Commentary & analysis of week's chip news
Greetings from Down-East Maine. We had our traditional, old-time, Fourth of July this week and the parades, theatre, games, et al were great fun. And because of coastal fog in the mornings, the max temperatures have been running at only 60F to 70F. It doesn't get any better than this. Have a great weekend and leave your worries about business and stock prices in the office when you leave tonight.
gives up for this year . . .
ASM Lithography has given up for this year. The Dutch lithography systems supplier now believes that a fourth-quarter recovery from the chip industry's downturn now appears unlikely.
"Recent comments from our customers about their Q2 and Q3 outlook lead us to believe that a market recovery in Q4 is more and more unlikely," says CEO Doug Dunn. "As a consequence, our customers are hesitant to commit to capital expenditures not necessary for their current business. This means," he says, "that this year's demand for our mainstream products could be lower than previously anticipated."
As a result, the litho giant is mounting a major cost cutting program across all of its divisions, including reductions in staffing levels, capital expenditures, and discretionary spending.
The company now expects lithography tool deliveries by ASML in Veldhoven to fall below 230 systems in 2001. Last year, this operation shipped 368 exposure tools, up from 217 systems in 1999.
As far as ASML USA is concerned, the new unit (formerly Silicon Valley Group) is now expected to report more than a 40% decline in revenues for the second quarter. Order cancellations and delays also resulted in a "sudden and sharp decline" in order backlogs and shipments in the second quarter. ASML will release its second quarter financial results on July 18.
(See July 5 story.)
. . . But the SIA is its
usual, optimistic self
But the Semiconductor Industry Association still believes that market conditions will begin to improve in the third quarter, even though chip sales in May were running at their lowest monthly level in nearly two years.
"Sales for May reflect the current inventory overhang that the semiconductor industry has been experiencing since November," points out George Scalise, SIA president. "We continue to believe that the industry will begin to see the final phases of the inventory correction late in the third quarter with a broad-based sequential recovery commencing in the fourth quarter." What an optimist!
But so far, May has been the worst month in the industry's current downturn in terms of a percentage decline in sales on a month-to-month basis. Excess inventories and a growing global economic weakness pushed May chip sales down 7.3% from April's $13.7 billion to $12.7 billion.
Global chip revenues have now dropped seven months in a row after sales peaked in October 2000 at $18.7 billion. May sales were off more than 20% from year-ago shipments.
On a regional basis, sales in the Americas fell the most, dropping 10.9% from April's total. Compared with year-ago figures, May chip sales were off a whopping 32%. Europe logged in with a 7.8% decline in May sales from April. Japan sales in May tumbled 5.9% from April, while Asia Pacific showed the best regional results with just a 4.5% decline in May.
(See July 2 story.)
Chip makers still
turning other cheek
Back in the early '70s, I remember chasing a story where a well-known calculator company would order chips at the highest discounts possible--in quantities of a million, for example. And every time the price would come down on chips, this OEM would cancel its original order to buy at the new low rate. No matter that they had purchased only a small percentage of their original commitment. Surprisingly to me, the OEM could always find a patsy for this kind of cheating.
Call me naive, but it still surprises and angers me when a manufacturer signs a deal to buy components in quantity at a higher discount, only to cancel it when business slows. You'd never see a Japanese equipment maker do that--they would continue to honor the original agreement to the letter.
No one wants to harm relations with a valued customer, but I was pleased to learn that Advanced Micro Devices had sued a major customer three months ago for allegedly refusing to honor commitments made in a two-year purchasing agreement for flash memories. The customer was Alcatel Business Systems, which was using AMD flash memory in its mobile phones.
The Sunnyvale chip vendor caused a stir in the industry by suing after a dispute had broken out between the two companies when flash prices were plunging and the cellphone market was going off as cliff. AMD said it had taken the unusual step of suing Alcatel as a last recourse after several months of negotiations.
But this week the lawsuit was dropped when the companies reached "an agreement." Neither AMD nor Alcatel was talking details, of course, though AMD vice president Walid Maghribi claims the settlement "acknowledges the validity and mutually-beneficial nature of our supply agreements." Yeah, sure.
(See July 2 story.)
Folding a document won't
hurt embedded RF ID chip
Here's my most interesting technology of the week. Hitachi has come up with a tiny chip that it believes will take over the market for conventional RF ID chips because of its smaller size. The chip measures only 0.4-millimeter on a side--about 1/10th the size of today's RF ID chip, which runs only several millimeters on a side.
The Hitachi chip, which stores security and identification information for items in which it is embedded, was originally developed to protect against counterfeit bills or documents. If the chip is embedded in a bill or security document, a chip reader can immediately determine if the item is authentic. The chip is so small that it doesn't damage material in which it is inserted--even if that material is folded, claims CEO Ryo Imura of Mew Solutions, a new venture company formed this week to market the new IC.
The 60-micron thick CMOS chip integrates RF wireless communications circuitry and 128 bits of ROM, and transmits data stored on its ROM, including encrypted data, over the 2.45-gigahertz band. The chip can transmit over a distance of 30 centimeters by connecting it to a metal antenna. Later on, Hitachi will offer a version of the chip with an attached antenna.
"We've got requests to add functions such as one-time writing or rewriteable capability to the chip," Imura says, "that we'll consider for the next-generation products." But for higher security, he adds, confining the chip to read-only memory prevents any falsification. Hitachi expects the chip to open up new applications, possibly by linking information on the Internet with the chip's ID number.
Samples of the Mew chip will available this autumn and the company will start marketing it next spring. The new company has big ideas: It expects to have annual sales of $145 million by 2005.
(See July 2 story.)
Downturn hits bottom (?) for
for IC packagers, assemblers
Leading providers of IC packaging and assembling now say the downturn finally has reached the bottom for them. But they don't seem to have much farther to fall--their current capacity utilization is running as low as 20%.
"It's pretty rough out there," acknowledges John Boruch, who runs Amkor Technology, world's largest IC packaging and test house. However, he adds, "we are at the bottom part of the cycle. But the question is, how long it will stay at the bottom?"
"Based on our customers' input, the rest of this year will be slow. Everybody in the IC-packaging and assembly business will have a down year," predicts Tien Wu, who heads U.S. and European operations for Taiwan's IC packaging and test giant, Advanced Semiconductor Engineering (ASE).
There could be some improvement in the second half of 2001, Wu says, but the business will recover to a larger degree next year. "We believe there will be a noticeable recovery in 2002, with a strong year projected for 2003," he predicts.
As for Amkor, the market is a moving target. Earlier, the company projected that its sales for the second quarter would be 20% below the first three months total of $481 million. Now the Pennsylvania company predicts that second quarter sales will fall to $350 million--a 30% sequential fall. Its utilization rates for its IC packaging and its silicon foundry businesses each have sunk to about 40%.
ASE, world's second largest provider of IC packaging and test services, reports that its revenues will decline as much as 25% in the second quarter from the $331 million reported in the first quarter.
(See July 2 story.)
Researchers try to come up
with quantum IC-making process
Now here's a challenge! A group of university researchers led by Ohio State University is trying to come up with a quantum chip-making process that is repeatable, reliable, and achieves good yields with a room temperature operation. The four-year-effort, being funded by a $1.6 million grant from the National Science Foundation, hopes to craft a manufacturing technology that will work with any of the quantum-computing architectures now being proposed.
"A lot of people have successfully demonstrated that quantum-dot nanoswitches of various architectures can work--such as resonant tunneling diodes, single-electron transistors, and quantum cellular automata--but there is no readily available chip-manufacturing process yet," notes project leader Paul R. Berger, associate professor at Ohio State.
The reason is that most quantum devices won't work at all unless they're very, very small, at least 10 times smaller than those proposed for future silicon chip technologies. These devices work by "tunneling"--instantaneously passing individual electronics across an insulator without taking any time to physically pass through it. Each nanosize domain also can store both a 1 and 0 simultaneously. So far, however, all quantum-dot projects have had dismal yields.
(See July 3 story.)
Hynix floats doubtful strategy
to turn around DRAM market
A South Korean DRAM vendor seems ready to employ an old Japan Inc. strategy in an attempt to stop the chip industry's bloodiest price war. But it seems doomed to failure.
Hynix Semiconductor, formerly Hyundai Electronics Industries, says it may cut its DRAM production by 20% as early as next week. "Beyond a certain threshold of pain, it doesn't make sense to keep on producing," says Farhad Tabrizi, vice president heading up memory marketing.
Hynix would cut output of its oldest DRAM products, those with geometries above 0.18 micron, which account for 20% of its total DRAM output. Such a move could cut world production by 4%, Tabrizi says. That could trigger a change in spot pricing and spark a price recovery, he claims. "There is a 5-to-8% oversupply at the moment and if we cut a week or two of production, it could change the psychology, Tabrizi claims.
There's no doubt the DRAM market is suffering something awful. Gartner Dataquest predicts the DRAM market will suffer its worst year ever, shrinking 55.5% from last year. Prices have plummeted 80% over the past year; spot prices are down to $2 and contract prices below $3 for a 128-megabit DRAM. Global DRAM revenues could fall to $18 billion this year, an $11 billion fall from 2000, says Semico Research.
Micron, which is now running neck-and-neck with Hynix and Samsung for the lead in the DRAM market, doesn't plan to reduce its output. "Gaining more market share is one element of what we're doing, but it's more about being a low-cost provider," says a Micron spokesman. "To reduce capacity means to run less efficiently."
Samsung suggests that Hynix would commit a major strategic blunder if it cut back production. "We have a different plan: We'd like to maximize production to maximize our market share," says the firm's DRAM marketing director, Il Ung Kim. "The beauty of the DRAM market is that . . . only the fittest survive."
The move by Hynix might have an impact if another major DRAM maker followed suit, admits IDC market researcher Soo Kyoum Kim, but he doesn't expect to see that happen. "Four percent is not enough to move the market. There is still channel inventory and oversupply of about 12-to-13%," he adds.
(See July 3 story.)
AMD joints crowd,
cuts sales forecast
Why should one of Wall Street's favorites be immune from the current semiconductor downturn? Yep, it was Advanced Micro Devices' turn this week to bite the bullet and lower its second quarter forecast.
AMD changed its forecast from a 10% sequential decline to a 17% drop to $985 million. Net will take a bigger hit as the company now sees quarterly profits of only 3-to-5 cents a share, down significantly from the 27 cents the street had estimated it would earn.
The company blamed the bad news on a collapse in the flash memory business and a drop in the average selling prices for its x86-based microprocessors. Its biggest competitor Intel recently started cutting prices on its flagship Pentium 4 processor lines by as much as 50%--a move designed to gain market share and fend off AMD.
Even so, AMD claims to have achieved record unit sales of both its Athlon and Duron processors in the second quarter. The chip maker will report its second quarter results next Thursday.
(See July 5 story.)
should start worrying
It's time for chip makers competing with Intel to start worrying. In case anyone had any doubts, it has become crystal clear that the chip giant doesn't plan to get caught short on capacity during the next upturn.
Yet another 300-mm wafer fab has popped up in the Intel game plan. Intel reportedly is planning to set up a new development fab in Silicon Valley, according to Prudential Securities. If true, this fab would become Intel's seventh 300-mm wafer plant being planned or built.
Intel has decided to convert its 200-mm D2 Santa Clara fab to a 300-mm fab in 2002, says the New York brokerage firm. When Intel was asked about the report, a spokesman would only say that "we have not made any announcements" about D2. Intel confirms it has completed the fourth module in the D2 plant to build chips based on 0.18-micron design rules. According to the Prudential report, D2 is also ramping Pentium 4 processors at 0.13-micron.
It's clear now that a big chunk of Intel's record $7.5 billion capital spending budget this year was earmarked for 300-mm processing lines and new copper-chip technologies.
(See July 5 story.)
UMC may mothball
one or more fabs
The downturn is really beginning to hurt the big Taiwan silicon foundries. Rumors have been building over the past few weeks in Hsinchu Science Park about United Microelectronics (UMC) mothballing at least one fabrication facility because of low utilization rates.
This week, the No. 2 foundry acknowledged to its employees that it might make sense to close temporarily one or more of its wafer fabs if the industry downturn gets any worse. But there are no plans to do so now, UMC says, and insists it would not lay off any workers.
UMC is moving forward, however, with its capacity expansion plans. A 300-mm wafer fab in the Tainan Science Park is slowly ramping up and is expected to be producing as many as 7,000 wafers a month by the end of the year. Because of such expansion plans, it doesn't make sense to let any employees go, UMC says.
(See July 3 story.)
Foreign chip gear makers pull
away in customer satisfaction
In some ways, semiconductor production gear is becoming just like automobiles--customers are becoming increasingly more satisfied with Europeans and Japanese vendors than they are with the Americans.
For the fourth year in a row, European-based suppliers of semiconductor production equipment made strong gains in customer satisfaction as U.S. tool vendors continued to slip in the ratings, according to a survey of chip makers by VLSI Research.
European suppliers wound up with 20 of the 50 top-10 rankings in five tool categories, a 50% increase from the 13 positions they earned last year. American-based companies held 19 positions, down from 24 in 2000. The Japanese suppliers held 10 of the top-rated positions, down from 12 last year.
Singled out most for lower quality this year were software and process support. Large suppliers of wafer-processing tools scored better than their smaller counterparts. Topping the list of companies with the highest ranking was Universal Instruments, which supplies assembly systems for chip packaging as well as printed circuit board products.
Varian Semiconductor Equipment topped the list for large suppliers of wafer-processing equipment, while Ultratech Stepper led the list of small suppliers of wafer-processing equipment. Karl Suss was No. 1 in test and handling equipment, and topping suppliers of process diagnostic equipment was Agilent. In assembly and specialty equipment, Universal Instruments led the way.
(See July 3 story.)
Still no high-k insulator
to replace silicon dioxide
Chip makers will soon have to make a decision on what material to use as a high-k gate insulator to replace silicon dioxide. But despite the importance of finding the right material, no obvious candidate has emerged, according to experts gathered at the VLSI Technology Symposium.
Little real progress has been made in the search and that will cause the traditional performance gains being made from CMOS scaling to trail off later in this decade, researchers predict. The industry should have one material that's proven at the research stage by now, but it doesn't. "This is why the industry is almost in a panic," comments one researcher.
Gate oxide is becoming the key limiter to performance gains. The ultrathin SiO2 permits electron tunneling through the insulator, causing current leakage that ruins power-consumption budgets. The industry goal is to find a new material that's thick physically so as to prevent electron tunneling but one that has an oxide thickness that allows performance scaling. That needs to be done within the next one or two years, experts say.
Intel will switch to a true high-k material for the 45-nm technology node, which start early manufacturing in 2007, predicts Robert Chau, director of transistor research for the big chip maker. He doesn't identify Intel's current gate insulator, but sources says that it already is using a stacked-gate oxide, with a nitride layer on top of an oxide at the interface to the channel.
Stacked oxides now being used by Intel, AMD, and IBM work well with available deposition tools and may allow microprocessor vendors to get through the 70-nm production node without having to switch to a new material. "The question is, can we meet the time scale set by the road map?" asks Professor John Hauser of North Carolina State University. He knows of only one company that has decided on a high-k gate material and is beginning process integration with it. Others have narrowed their search to a few materials, including oxides and silicates of aluminum, hafnium, and zirconium.
These corporate efforts are being bolstered by programs underway at consortia such as IMEC in Belgium, Sematech in Austin, and a Japanese development project. "Without global cooperation, this might be the show stopper," says Mark Heyns, who runs IMEC's high-k research effort.
(See July 2 story.)
A much cheaper way to ensure
that photomask quality is OK
Because inspecting for quality is so difficult with attenuated phase-shift masks, chip makers and photomask shops are going overboard to make sure the photomasks are good.
"The mask industry has been operating on the basis of inspecting everything and repairing everything whether or not it has a measurable impact on yields," says J. Tracy Weed, senior marketing director at Numerical Technologies. "But as the complexity increases, that is no longer economical," he adds. "The practice of repairing everything impacts costs significantly," he says. With attenuated phase-shift masks, the only way to make sure mask quality is okay is to run test wafers.
To get around this problem, the San Jose company came up with a new software module that lets IC makers and photomask shops to quickly inspect and determine the quality of attenuated phase-shift masks before using the reticles to print circuit features on silicon substrates. These masks are used to improve the edge contrast features when chip makers are attempting to push the critical circuit dimensions below the wavelength of light in an exposure tool--known as subwavelength lithography.
This new tool will not only enhance yields on next-generation chips produced with subwavelength techniques, but should enable large suppliers to save millions of dollars in time and materials annually by avoiding unnecessary repairs of the expensive phase-shift masks. KLA-Tencor is the first inspection tools supplier to offer the new software with its systems.
(See July 2 story.)
Qualcomm claims 3G lead--
but will market ever take off?
If it ever develops like the current hype is predicting, the third-generation (3G) wireless market will be incredibly important for the chip industry. So far, though, this technology is slow in coming, slow to be adopted, and the potential market still remains to be proven. In other words, will the consumer buy it?
Right now, it's a war of words. In spite of the high-profile, 3G deployments of Japan's NTT DoCoMo, San Diego's Qualcomm still claims it is leading the 3G wireless technology race. The current word is that DoCoMo will finally begin offering 3G services in Japan in October after experiencing delays this spring, which reportedly were due to software glitches with its 3G network.
The Japanese network is based on a standard called wideband code-division multiple access (W-CDMA), which reportedly uses baseband ICs from Texas Instruments.
Qualcomm's 3G offering, called cdma2000, has a major lead over W-CDMA in the market, the company claims. South Korea's SK Telecom already is running the world's first 3G network, which is based on Qualcomm's 3G technology--cdma2000 1x. In May, two other Korean carriers--KT Freetel and LG Telecom--also began offering 3G services in that nation based on the cdma2000 1x technology.
Korea is one year ahead of Japan in 3G, claims Qualcomm's CDMA Technologies vice president, Johan Lodenius. He also says Qualcomm is gearing up for cdma2000 1x in Japan, U.S., and South America.
3G may finally become a reality in the U.S., but domestic carriers are expected to deploy both cdma2000 and W-CDMA. U.S. carriers such as Verizon and Sprint plan to deploy their initial 3G services later this year using cdma2000, according to Qualcomm. But in the U.S., domestic carriers are expected to deploy both cdma2000 and W-CDMA. The European market is talking about deploying W-CDMA by 2004-05
(See July 5 story.)
If you have any comments, criticisms, or questions, don't hesitate to E-mail us at email@example.com. Have a great weekend!
(Click here for last week's Semiconductor Alert!.)