Dataquest
The Semiconductor DQ Monday Report, Issue 27 (exerpts)
July 13, 1998
1. IBM and STMicroelectronics Forge Partnership: very positive comments from DQ
2. IBM's Interconnect Announcements -- There's More Here than Just Packages! :contains interesting speculation about possibility of Moto going fabless and using IBM as their foundry!
3. Analyst Note: Wafer Fab Equipment Market Forecast
Update: This Story Is Getting Really Old: now they say things won't get better till 2000
1. IBM and STMicroelectronics Forge Partnership
IBM and STMicroelectronics (ST) have joined forces to
accelerate the development of system-level ICs or systems-on-a-
chip. The two companies have agreed to exchange intellectual
property (IP) and jointly develop ICs for current and future
data storage applications and PC-compatible information
appliances. With a focus on hard disk drives, ST will gain
access to IBM's PowerPC technology and IBM to ST's next-
generation DSPs and read channel technologies. The companies
will also jointly develop PC-on-a-chip ICs. The alliance is
quite deep, in that it includes a broad patent cross-licensing agreement.
Dataquest Analysis
This is clearly a major alliance considering that IBM is the
largest U.S. ASIC supplier ($1.5 billion) and ST is the largest
European ASIC supplier ($400 million). The combined effort of
these two companies is sure to expedite their presence in the
rapidly expanding system-level ASIC market. This could give a
customer the option for a second source on a large-volume
design. The system-level ASIC market is expected to grow from
around $4.5 billion in 1998 to $21 billion in 2002. Having a
comprehensive IP offering targeted at specific applications is
the ticket to play in this market. The combined IP offering by
both of these companies is impressive. In the last six months,
IBM added to its PowerPC cores PicoJava, Arm, and a new TI DSP
knock-off. ST has a solid line-up of IP and has some unique
mixed-signal capabilities. Specifically, it has a license for
Chromatic's Mpact media processor. System knowledge will also
play a lead role in knowing exactly what the system designers
need in their next-generation system. Companies such as LSI
Logic, with its pending purchase of Symbios to penetrate the
disk drive market, and National Semiconductor, with its focus
on the PC-on-a-chip market, are sure to take a long hard look
at the competitive landscape of these markets. Alliances will
play an increasing role as we move into the more complex world
of system-level integration.
By Bryan Lewis (bryan.lewis@dataquest.com)
Dataquest Analysis
This collaboration is of major importance for both companies
and has significant implications for the future. IBM and ST are
among the small number of vendors already possessing real
system on chip technology, and this agreement will boost the
position of both companies in the growing number of sectors
requiring this technology. This agreement probably creates the
largest (virtual) system on chip design team in the world. Yet,
the number of cores needed in this market is growing daily and
includes microcontrollers, DSPs, MPEG, high-speed interfaces,
memory, and many more. This is too much for any company with a
relatively broad application focus to attempt to develop.
Alliances and pooling of resources will become increasingly
necessary.
Storage IC applications are attractive, offering high volumes
in a growing market. This is an area where the companies have
complementary technology and therefore a good synergy in
joining forces. The deal is clearly good for the companies
today, but where could it lead in the future? On the face of
it, closer ties with ST could help IBM to better penetrate the
European semiconductor market. However, few hard disk drives
are manufactured or designed in Europe and this agreement may
have little visibility for semiconductor customers in Europe.
This will certainly be a great way for the companies to find
out how they can work together and could pave the way for
agreements in other market sectors, notably in digital
communications and consumer applications.
The PC on chip agreement could trigger market growth and
benefit both companies in different ways. Second sources are
often unnecessary, especially in ASIC markets. However, the
absence of a second source can inhibit a market, especially in
specific sectors such as automotive, which have traditionally
expected this. Second sources are also beneficial in high-
growth markets in which future demand is unpredictable. This
may apply to network computers and other information
appliances. In these markets, OEMs may need to increase volume
tenfold in one year and therefore need to be convinced of
security of supply. ST and IBM should jointly be able to
satisfy customers in this respect. However, the details of this
arrangement are unclear, as the two companies will not operate
compatible processes.
We noted in a previous DQ Monday article that for ST to grow
from its present worldwide market position as No. 10 toward the
top five, it must more aggressively target the PC markets. The
company is now entering this market through the back door, not
targeting the desktop market, but the higher-growth information
appliance markets. IBM's strength in system design coupled with
ST's strength in x86 and related core design could make a
powerful partnership.
The patent cross-license involves a huge portfolio, especially
from IBM. However, we believe that these patents are unlikely
to be used widely in design projects. Instead, the agreement
will foster closer work in common technology areas while
removing the problems of accidental patent infringement, which
are so common in R&D.
By Jim Tully (jim.tully@gartner.com)
2. IBM's Interconnect Announcements -- There's More Here than Just Packages!
Semicon West will be the staging area for IBM's announcement of
three new interconnect offerings. These new products are as follows:
- Ultra Fine Pitch (UFP) Ball Grid Array: a BT resin based
four-layer (max.) BGA package with a maximum I/O of 1088,
targeted for wireless applications
- Multichip Module-Laminate (MCM-L): MCM-L solution using flip
chip, wire-bond, or SMT interconnect with multiple substrate
materials, as well as edge connector design flexibility
- Glass ceramic package for high-end ASIC and MPUs, with off-
chip maximum I/O of 5500 and on-chip I/O of 1700. There are
current telecom designs at 1657 I/O.
IBM's design centers for these interconnect products are
located in Italy; Yasu, Japan; and Endicott, New York. IBM is
now more than a captive and merchant supplier of systems and
semiconductors. It has become a full turnkey foundry supplier.
Dataquest Analysis
IBM's interconnect products will be offered not only for IBM's
semiconductor products but also those from the merchant market.
IBM is essentially a large system operation whose revenue has
historically been generated by its electronic equipment
business and service sectors. From the semiconductor or IBM
Microelectronics side, IBM is basically in the category of an
integrated device manufacturer (IDM). An IDM by Dataquest
definition is a semiconductor supplier, merchant or captive,
that manufactures its own products. To meet its ever-increasing
internal silicon demands and reduce the increased cost of
advanced fab spending, IDMs, like IBM, or Intel and Motorola,
are increasing their usage of semiconductor contract
manufacturing services from foundry companies such as Taiwan
Semiconductor Mfg. Co. (TSMC) and United Microelectronics
Corporation (UMC). Very few companies fall into the category of
an IBM, which can be referred to as a full turnkey foundry
service. IBM now offers basic wafer processing, the subsequent
manufacturing operations of packaging and assembly, testing,
and drop shipment of finished IC products to the end customer
or distribution channel.
IBM Microelectronics' package foundry announcement is key as it
offers state-of-the-art architectures that very few others can
produce, especially for new RF wireless and satellite
communication frequencies. This is also beneficial for the
Americas region's fabless companies that need quick-turn design
support and manufacturing capability. The only dedicated
foundry service providers in North America are Orbit
Semiconductor and WaferTech. While there are dedicated package
and board suppliers in North America, only a few have the
interconnect capability being offered by IBM, and none offers
all of the services that IBM can provide.
IBM Microelectronics is positioning itself for the future as
both an assembly and chip-level business. IBM also maintains a
very successful board assembly business in North Carolina. The
packaging operations enhance its semiconductor and system road
maps. IBM becomes a full-service foundry for companies besieged
by the high cost of semiconductor fab and assembly operations.
IBM's announcement is key at this time. The semiconductor
industry is in the throes of a major industry downturn, with
rampant overcapacity and underutilization. The companies that
will survive this downturn are those that retrench and save
their R&D investments for future technology roadmaps. They will
not be spending on fabs or assembly operations. After years of
downsizing and reorganization, IBM is capable of surviving this
downturn by utilizing all of its technological capabilities to
the benefit of start-ups, fabless, and struggling semiconductor
companies. The fabless industry was born out of the last major
industry downturn, in 1985 through 1987. Start-up companies and
existing companies gave up on ideas of costly fab ownership and
contracted their semiconductor designs to SCMs. IDMs such as
IBM maintained in-house semiconductor R&D and kept their
advanced fabs, but mothballed their older fabs and became large
users of SCMs, saving huge capital spending dollars on new fabs.
Although this is pure speculation from Dataquest, a company
like Motorola could benefit tremendously as an IBM partner.
IBM's strategy is to expand beyond its services and data-
processing equipment sector. It has made small investments in
software and communications businesses/partnerships to grow
near-term revenue. Long-term, it will need to expand its
manufacturing capacity to maintain the estimated growth
expected from the communications industry. Motorola's
semiconductor strength lies in its microcomponents, logic,
SRAM, analog, discrete, and opto products. It also has R&D
projects in 1V GCMOS, ferroelectric memory, SiGe HBT, RF LDMOS,
smart sensors, and smart cards. It has multiple fabs, focused
on both leading-edge and lagging technology. Motorola's revenue
strength lies in its communications business operations.
Could Motorola go fabless' Would Motorola partner with IBM?
Motorola has been downsizing its role in certain areas of
costly semiconductor manufacturing since 1994. Its departure
from this market has been an emotional, hand-wringing
experience for its entire semiconductor segment. Motorola
departed from DRAMs in July 1997. During the first half of
1998, Motorola posted disastrous sales results from the
semiconductor sector. According to recent company reports,
Motorola's semiconductor products fell nearly 11 percent to
$1.81 billion. Motorola knows that it will have to continue
restructuring its semiconductor operations to enhance its
ability to compete. As of July 9, it had reorganized its
communications segments into one division. Motorola is
aggressively pursuing global participation in the
communications business, expanding into Latin America as well
as China and other Asia/Pacific countries. Motorola's biggest
competitors are AT&T (systems) and Lucent Technologies
(microelectronics). It is questionable whether Motorola will
continue to pour investment dollars into semiconductor
manufacturing, if it can find a solid foundry partner.
Together, Motorola and IBM would be formidable competition,
worldwide. Just food for thought!
By Mary Olsson (mary.olsson@dataquest.com)
3. Analyst Note: Wafer Fab Equipment Market Forecast
Update: This Story Is Getting Really Old
The following is an excerpt from a Dataquest Alert sent to clients last week.
For the past 18 months, we have projected that the 1998 wafer
fab equipment market would be "frustrating." 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. Aggressive investment in 0.25-micron technology
throughout 1997 contributed to the continuing overcapacity in
the industry. However, the economic slowdown in Asia and Japan
has made the 1998 spending environment downright ugly.
DRAM Overcapacity Continues, but Finally Constraints on Supply
In DRAM, there has been a net capacity addition in the last two
years beyond the requirements for silicon area. Presently, we
are estimating the overcapacity in DRAM to be between 20 and 25
percent. The movement of the industry to the more silicon-
efficient 64Mb density will sustain the oversupply throughout
1999. Our silicon demand model shows that with the forecast 60
to 70 percent bit growth rate in 1998 and 1999, about 6 to 9
percent less silicon will be required by the end of 1999 than
is currently consumed.
However, net supply will be trending down as well over the next
18 months as capital spending cuts make their mark on reducing
the rate of supply increase. As time passes, larger linewidth
capacity naturally exits the market, and we are estimating that
this "attrition" will actually outpace new supply. Also we
expect capacity to actively exit the market, meaning fabs
currently in commission will be closed or mothballed. This
could take several forms, including companies exiting the
market, consolidation in the industry, and outright mothballing
of fabs. A recent example is the net loss of the TwinStar fab
in the United States, which was part of the sale of Texas
Instruments' memory business to Micron Technology Inc. This fab
is being mothballed as a result of the consolidation of
capacity, and the equipment is being reallocated and sold. The
situation should set up nicely for a shortage of capacity in
the year 2000.
Semiconductor Demand Has Stalled: One-Year "Holding Pattern"
The Japanese and other Asian economies have slowed tremendously
in recent months, and this has impacted electronic equipment
demand in the areas of consumer electronics and automotive,
with a slight impact to computing. Dataquest recently reduced
the electronic equipment production forecast for 1998 from 6.9
to 4.4 percent growth. This reduction has made a dramatic
impact on semiconductor demand from a unit perspective, and our
most likely forecast for semiconductors in 1998 now calls for a
1 percent growth, essentially flat.
One way to view 1998 is that the semiconductor industry demand
profile has inserted a one-year "holding pattern," and likewise
a one-year delay, in any recovery that was previously forecast.
This has made the secondary downturn we were expecting appear
much more severe, as the industry will be making only "minimum
investments" during the next 12 months.
Since a sustainable recovery in spending and the equipment
market must come from capacity buying, we must have a healthy
chip market in 1998 to have a growing equipment market in 1999.
It does not appear to us that the chip market can mount a
recovery until next year, so we have downgraded our 1999 wafer
fab equipment forecast to be essentially flat from 1998 levels.
The Foundry Industry: Oversupply Just Became "Acute"
Dataquest's analysis of supply and demand in the foundry
industry has been showing, for about a year, supply base plans
are about three months ahead of demand for the leading-edge
0.35-micron technology through 1999. With more suppliers
entering the foundry business, this situation was forecast to
deteriorate into a 15 to 20 percent oversupply generally in
1998 and 1999. We had anticipated that foundry spending plans
would be untouched through 1998, thinking that the supplier
base would react in a way similar to the DRAM suppliers.
However, the foundry suppliers are reacting much faster, since
they are much more driven by profitability in their business
model than the DRAM industry appears to be. Many suppliers have
cut back spending from original plans for 1998 by 20 to 25
percent. Since these adjustments will be in the second half of
the year, the actual spending rates are being cut by about half
from first-half 1998 levels.
Unfortunately, the stall in semiconductor demand has made the
forecast oversupply much more acute, now calculated to be
between 30 and 40 percent. We are expecting foundry investment
to be cut significantly in 1999 relative to overall 1998
spending, perhaps by 20 to 30 percent.
The Wafer Fab Equipment Forecast
The movement to a "minimum investment" pattern in 1998 has
meant a severe cut in spending levels, with a 22 percent drop
in capital spending and a corresponding 17 percent falloff in
wafer fab equipment compared to 1997. The semiconductor demand
stall in 1998 actually pushes the sustained recovery into early
2000, and therefore the growth forecast for 1999 is essentially
flat overall.
We do see a bright spot, albeit moderately in the distance.
While the next six months will be extremely difficult, and 1999
is shaping up to be a flat year, we see the first fundamental
signs aligning to create a spending boom during 2000-2001, with
a shortage in the DRAM market emerging early in 2000.
By Clark J. Fuhs (clark.fuhs@dataquest.com), Ronald Dornseif
(ronald.dornseif@dataquest.com), James Hines
(james.hines@dataquest.com), Klaus-Dieter Rinnen
(klaus.rinnen@dataquest.com), and Takashi Ogawa
(takashi.ogawa@gartner.co.jp)
|
