|
|
 |
INDUSTRY NEWS
Fabless industry segment remains upbeat despite downbeat
news
 |
| BOUNCING BACK: Revenue is expected to rebound
following a 21% drop in Q1 2001 caused by low ASPs and a 13% drop
in wafer demand. |
It is an article of faith in the fabless semiconductor
community that its business model is more recession-proof than the version
operated by integrated device manufacturers. In its just-released 2001
wafer and packaging demand survey, the Fabless Semiconductor Association
(FSA) admits to spending "a little time gloating about 2000" because
its members performed better on key financial markers than its IDM competitors
"despite supply constraints."
Revenues in the fabless semiconductor industry segment
soared 68% in 2000, compared with 39% growth for IDMs, the report
points out. Its faith in the future is such that FSA predicts fabless
operations will generate 50% of all IC revenue by 2010. Today that
figure stands at 15%.
In 2001, however, that faith will be severely tested.
While still expressing optimism, Jodi Shelton, FSA's executive director,
told a luncheon audience at the launch of the annual report in San
Jose that the picture was changing daily, "and unfortunately for the
worse."
In an interview two days after the April luncheon, Shelton
was asked whether she believes the market is as healthy as a quote
in a news story about the luncheon had implied. "I certainly wouldn't
want to use the word 'healthy,'" she replied. "Fabless companies are
suffering from a combination of issues. One, of course, is demand.
The others are pricing erosion and an economic slowdown. And all three
things are happening at once."
The FSA survey touches on this convergence of factors,
noting that responding firms had begun feeling the first rumblings
of a downturn as the survey got under way. "The aggregate wafer decline
of 13% from actual fourth-quarter 2000 numbers to forecasted first-quarter
2001 numbers is the first such decline recorded in many years for
the survey," the authors point out. Many respondents reduced their
demand projections for the first quarter of 2001 to adjust for their
high inventory levels at the end of 2000.
"This expected demand dip is entirely consistent with
recent industry announcements of dramatically lowered anticipated
revenues for Q1 2001," the FSA report notes. "Thus, the root cause
of Q1 2001's drop in demand can be attributed to anticipated inventory
corrections."
Unlike previous downturns driven by excess capacity
alone, analysts point out, this slump features a toxic mix of overcapacity
and lessening demand. FSA's overview predicts a 13% drop in
wafer consumption between the last quarter of 2000 and the first quarter
of 2001 with an accompanying 21% dip in revenues during the same period.
These results suggest what FSA calls a "meaningful decline" in the
average selling prices of finished wafers.
Despite having a slightly darker cast, the survey's
overall picture remains rosy. Fabless wafer demand will grow approximately
4% in 2001, while IDM demand is forecast to decline by at least 8%.
Foundries posted 72% growth themselves last year over 1999, the report
notes.
These positive signs reflect the nature of the fabless
and foundry business, Shelton says. Despite the triple dose of bad
news, "fabless companies," she notes, "have traditionally always fared
better [than IDMs]. They fared better in the good times, and they
fared better in the bad times."
The reason? "They seem to do better because they don't
have fixed costs. All they have are variable costs." IDMs "running
at 40% to 50% utilization...don't have a broad array of products.
They have one type of product," Shelton argues. Unlike manufacturers
operating their own fabs, "when our foundries have low utilization,
the costs are not passed on," she emphasizes.
 |
| BIG PIECE OF PIE: In 2002, a majority
of fabless wafers will be made at processes 0.18-µm
or lower, the FSA survey predicts. |
In addition to the 68% rate of growth in 2000, the
fabless industry, now numbering approximately 625 companies around
the globe, is a growing presence. The survey points out that most
semiconductor start-ups now "take on the fabless business model."
According to FSA, at least 10 of the firms launched successful
IPOs last year. Among them were Quantum Effect Devices, Silicon
Laboratories, Centillium Communications, and Marvell Technologies.
Startup Malaysian foundries, such as Silterra and 1st Silicon,
and Tower Semiconductor in Israel are poised to give Taiwanwhere
a concentration of firms such as TSMC and UMC provide more than
65% of all worldwide foundry servicesa taste of competition,
Shelton points out. As noted in the report, 1st Silicon expects
to reach 20,000 wafer starts per month this year, while Silterra
announced an $80-million order less than three months after installing
its tool set.
An increase in outsourcing by IDMs has contributed to
the good fortunes of the fabless and foundry model, the survey says.
Between 1999 and 2000, IDMs doubled the amount of funds spent on outsourcing.
By the end of last year, FSA notes, the trend appeared ready to slow
somewhat as several chipmakers found their fab utilization rates tumbling.
Although Shelton believes fabless companies can
withstand the rigors of the downturn, potential problems loom.
Rapid shifts in demand, further cuts in foundry capital outlay,
and an upsurge in foundry wafer demand by IDMs could all cause
the pain to continue after the worst of the industry's slump has
passed. Shelton singles out one risk in particular: Nearly 60%
of fabless companies "are geared toward the communications markets.
In the overall industry, PCs are still the dominant sector. Interestingly
enough, that [end-market] has sort of stabilized. That is one
risk that may change [the forecast]."
Still, some analysts are bullish on the foundry segment
despite the downturn. In March performance and market evaluations
of TSMC and UMC, Thomas Weisel Partners forecasts a resumed increase
in demand for communications, multimedia, and consumer chips late
in 2001 and continuing into 2002. Foundry wafer shipments will bottom
out in the second quarter of this year, according to the firm. Weisel
pointed to outsourcing by IDMs and fabless companies, which are core
customers of the foundries, as principal drivers for long-term growth
in the fabless market segment.
The survey notes that the amount of wafers fabless companies
devote to PCs and peripheral devices continues to diminish as the
end-market demand slides toward 50% of IDM revenues. With its emphasis
on advanced logic and mixed-signal customer needs, the fabless segment
saw its share of wafers sold for communications-related applications
rise 5% from 2000. Many of the additional processed wafers were targeted
for wireless communications devices. In contrast, the share of wafers
fabless companies allocated to the PC end-market shrunk by 8% in 2000,
the survey points out.
Shelton emphasizes that fabless firms are more than
equal to the technological challenges of the submicron era. Each year,
more companies are shifting their wafer process to advanced nodes.
In 2000, 0.35-µm processes accounted for 39% of all wafer output.
The demand for 0.18-µm processes stood at a paltry 1%. However,
by the fourth quarter of 2000, approximately 18% of all fabless wafer
demand was for 0.18-µm processes, she notes. That figure put
the fabless sector on a par with its IDM counterparts, which showed
an industry average of 19% at that technology node.
The survey forecasts a majority of fabless wafers in
2002 will be processed for the 0.18-µm node or below, a demand
that the existing infrastructure can support, particularly at the
three main foundries. According to the report, four metal layers were
used almost exclusively for the 0.35-µm process node in 2000.
A "significant number" of responding companies have been successful
with four-metal-layer processes at 0.25 µm. By 2002 more than
80% of all fabless wafers will feature five or more layers of metal,
with 75% of them at or below the 0.18-µm node and the rest at
0.25 µm.
Fabless firms have made "unbelievable investments" that
have contributed to the accelerated growth at the advanced nodes.
"I do believe the fabless companies are ahead of the IDMs in taking
advantage of that," the FSA executive director asserts. "I heard the
other day from Applied Materials that they go to the fabless community
when they're building their next generation of products. Fabless companies
are on the leading edge of design. That makes sense because they focus
on design."

MicroHome |
Search | Current Issue | MicroArchives
Buyers Guide | Media Kit
Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com.
© 2007 Tom Cheyney
All rights reserved.
|
|
 |
 |
 |
|