Commentary & analysis of week's chip news
Greetings from Down-East Maine, where you know spring has arrived when the sea breezes keep our coast 10 degrees cooler than it is a few miles inland. That means we are enjoying temperatures in the high 50s to low 60s. It also means we have started all kinds of outside projects. This week, for example, we have a team of woodsmen cutting up trees either knocked down by 70MPH winter windstorms or dying of old age.
Next week we'll bring in a portable saw mill to saw up these 100-year-old logs into lumber and store them in our boathouse to age. We've also opened up our houses and will begin the many odd jobs that 200-year-old homes require. So it's a great time to be outside, which is where you'll find me when I'm not writing columns on my computer.
Semicon worries about
drop in IC unit sales . . .
Guess what topic A was in Munich this week at the Semicon Europa conference? Where the chip market was going, of course. But let's face it--no one really knows what's going to happen. The only thing industry managers could agree on was that they had just gone through a terrible first quarter. After IC shipments peaked in October at 7.7 million units, they fell off sharply to 6.5 million chips in February.
Dataquest once again revised its forecasts downward for both the semiconductor and chip-making equipment markets. During a session at the Semicon Europa trade show here this week, analyst Klaus-Dieter Rinnen, director of semiconductor manufacturing analysis at Dataquest, said the research firm
It is now expecting chip revenues this year to fall by 16%, says Dieter Rinnen, director of chip making analysis at the research firm. And it gets worse: If overall business conditions continue to erode, he warns, the decline could be more like 23% in 2001. The San Jose researcher also cut its forecast for 2001 chip equipment revenues from a dip of 6% to a fall of 24%.
What's scaring analysts now is the growing possibility that even unit sales will fall this year at a time when industry capacity is so under utilized. "There has only been one year in history when IC unit volumes declined that was in 1985," points out Elizabeth Schumann, research director of the SEMI trade group. And now it's beginning to look like 2001 will be the second year in history that IC units volumes decline, she predicts.
(See April 25 story.)
. . . But some optimists see
March as bottom of slump
But not every market researcher and analyst believes that unit sales of chips would drop for the first time in 16 years. Dan Hutcheson of VLSI Research maintains unit shipments are not dropping, and he even believes that March was the bottom of the current slump.
"Pricing on ICs has held firm since September," he argues, and it's now strengthening in some segments," he tells SBN. "This is the light at the end of the tunnel--but is it a train?" he asks. Hutcheson was only half joking about the "train." While he is convinced that the industry has reached the bottom of the downturn, he believes it will take longer for the chip and equipment markets to fully recover.
"We certainly won't bounce off the bottom like we normally do," Hutcheson comments. VLSI Research predicts 2001 chip sales will drop by nearly 20% and chip industry capital spending will fall by 19%.
One reason why he calls March the bottom is that new data shows chip makers acted faster than they normally do in shutting down production lines in reaction to a business slowdown. "In 1996, when the business started to decline, chip companies didn't respond immediately and it wasn't until 1998 that they began to close fabs," notes Hutcheson. "That delayed reaction is happening in this downturn."
He now expects the chip industry's capacity utilization to edge up slightly in April after hitting a low point of 80.8% in March. Fab utilization peaked at 98% last August. Hutcheson predicts fab capacity utilization will slowly move up to around 91% by December.
But other industry analysts put the fab utilization figure in the 70% range for the second quarter and they say it isn't clear if that percentage will move up much higher before the end of the summer. Hardest hit are the big silicon foundries, which were hit by integrated device manufacturers pulling back outsourced production to their own plants. Some Asian foundries already have fallen "well below 40%," Hutcheson points out.
(See April 25 story.)
SVG delays shipment of
of latest 193-nm scanner
The chip-making equipment market may have taken a dive, but it's still hard to get an on-time delivery of the latest scanner models.
Silicon Valley Group, for example, has confirmed it will not be on time with the first delivery of its latest tools based on 193-nm, argon-fluoride (ArF) technology. The three-to-four month delay will impact Intel and several other big chip makers, say analysts.
CEO Papken S. Der Torrossian says his company has pushed out deliveries of its Micrascan 193 high-numerical aperture (HNA) scanner from April to July. "It's late," he acknowledges. "We would have liked to have shipped them in April. But the good news is that our competitors are late as well."
Last year SVG got a $100 million order for this Micrascan 193 tool from "a leading microprocessor manufacturer," which is believed to be Intel. Cypress and IBM are two other chip makers that use SVG lithography systems.
The SVG Micrascan 193 HNA system is based on a new platform that improves the throughput of its older 193-nm tools, and provides a resolution of better than 130-nm. It will also enable chip makers to develop devices at the 100-nm node, SVG claims.
But SVG and other tool suppliers are having problems with their 193-nm models, due partly to a shortage of calcium-fluoride for making the lens.
ASM Lithography, Canon, and Nikon, as well as SVG, have been shipping 193-nm tools in small quantities for R&D purposes since 1999. But they are still unable to deliver these high-priced systems in large volumes to their chip customers--most of which are now attempting to step up their use of 193-nm tool for production purposes, sources say.
The existing krypton-fluoride, 248-nm lithography tools are quickly running out of steam. While some experts believe that 248-nm tools will process devices down to 0.13-micron or so, the newer 193-nm scanners are geared to more easily deliver chips at 0.10-micron and below node.
(See April 25 story.)
ASML-SVG merger was
not a done deal after all
It looks like I might have been wrong after all in predicting ASM Lithography's deal to buy San Jose's Silicon Valley Group would be okayed this week. The pending $1.6 billion acquisition has been stalled since early March by U.S. government reviews.
Management at both companies are growing frustrated over the lack of a decision by the government. They believed the deal had cleared regulatory hurdles back in February, but the U.S. Committee on Foreign Investment initiated a 45-day review of the transaction because of concerns over defense-related technologies. But the committee was unable to agree by the end of the review on Monday, so the hot potato was passed on to President George W. Bush, who now has 15 days to decide if the merger can proceed as planned.
U.S. defense concerns regarding the merger is the ability of SVG's Tinsley Laboratories unit to produce highly accurate lenses, which have been used in U.S. spy satellites during the 1990s. ASML says it is willing to consider selling Tinsley if it would help ease U.S. security concerns over its planned purchase. This "has been discussed during the last 45-day review," says Rob van Vliet, ASML investor relations manager. "All types of options have been reviewed, including keeping Tinsley inside after the acquisition but providing certain guarantees to the U.S. government for protecting the lens technology," he adds.
(See April 26 story.)
They didn't work in 1998, but
'bridge tools' are big time now
Even though it didn't work three years ago, chip makers are reaching for an old tool to help them fight the high costs of 300-mm wafer production tools.
These "bridge tool" platforms, which were a controversial approach in the industry's failed attempt to move to 300-mm wafers in 1998, now appear to be taking off in a big way as the migration to larger-diameter wafers gets rolling. Chip makers are embracing the concept of using these platforms to do process development and tool-set integration on 200-mm wafers first, then upgrade the tools so they can produce the 300-mm substrates. This approach didn't take the first time because it added costs to already expensive 300-mm tools.
But this time around, uncertainty in the marketplace plus the need to speed up development of next-generation process technologies has given bridge tools a big boost. In fact, bridge tool shipments are expected to surpass dedicated 300-mm systems next year in number of fab tools shipped, according to a new forecast from VLSI Research.
Bridge tools will grow from just 11.2% of this year's total equipment shipments to 52.5% in 2005, predicts the market research firm. In contrast, it believes that 300-mm dedicated systems will go from 11.5% of tool shipments this year to just 5% in 2005.
"The bridge tool is going to account for the vast majority of 300-mm capable systems during the next fives years or more," predicts Dan Hutcheson, VLSI Research president. "Back in 1998, the concept touched off a huge controversy, but it did not make sense for equipment suppliers to be developing two different platforms for one generation of process technology," he says.
Some early opponents to bridge tools have turned into backers of the flexible-system concept, which make it possible to convert a platform from 200- to 300-mm wafers within hours. "At one time, I felt it was fairly impractical, but now I feel it is absolutely the right way to go," says Semitool CEO, Raymon F. Thompson. "It is crucial to where we are headed," he says. "Our 300-mm tools will be 200-mm compatible because it is now necessary."
(See April 26 story.)
IBM says nanotubes may
keep Moore's Law going
It seems like a wild coincidence to me. At a moment in time when the creator of Moore's Law is retiring from Intel, and researchers are questioning whether a new technology will come along to keep Moore's Law going (You know, every 18 months the computing power of a chip doubles), we may have a candidate.
Gordon Moore himself said four years ago that Moore's Law is coming into direct conflict with the law of nature, that chip making was going to start butting up against the finite size of atomic particles. "Some time in the next several years we get to some finite limits," Moore said. Physical limitations could be reached by 2017, according to one study.
But on Friday, IBM researchers reported they have come up with a new way to build the world's first array of transistors out of carbon nanotubes, which are tiny cylinders of carbon atoms measuring 10 atoms across. These structures are about 500 times smaller than today's silicon transistors. The IBM researchers believe their efforts could help clear the way for production of ICs once solid-state physics runs out of gas in silicon during the next 10 to 20 years.
IBM came out strongly for the new technology. "Our studies prove that carbon nanotubes can compete with silicon in terms of performance, and since they may allow transistors to be made much smaller, they are promising candidates for a future nanoelectronic technology," declares project head Phaedon Avouris.
Researchers already have been able to fabricate field-effect transistors (FETs) from carbon nanotubes with any variable band-gap desired. Data shows that if carbon nanotubes were scaled up to the size of today's silicon transistors, IBM figures they would perform at the same level as current semiconductors.
(See April 27 story.)
Seems to me I've
heard this song before
So--are you ready for your weekly infusion of hype--oops, I mean optimism? I'm the first to admit we can talk ourselves into a recession, but this is ridiculous. I question whether we can talk ourselves into a boom.
Cahners In-Stat is at it again with the Bluetooth. This week the market researcher declares that the emerging wireless connection format will survive the economic slowdown, product delays, and a wave of negative reports to grow explosively over the next five years. You won't believe this, but the Scottsdale firm predicts that Bluetooth will shoot up to 955 million units in 2005--that's nearly a billion from almost nothing now!
That means the five-year compound annual growth rate for Bluetooth will run at an incredible 360% between now and 2005. It also means that the semiconductor potential from Bluetooth will reach $4.4 billion in 2005, In-State estimates.
Last year was "a year of trials and tribulations for Bluetooth," acknowledges Joyce Putscher, In-Stat director. "However," she hastens to add, "positive signs are here now as more silicon is going into production, more products are closing in on production schedules and are coming to market very soon."
The first "hot spot" projects already have appeared in hotels, shopping malls, golf courses, airports, she says, and more are expected to come to fruition by the end of the year.
(See April 25 story.)
3G wireless another hyped
market that isn't making it
Here's another highly promoted future market that keeps moving further and further out.
NTT Docomo, which is deploying the world's first 3G (third-generation) wireless network in Japan, is delaying that service from May to October, a move that some say casts more doubts about the technology, according to reports this week from Reuters.
And a new report from the Cahners In-Stat Group declares that NTT Docomo will not have "nationwide coverage" for its 3G network until late 2002.
In fact, the new wireless service is behind schedule in other regions. Europe has made several commitments to deploy 3G, but there will not be a "moderate subscriber uptake" until mid-2003, In-Stat says. And it's unclear now when or even if the U.S. will deploy 3G.
One reason for all the delays, analysts say, is that 3G may not be ready for prime time. High costs to buy spectrum and delays in handset introductions continue to push out the deployments of 3G services, In-Stat says.
"3G is not for those with weak hearts or shallow pockets," declares In-State analyst Ray Jodoin. "The financial impact of mega-spending for spectrum followed by seemingly unreal financing arrangements has taken its toll."
The short-term prospects for the technology that will make possible video calls, high-speed Internet access, multimedia services, and other features are bleak, he concludes.
But ever the optimist, In-Stat predicts the 3G market may still take off in a big way in a few years. 3G services make up only 4.7% of worldwide wireless market this year, In-Stat estimates, but if the market develops as expected, that share could jump to nearly 50% in 2005. That's another incredible growth rate.
(See April 24 story.)
Agere profits evaporate
in quarter ended Mar. 31
It looked though the child was imitating its parent this week when Agere Systems--which was spun out of Lucent Technologies in March--had a lousy quarter with net profits diving a whopping 94% in the quarter ended March 31 vs. a year ago. Not only that, the communications IC and optical module maker predicts that it will lose money in the current quarter.
Revenue declined too, but not at the same rate. It dropped 13% from the previous quarter to $1.2 billion.
As a result, Agere plans to slash 2,000 jobs out of its headcount of 18,500 by the end of June, with most of the reduction taking place at its plants in Allentown, Breinigsville, and Reading, Penn., Irwindale, Calif., and Orlando, Fla.
But revenues in its optoelectronics business climbed 42% over the same quarter last year to $379 million. This was driven by growth in the metro and submarine product lines, as well as strong demand for transponders. But revenue was down 11% sequentially from the previous quarter due to continuing softness in the high-speed long-haul product line.
>P>The IC business hit $812 million for the quarter, up about 1% from the year-ago quarter, but down 13% from the previous quarter. Demand was down for all of its IC lines except for wireless LAN.
The June quarter will be even worse. Agere estimates revenues will fall 14 to 20% from the quarter ended March 31, or total revenue of between $950 million and $1.025 billion.
(See April 24 story.)
Are foundries moving
into the driver's seat?
There seems to be a fundamental power shift going on in global chip making.
Taiwan Semiconductor Manufacturing Co. now says it will offer only a single version of its upcoming 0.10-micron process technology, which it hopes to roll it out in the third quarter of next year. This is in sharp contrast to the foundry market's current strategy of offering as many as five modified versions of its processes to match the large chip companies' production technologies.
TSMC had to do this because the integrated device manufacturers (IDMs), with their inhouse fabs and proprietary chip-making, weren't able to shift production easily to foundries. The IDM chips were designed to use the big chip makers' own manufacturing processes, and as a result, the foundries have had to tweak their processes--at significant cost--to land the big orders from IDMs.
"If IDMs insisted on using their own process, we would have to modify our process to match theirs, and that can take up to six months," says Shiang-Yi Chiang, R&D vice president. "But for the 0.10-micron process, we will not need to develop any revised design rules--this is a significant change."
TSMC has about five different variants of its 0.13-micron process, each one specific to a single, large-volume IDM. "The IDMs started developing their 0.13-micron processes two or three years ago, and TSMC was weak then," notes F.C. Tseng, TSMC president. "We couldn't negotiate."
"Creating different versions of their processes is something that TSMC and the other top foundries have been doing for some time in order to accommodate the IDMs," says Jim Hines, analyst covering the foundries for Dataquest. "But I think the foundries would much prefer if they could develop just one process."
The chip industry is moving toward a consolidation of manufacturing technology, Hines says, and TSMC's standard process has nearly taken on the mantle of a de facto standard. "As time goes on, we will see fewer and fewer IDMs developing their own processes," he predicts, for designs based on standard CMOS logic manufacturing. "Within three to five years, if not sooner, there will only be one or two standard processes within the semiconductor industry," Hines predicts.
(See April 25 story.)
Infineon starts equipping first
300-mm wafer production fab
Not everyone is delaying the startup of their 300-mm wafer fabs because of the economic downturn. Infineon Technologies says it has completed construction of its next-generation fab in Dresden, Germany, and is beginning to equip the cleanroom for fabrication of 256-megabit DRAMs using the latest 0.14-micron design rules.
A handful of other chip companies are preparing to ramp production in pilot lines and small startup fabs, but Infineon claims that its Dresden facility will be the world's first 300-mm wafer production plant. Production at the $1 billion production module should reach full capacity by the end of next year. First production is expected from the new 300-mm frontend line during the second half.
Like everyone building the new generation of fabs, Infineon expects to be able to slash its production costs dramatically. Even though the cost of a finished wafer will be the same as the current 200-mm wafers, the 12-inch platters will be able to carry two-and-one-half times as many chips.
(See April 23 story.)
Samsung surprises some
with first-quarter results
Think Samsung Electronics, and you come up with huge bank debts, big losses on DRAM sales, and a shaky local economy. Not good. But the world's largest DRAM supplier surprised some observers (including me) by turning in a fairly decent first quarter.
Net income amounted to $950 million, down only 22% from the first quarter last year. Sales hit $6.62 billion, down only 5.5% from last year. Samsung Semiconductor reported $2.3 billion in first-quarter sales on an operating income of $792 million and a profit margin of 35%.
Sales and earnings were both helped by Samsung's ramp-up in Direct Rambus DRAM production. Analysts estimate that Samsung has nearly an 80% share in this emerging global RDRAM market.
First quarter memory chip sales totaled $1.6 billion, LCD flat panels hit $385 million, and large-scale integrated circuit sales were $361 million. Samsung Electronics' information and communications business accounted for $1.5 billion in sales, operating income of $177 million, and a profit margin of 12%.
But Samsung Electronics reduced its 2001 capital expenditures to $4.7 billion, down nearly $1 billion from the originally projected $5.6 billion. All of the reduction was in planned spending for the chip business.
(See April 23 story.)
March chip gear sales are no
surprise, book-to-bill at 0.64
Those people who say the semiconductor equipment industry is turning around are smoking something. The chip-equipment industry is still in one deep slump. Look at the orders books for North American-based suppliers of chip gear in March.
Orders fell 23% to $1.3 billion from February's revised total of $1.7 billion, according to the SEMI trade group's monthly report. Shipments amounted to $2.04 billion, 11% below the revised February total, but is 17% above the March total last year.
The book-to-bill also continued to fall, hitting 0.64 compared to February's .073 ratio, meaning that only $73 of new orders were written for every $100 worth of products shipped in March.
"The March numbers reveal the continued erosion of bookings during the first quarter, which is not surprising given the continued softness in chip markets," explains CEO Stanley T. Myers of SEMI.
(See April 23 story.)
Pentium 4 may ignite price
war, but could restart PC biz
The Intel "battlestar" launched its latest strike this week against its chief and virtually only competitor, Advanced Micro Devices, and the only real winner would seem to be the consumer. Certainly not either of the protagonists, who are doomed to face increasingly narrow profit margins despite the claims of lower production costs.
Intel officially launched the superspeed 1.7-gigahertz Pentium 4 microprocessor, but in a bit of a surprise the chip giant priced its highest performing CPU at only $352 each in 1,000 quantities, only half of what the chip reportedly was going to be priced at originally.
Intel's price cuts, which extend to other CPU models as well, could also be what the doctor ordered to restart the PC business. Toward that end, Intel plans to spend more than $300 million in a campaign to boost PC sales, which have slipped significantly in recent months.
Intel called the Pentium 4 its first completely new desktop processor design since 1995, and will speed up such jobs as video encoding and other entertainment functions. But its major impact undoubtedly will come from its pricing. The new prices were obviously motivated by AMD's growing market share with its high-end Athlon chip. But AMD promises to match Intel price cuts.
(See April 23 story.)
If you have any comments, criticisms, or questions, don't hesitate to E-mail us at firstname.lastname@example.org. Have a great weekend!
(Click here for last week's Semiconductor Alert!.)