steps up to buy shares of the two as a
way of lifting the overall market.
Investors, should also be concerned about the valuation disparity between the
shares traded in New York and Taipei. For example, ADRs of TSMC trade at a 70%
premium to the value of the local shares, while ADRs of UMC command a 44%
premium. At their peak in February 2000, however, the premium on TSMC ADRs
topped 100%. Demand for the ADRs is so high, skewing the prices, because only
approved foreign institutions are allowed to own local shares.
So while TMSC shares look particularly cheap in Taiwan, trading at just 18
times estimated 2002 earnings, as the table on this page indicates, the ADRs,
with a P/E of 30, appear less so. Likewise, UMC shares trade at 13 times
estimated 2002 earnings, while the ADRs fetch a pricier 19 times earnings.
Ask Morris Chang what TSMC will look like in five years, and he looks at a
chart of TSMC's sales by customer and by technology. Right now, a third of
TSMC's production is in communications chips, another third in computer chips,
16% is consumer, and 18% is memory. In five years, consumer chips will take a
greater share, memory will shrink, and the sales volume of communications chips
will exceed that of PC chips.
And given Moore's Law, which says that chip capacity doubles every 18 months,
the complexity of chip manufacturing will continue to rise. Already, the cost
of building a new cutting-edge fab has grown to $3 billion. (More basic fabs,
like those being built in China, cost far less and, as such, will not compete
in the high-end, high-margin chip business.) And without detailing the physics
of how chips will process ever more information ever faster, it's safe to
assume that the cost of entry into state-of-the-art chip manufacturing will
continue to rise. In other words, the foundry model clearly has legs.
Indeed, during the first quarter, UMC's sales to IDMs rose to 28% of total
sales from 26% in the fourth quarter. Within 10 years, says Morris Chang of
TSMC, IDMs will go fabless. The last to give up will be the makers of
microprocessors and DRAMs, because manufacturing is such an integral part of
the business. But the chance to increase margins will become increasingly
appealing, and even the holdouts will eventually turn to foundries.
In the meantime, with demand still anemic, consolidation rumors have begun to
spread. Specifically, they involve Chartered Semiconductor, which relies
heavily on IDMs such as STMicro for sales and is highly dependent on
communications chips. What's more, its fabs are not as technically advanced as
those of TSMC and UMC. Both, it has been said, are interested in buying
Chartered, though both have denied it. Additionally, Chartered is controlled by
the Singapore government -- not a ready seller -- and TSMC and UMC are hardly
likely to need additional capacity in last-generation manufacturing. Still,
says Peter Chang of UMC, "in this industry, anything's possible." A Chartered
spokeswoman said the company doesn't comment on market rumors.
Ultimately, the businesses aren't likely to improve until they see an
improvement in demand from the foundries' biggest customers.
What the stock market cares about, however, is an inflection point. Sure, the
global economy remains weak, and pricing pressures are on the rise. But Lucas
Ward, the lead semiconductor analyst at J.P. Morgan, maintains current
valuations will reward investors. The stocks, he says, discounted the downturn
well before other segments of the technology market, and up to a year before
business deteriorated. "You have had every negative data point thrown at these
stocks -- the billions of dollars in inventory, the preannouncements, the bad
news in the economy, the bad utilization rates -- and they haven't tested the
lows they hit in October. The reason is they're trading at rock-bottom levels." Investors can only hope he's right. |
