Greetings from Down-East Maine--it's good to be back home after a 10-day trip to Kansas. Believe it or not, flying from Maine to Kansas takes longer and is more complicated than flying to Tokyo from JFK. We flew in everything from a Beechcraft to a Boeing 737 in a trip that included four hours on the ground in Philly. Smokers don't know what they're missing if they haven't spent time at that no-smoking airport. This was the week of the market forecast. While many of them these days are about as accurate as this fall's Presidential polls, they're still a lot of fun to read about.
SIA forecast is exciting,
but just how real is it?
Granddaddy of all chip market forecasts is the annual outlook of the Semiconductor Industry Association. The trade group, however, is usually too bullish and this year's predictions probably won't be an exception. The SIA sees neither flat nor falling sales over the next three years.
Worldwide chip sales will rise 37% to $205 billion this year for the first time, the SIA predicts. Now that looks very doable, but the group's predictions for the next three years don't. The SIA expects sales next year to rise 22% to $249 billion and then slow to 10% in 2002. Then sales in 2003 will climb 16% to $319 billion, or so the SIA says.
The PC will still be a major growth driver, but other markets will grow even faster, the trade group says. The communications business, for example, will play a major role in driving the chip industry to the SIA's $1 trillion (that's with a "t") goal in the next decade. Now that's optimism!
For the next three years, the Americas will remain the world's largest semiconductor market for the next three years, growing from $64 billion this year to $96 billion in 2003. But Asia Pacific, the No. 2 market, is expected to close the gap by growing to $85 billion in 2003.
Not surprisingly, digital signal processors and flash memory are expected to be the fastest growing chip market segments over the next three years. DSP sales will more than double to $13 billion in 2003, the SIA predicts. Flash memory sales are projected to grow from $10 billion in 2000 to $23 billion in 2003.
DRAMs are projected to grow from $31 billion this year to $52 billion in 2003, while microprocessors will climb from $30 billion to $39 billion during the same period, according to the SIA. Analog chips will climb from $31 billion this year to $47 billion in 2003. And microcontrollers are expected to jump from $19 billion this year to $34 billion in 2003. In sum, an exciting story, but if history serves there certainly will be some real dips ahead.
(See Nov. 1 story.)
Sanders calls observers 'jerks'
and 'idiots' for calling downturn
Jerry Sanders has been around for a while--in fact he's the only CEO that Advanced Micro Devices has ever had. So I always listen when he talks. But this week he called industry observers like me "jerks" and "idiots" for confusing slowing industry growth rates with the peaking of revenues. I don't think so.
"I think the semiconductor industry is about a year-and-a-half into what should be a four-year cycle," Jerry maintains, refuting recent warnings that industry growth and revenues will hit a slump in 2001 or 2002 due to an over buildup of production capacity. But Jerry, it wasn't that long ago when executives such as yourself were telling us the chip industry had matured and there would be no more of the deep business cycles that periodically knocked it for a loop. Yeah sure.
But Jerry wasn't the only chip executive who was confident the industry wouldn't suffer a repeat of the three-year slump that hit semiconductors in the late 1990s. The buildup of new communications infrastructures for broadband Internet access, data and voice convergence, and wireless systems was cited as the main reason for the brighter picture.
"What is really changing the amplitude of cyclicality is the fact that communications is the driver today--this will tend to mute any slowdowns," declares Wilf Corrigan, CEO of LSI Logic. "There will still be cycles," he now acknowledges. "But it is just that the amplitude of the cycles will be muted somewhat." This muted slowdown should happen somewhere around 2002, he guesses.
(See Nov. 2 story.)
Lower order rate again
spooks Kulicke & Soffa
Is Kulicke & Soffa telling us something? Late Thursday evening, the Willow Grove, Penn., production equipment maker once again reduced its revenue projections for the quarter ending Dec. 31 because of additional delays in new orders for wire-bonding equipment.
The world's largest supplier of chip-assembly and packaging tools now expects quarterly revenues to be in the $140-to-$165 million range, down from $180 million in the same period last year. The downgrading most likely will renew Wall Street concerns over the buildup of too much chip capacity--especially in the backend assembly and packaging segment.
K&S had warned investors this past summer that chip-assembly and packaging operations were pushing out orders for equipment due to a lack of processed wafers. That warning caused a sharp drop in stock prices of nearly all semiconductor equipment companies-even suppliers of frontend wafer-processing tools.
K&S says that its customers notified the company this week of additional push-outs of orders and new delays in bookings. "Orders for our wire bonder equipment are very volatile right now and subject to modifications," warns C. Scott Kulicke, CEO. Additional push-outs of previously booked tools were made in addition to deferring new orders, he notes.
(See Nov. 3 story.)
Chip sales still show
no signs of slowing down
No matter what you hear about any slowdown in the general economy or in PC and cell phone growth, the chip business is still going hell bent for election. September was another record month for global chip sales. They hit $18.4 billion, up 45% from the same month a year ago and 1.9% higher than they were in the previous month.
Chip sales were strongest in Japan, jumping 51% to $4.2 billion vs. September 1999, and moving 4.3% higher than August sales, according to the SIA. The Asia Pacific region climbed 47% over year-ago figures to $4.8 billion in September, moving 1.2% higher than the previous month, according to the San Jose trade group.
Sales in the Americas hit $4.7 billion, 46% higher than the year-ago month and 1% higher than the previous month. In Europe, chip sales grew 36% to $3.6 billion vs. September 1999 and $1.8% higher than in the previous month.
(See Nov. 2 story.)
VC funding rises 17% in Q3
for chip and related startups
It isn't surprising that U.S. venture capital funding dropped 7% in the third quarter, what with the collapse of Internet stocks. But what was interesting to semiconductor industry observers was chip startups and electronics companies attracted $2.15 billion in VC money, an increase of 16.9% over the second quarter.
Semiconductor and electronics companies pulled in 8.3% of the venture capital investments in the third quarter, while Internet-related firms received 44.6% and communications companies pulled in 17.7%, according to data from Thomson Financial's Venture Economics unit and the National Venture Capital Association.
During the quarter, VCs invested in 92 semiconductor and electronics companies. These deals averaged $23.4 million.
(See Nov. 3 story.)
Intel spins improving
outlook to Wall Street
Intel worked hard this week to ease fears on Wall Street that its business was slowing down.
The chip giant put its best face forward in a live Webcast to analysts and the press, giving an update on its disappointing third quarter financials and product strategies. "We're bullish," declared CEO Craig Barrett "I expect our communications products will grow by 50% a year."
Intel also blamed an assortment of product delays and problems during the year on an inability to bring costs down for the much-troubled Direct Rambus DRAM memory architecture. "One of our problems was RDRAMs," said EVP Paul Otellini, referring to the next-generation Rambus memory chips. Rambus was unable to bring up its memory-chip technologies "at acceptable price points and volumes," he said.
Intel put out an upbeat story for its current business. "PC sales are brisk," declared Sales VP Sean Maloney. "Outside of the PC market, we have seen strong growth for our servers and communications products."
The company also is beginning to see a slight uptick in the European market, which had been a sore spot for the company; Intel blamed its third-quarter results on weak demand in Europe. "Europe was disappointing in the third quarter, but we are starting well in Europe in the fourth quarter," Maloney noted. He said Intel also is experiencing brisk demand for PC and other products in Asia and especially Japan.
(See Nov. 1 story.)
Is allocation coming
for wafer substrates?
The bare silicon-wafer market could turn chaotic next year because analysts are now projecting a shortage of 200-mm wafers and very expensive price tags for 300-mm substrates.
Already in tight supply, 200-mm wafers could move into allocation next year because of the high demand and the lack of new capacity to turn out the bare substrates for chip fabs, warns Klaus-Dieter Rinnen, Dataquest's chip-making analyst. "We are looking at shortages of products in 2001," he believes.
One reason why is that several silicon substrate suppliers failed to put new capacity in place to meet anticipated demand next year, Rinnen says. This happened because the wafer material suppliers struggled in the late '90s with huge losses due to an oversupply of starting substrates and slower than expected industry growth.
But now wafer demand is up big time as more new fabs are started up. This year 24 new fab projects are in the works, and the total is expected to grow to 30 plants in 2001 and 14 in 2002, according to Dataquest. By 2001, eight-to-ten 300-mm lines will be in place, Dataquest figures. And by 2003, there will be 20-to-25 300-mm fabs.
(See Oct. 31 story.)
Contract chip packaging
could reach 50% by 2010
A major trend to outsourcing in packaging will result in 29% of all integrated circuits being assembled this year by third-party packaging houses, up from 26% in 1999. By 2003, predicts Dataquest, 38% of all chip products will be outsourced to third-party vendors. "By 2005 we'll see that number climb to 40% and by 2010 it could be 50%, predicts Jim Walker, who tracks this industry segment for the market researcher.
This trend will help to more than double the worldwide IC-packaging and test market over the next five years, Walker predicts. This market is projected to grow from $25.5 billion in 1999 to $36 billion in 2000 and to $53 billion by 2003, he says.
The trend to third-party packaging is also being fueled by a shift towards higher-end products such as chip-scale, system-in-package, stacking, and wafer-level packages, Walker notes. Chip-scale packaging is expected to grow at an annual compound rate of 96.5% from 1998 to 2004, the analyst predicts.
(See Oct. 31 story.)
EBN story roils
Rambus stock
A story written about Rambus in CMP Media's Electronic Buyers' News hit the chip designer like a ton of bricks early this week, creating all sorts of excitement among investors watching and investing in this volatile stock. With trading volume nine times higher than normal, the Rambus share price fell by 25% on Tuesday before partially recovering.
A confidential roadmap obtained by EBN showed Intel dropping Direct Rambus DRAM from every computing platform but high-end workstations by mid-2001. According to the document, Intel will phase out the slow-selling Direct RDRAM-enabled 820 chip set in the first quarter of next year, while the yet-to-be-introduced Intel 850 chip set will be dropped in the middle of the third quarter. By that time, Intel's only remaining Rambus chip set will be an enhanced 850 device code-named Tehama-E, which the company is rolling out for workstations and PCs costing more than $2,000.
Other publications maintained there really wasn't anything new in the EBN story. Back in July, they reported, Intel conceded that its next-generation Pentium 4 chip wouldn't solely support RDRAM.
(See Oct. 30 story.)
Foundries won't get hit
by capacity glut until '03
Demand for silicon foundry services will remain strong for the next two years, according to Dataquest, but the market researcher now figures a capacity-glut could hit this booming market with a bang by 2003.
For now though, there's a big shortage of worldwide foundry capacity. Foundries are fully booked with demand robust from both the fabless semiconductor houses and integrated device manufacturers (IDMs), according to James Hines, Dataquest analyst who tracks foundries for the market researcher.
"We see the foundry market tripling by the time we get to 2003," he says. "But I expect the market will move into an overcapacity situation in 2003 and 2004."
The worldwide foundry market will show an annual compound growth rate of 25.4% from 1999 to 2004. The foundry market is projected to grow from $7.5 billion in 1999 to $11.5 billion in 2000 and $16.2 billion in 2001.
The foundry market will "flatten out" then due to a period of "overspending" in fab capacity, he says. The foundry business will grow from $22.3 billion in 2003, to $23.3 billion in 2004, he says. Helping things is the current trend among IDMs to move to a fabless model.
(See Oct. 30 story.)
Sumitomo expands
in 'Silicon Forest'
Japan's Sumitomo Electric Industries is going to build its first U.S. plant to turn out blank compound semiconductor wafers. Construction will begin this month in Hillsboro, Ore.
"The 'Silicon Forest' has excellent infrastructure for semiconductor manufacturing, including highly reliable utilities, a well-educated work force, and proximity to key sources of supply," notes Kenichi Yoshida, Sumitomo Electric managing director. "The need to move quickly was a key factor in making this investment," he adds.
The plant is slated to be in test production by next August, with full production beginning late next year on 6-inch gallium-arsenide wafers. Cost of this first phase of the plant will be $20 million; a planned $70 million second-phase expansion will add large-diameter indium-phosphide wafer production
The Hillsboro plant is being built to serve the exploding market for wireless communications products, such as cellular phone handsets. Sumitomo says it has a number of major GaAs wafer customers in the U.S., including TriQuint Semiconductor, TRW, and Motorola.
(See Nov. 2 story.)
Critical process developed
for 0.07-micron production
Two Dutch companies have come up with a thin dielectric layer that's similar to what will be needed for the 0.07-micron (70-nanometer) process coming later in the decade.
ASM International and Royal Philips Electronics say they have developed what they believe to be the thinnest gate dielectric with a commercially acceptable leakage current.
The gate dielectric was produced with atomic-layer chemical vapor deposition technology and equipment developed by ASM's Microchemistry subsidiary. To produce such ultra thin layers, the two companies had to change from silicon oxide to a layer consisting of three elements: zirconium, aluminum and oxygen.
The thin-film dielectric layer has an effective thickness of 1.1 nm--comparable to a layer of silicon oxide four-or-five atoms thick. The thin dielectric layer is one million-times better than silicon oxide as an insulator in transistors.
The first wafers made with the high-k dielectric film were shipped to International Sematech where they will be employed in a joint development program for advanced CMOS transistor gate structures.
(See Oct. 31 story.)
Cymer claims 81% share
of light-source market
Things are going well for San Diego's Cymer. The excimer light source supplier reported a 67% increase in third-quarter revenues to $98.4 million, up from $58.9 million last year.
Net income jumped to $18.1 million compared to $3.5 million in the quarter last year, with earnings per share hitting 58 cents, easily beating Wall Street's consensus forecast of 52 cents. Cymer's light sources are used in deep-ultraviolet (DUV) lithography.
"During the third quarter, 101 Cymer light sources were installed at chip makers, bringing our year-to-date market share for 2000 to approximately 81%," says CEO Bob Akins.
Accounting for 77% of the Cymer systems shipped in the third quarter was krypton fluoride light sources. Some 128 units were shipped compared to 112 in the second quarter, the company says.
Cymer predicts unit shipments will rise 8% in the fourth quarter over third-quarter levels, with revenues expected to increase 4-to-5% in the fourth quarter over the third.
(See Oct. 30 story.)
AMD takes advantage
of Intel's Rambus woes
Advanced Micro Devices is trying to make hay in the microprocessor market while Intel struggles with its next-generation memory problems. The No. 2 MPU supplier is trying to take advantage of Intel's pushing Direct Rambus memory, which has been plagued with delays in chip sets that can run it and higher costs compared with mainstream synchronous DRAMs.
Now AMD is claiming to be the first chip supplier to offer a PC platform supporting double-data rate (DDR) memory. This includes the AMD-760 chip set and a new Athlon processor with a 266-MHz front-side bus--what it calls the highest grade of DDR memory available.
"DDR memory technology doubles a PC's available memory data transfer rate at comparable costs to today's SDRAM solutions," maintains Richard Heye, AMD vice president. "With DDR memory, a major performance bottleneck is removed, allowing PCs to take advantage of increasing processor frequencies," he says.
PCs featuring the new Athlon processors and DDR memory technology already are available in the U.S. from Micron Electronics and in Europe from NEC Computers. Other major PC makers, including Compaq Computer, are now evaluating AMD's DDR platform for product introductions in early 2001, AMD says.
(See Oct. 30 story.)
Another vote for booming
chip-gear sales to continue
The semiconductor capital equipment market is neither overheating nor slowing down. The next two years still look very good for this business, according to a new forecast from The Information Networks.
This optimistic outlook comes despite the recent downgrading by Wall Street of production equipment stocks, the growing softness in cellular handsets and PCs, and the fall in the book-to-bill ratio for North American-based equipment suppliers from 1.23 in August to 1.16 in September. . The research firm believes this decline was a seasonal adjustment and not an indicator of a real slump.
In fact, Robert Castellano, president of the New Tripoli, Penn., research firm, believes the perception of a slowdown will help to buffer a downturn in semiconductor equipment because device makers will begin to spread out their tool purchases.
"Every equipment company . . . is observing greater bookings and revenue as the semiconductor manufacturers are ramping for increased capacity and transitioning to 300-mm wafers and copper technology in 2002," declares Robert Castellano, president of the New Tripoli, Penn., research firm. He sees new capacity being added at a super-fast rate through 2001. "Thirteen new manufacturing fabs and 16 major expansions are slated for 2000 and 16 new fabs and 15 expansions in 2001," Castellano says. "A total of 26 300-mm fabs will be build over the next few years," he believes.
Semiconductor capital equipment purchases will surge 76% to $44.7 billion in 2000 followed by another 46% increase to $65.2 billion in 2001, according to the new forecast.
(See Nov. 1 story.)
National's Geode processor
seems to be picking up speed
National Semiconductor seems to be getting off to a good start with its Geode reference design for home Internet appliances--better than I expected.
The chip maker won its third design in October as Compaq Computer picked Geode as the architecture for its iPaq IA-2 home Internet appliance. It previously chalked up a design win with Honeywell's WebPad and struck an alliance with eMachines to design its MSN Companion around the Geode architecture.
The potential market here seems huge, but it undoubtedly will be a cutthroat business when it comes to pricing. "By 2005, more than 70% of the world's Internet users will use information appliances to access the Internet," maintains Egil Juliussen, president of eTForecasts. And he figures National is positioned well for that market. We'll see--historically Egil has always been a big market bull.
International Data (IDC) is one of the very optimistic. It figures the worldwide market for home Internet appliances will zoom to more than 10 million units in 2002, up from just 1.6 million units in 1999. Geode uses an x86 architecture to integrate on-chip graphics, audio, memory control and a PCI interface. This amounts to more than 80% of the silicon that will make up Compaq's iPaq IA-2.
(See Oct. 31 story.)
Dataquest looks at lowering
its wireless handset forecast
Now even Dataquest is wavering on this year's wireless handset market.
The market researcher in recent weeks had resisted cutting its bullish forecast even as others were doing just that. Earlier this year, Dataquest predicted the global unit market would grow from 290 million in 1999 to 434 million this year and 567 million by 2001.
Analyst Stan Bruederle now says that it looks like the 2000 forecast "will be a little lower." But he was still sticking with his highly optimistic 2001 forecast.
(See Oct. 31 story.)
Hyundai's second thoughts
over de-emphasizing DRAMs
Hyundai Electronics' semiconductor group is busily trying to reinvent itself now as a hybrid DRAM vendor and logic foundry supplier in an effort to escape the cyclical nature of the memory market. But the real question now is: Does the company really want to move away from DRAMs?
Hyundai executives now believe they see the see-saw DRAM cycle leveling out and that consolidation is yielding a better balance between DRAM supply and demand. "We think that with fewer people in the market, and with fabs becoming more and more expensive, the overall market will be stabilized," says Sang Park, COO of the group. "This could be the end of the semiconductor cycle."
DRAM vendors have typically made the bulk of their profits during periods of shortage, when they could charge higher prices for the memory chips. Then they would use those profits to increasing their production capacity. But, the proliferation of fabs soon creates an oversupply of DRAM chips and prices plummet.
Now, Park says, the DRAM market has consolidated and just four vendors account for three-quarters of all DRAM chips sold. That's down from eight vendors in 1996. Over the same period, the cost of a fab has jumped from $1.5 billion to an estimated $2.5 billion now for a state-of-the-art 300-mm wafer fab.
With fewer players, Park expects the long-term result will be a balance between supply and demand. "We believe this will be a very stable market in the years to come," Park predicts.
Some DRAM experts don't share Hyundai's position that the DRAM cycle may soon be over. They "seem to think that with fewer players, there's more visibility into capital spending, and the higher costs of fabs will keep out any new players," says Dataquest memory analyst Jim Handy. "We just don't subscribe to that theory."
Handy still expects to see memory vendors continuing to chase profits during boom times and then overbuild to serve the demand surge, swinging the pendulum back toward overcapacity. He also disputes Park's assertion that the DRAM market is more consolidated than ever before. The downturns of the mid-80s and '90s demonstrated that a smaller field does not necessarily lead to smaller swings in supply and demand. "This is not rocket science," declares Handy. "The semiconductor cycle will continue."
(See Nov. 1 story.)
New Motorola product
brings back memories
It seems like Motorola Semiconductor has been interested in "tonnage" ever since it started making semiconductors products back in the '50s. I can remember visiting Phoenix in the early 60s and seeing their diode production line spew out parts by the millions for automobile alternators. Now the company is about to roll out another semiconductor part that could sell in huge volumes.
This "electronic tag" technology, which will be introduced in Chicago next week, could add automated identification to every day, low-cost, disposable products ranging from lipstick boxes to breakfast cereal containers. The tags are expected to enter the market late this year following the launch of the "smart' cardboard box partnership between Motorola and International Paper.
"There are going to be trillions of tags like these on all kinds of consumer products, and they'll tell us exactly where those products are in the supply chain at all times," comments Larry Kellam, director of worldwide-supply-chain innovation for Procter & Gamble. "This is going to be huge."
The technology, which uses radio-frequency transmission techniques, dramatically cuts the cost of automated product identification. Microchips now used in R-F identification cost from 50 cents to more than a dollar each. But Motorola's new chips will cost only 10-to-30 cents a piece and ultimately might drop to a few pennies each, engineers at the company believe.
Moto's new technology departs dramatically from other electronic tagging systems. Conventional R-F ID systems use inductive methods, while Motorola's system employs a capacitive technique called BiStatix. It uses an electric field to couple the chip, or tag, capacitively to a reading device. The electric field becomes the tag's source of power and its master clock, allowing the tag to modulate its data regularly and transmit the data to the reader's receiver circuit.
BiStatix is lower in cost because it eliminates the need for the wire coils or batteries used in inductive systems. Instead of a wire coil design, the BiStatix configuration mounts a tiny silicon chip atop printed-carbon-ink electrodes. The electrodes, which use conductive ink, act as an antenna.
(See Nov. 3 story.)
If you have any comments or questions, don't hesitate to E-mail us at bhenkel@aol.com. Have a great weekend!
(Click here for the previous Semiconductor Alert!.)
