Commentary & analysis of week's chip news
Greetings from Down-East Maine. Returned this week from a 10-day family vacation to Quebec only to find the chip business still sliding. In fact, it now looks like we're running out of optimists. Before I left, there were still analysts and vendors arguing that the chip market would be ending its sales plunge by summer and be on its way back up in the third quarter. Well, June is turning out to be worse than most people expected. As for me, I'm still sticking to my prediction that the industry won't really recover until the fall of next year. For now, take a vacation, forget about the market, and enjoy the Fourth of July!
Talk about bad news:
look at wireline-chips
Late last year, wireline-chip makers started seeing a slowdown in orders but dismissed it as a minor inventory correction. But the so-called correction has turned into a major disaster. This week the In-Stat Group predicted that chip consumption in the communications market would fall 15.1% this year, but that might be the understatement of the year.
Given the news coming from several wireline-chip makers this week, the communications IC market is probably falling at an even faster--and perhaps unprecedented--rate. The bad news began on Tuesday when Applied Micro Circuits revealed what might be the largest sequential decline reported by a major chip maker.
Then on Wednesday, AMCC's archrival, Vitesse Semiconductor, warned it might report a 51% sequential drop in revenues for the current quarter. Then Xilinx, which gets the vast majority of its programmable logic sales from the communications market, said that it now expects revenues in its current quarter to drop 32% sequentially. That same day, its main rival, Altera, reiterated its previous projection of a 25% drop in second quarter sales. On Thursday, Agere (formerly Lucent Semiconductors) joined "the bad news bears" by announcing a massive restructuring as it anticipated quarterly sales to fall by 23.3% over the previous quarter.
Wireline-chip makers also may be last semiconductor vendors to feel the recovery, says one executive who tracks the industry closely. "The consumer-chip market will be the first to improve," predicts Tien Wu, who heads U.S. and European operations for Advanced Semiconductor Engineering, the Taiwan IC-packaging and test giant. "The next market to improve," he says, "will be the PC, and then wireless."
Almost no one sees a recovery anytime soon. "I'm thinking it's mid-2002," says Bob Pierce, vice president for Infineon Technologies' Communications Group. But he sees orders starting to come in again by November.
(See June 28 story.)
Agere fires 1/3rd of staff,
takes $900 million charge
The news out of Allentown is grim. Agere Systems--formerly Lucent Microelectronics--in just three months will have laid off one-third of its employees and taken a $900 million charge to cover new plant closings, layoffs, and other restructuring measures.
In a massive restructuring, the chip maker is laying off 4,000 more workers and will close down its wafer fab in Madrid, Spain, by the end of the year. It will also consolidate several "satellite manufacturing sites" as well as leased corporate offices. The company is trying to adapt its business in the face of deteriorating conditions, says CEO John Dickson.
Elimination of the 4,000 jobs comes just two months after Agere said it was cutting 2,000 employees. About 1,000 of the job cuts will come from the closing of Agere's Madrid plant. At the end of March, Agere employed 18,500 workers. Agere expects the restructuring and workforce reduction to result in an annual pretax savings of $520 million.
The company also cut its estimate for revenues in the current fiscal quarter. Sales will fall by 23.3% to $920 million, Agere predicts. It had been projecting a 20.8% drop from the quarter ended Mar. 31.
(See June 29 story.)
June quarter even worse;
AMCC sales to fall by 2/3rds
I've been covering the semiconductor business since 1958 (yep, 43 years ago) and I can't remember a steeper industry downturn. And we still haven't hit bottom. This week the bad news kept coming.
Applied Micro Circuits (AMCC) may have taken the prize for the largest sequential sales decline ever reported by a major chip maker. The San Diego company says sales for its first fiscal quarter ended June 30 will fall a staggering 62.8%-to-67.9% sequentially to $40-to-$45-million. It had expected sales of $70-to-$85 million.
"The magnitude of this downturn is greater than we, or our customers anticipated, and the progress in reducing excess inventory in the channel appears to be slower than they originally thought," says CEO Dave Rickey. "New order activity has remained weak throughout the quarter and our overall backlog has declined."
(See June 25 story.)
Xilinx also had a lousy June
The first three weeks in June were awful for Xilinx as well. As a result, the programmable logic supplier again slashed its sales forecast for the June fiscal quarter due to an unexpected drop in customer orders. It now expects revenues to drop 32% sequentially from $407 million in the March quarter. Just three weeks ago, Xilinx was predicting a 15-to-25% sales decline.
The culprit was a big falloff in "turns business"--orders that are booked and shipped in the same quarter. A bigger percentage of the chip maker's revenues now come from "turns business" due to customer ordering during the downturn.
Gross margins will decline to 52% in the first fiscal quarter, which ends June 30, because of lower-than-expected sales and a mix shift to higher density devices, the company predicts.
(See June 26 story.)
Vitesse joins freefall
with 51% drop in sales
Vitesse Semiconductor also is now watching its sales fall off a cliff. The Camarillo, Calif., vendor now expects a 51% drop in revenues during the current fiscal quarter to about $60 million from the previous quarter.
Until the new forecast, analysts were expecting operating net profits of 4 cents a share, but now the company expects to report a net loss of 6 cents a share, excluding charges for restructuring and other items.
"We have seen no improvement in visibility at a majority of our customers," says CEO Lou Tomasetta. "Inventories in our customers' channels remain at high levels and the process of reducing excess inventories is taking much longer than anticipated."
(See June 26 story.)
To fight slowdown, Altera
makes "difficult decisions"
The ongoing business slowdown is continuing to hit Altera hard. The San Jose programmable logic vendor is now making what it calls some "difficult decisions" in order to keep operations going.
It is laying off 152 people, or 7% of its workforce, as part of a restructuring. In addition, Altera is reducing executive officer pay by 10% and postponing employee merit increases.
At the same time the chip maker is taking a one-time, pretax charge of $115 million for excess inventory in second quarter. The inventory and restructuring charges are expected to amount to $105 million on an after-tax basis.
Altera continues to expect a 25% drop in revenues in the second quarter from the $287.4 million in first quarter sales. International business remains soft, although the company sees signs of stability in North American markets.
(See June 27 story.)
Over 40 fab shutdowns are
scheduled for this summer
The industry is using plenty of summer factory shutdowns to try to reduce its inventories. "We have been notified of more than 40 wafer fab shutdowns, varying in length from one to three weeks, that will take place before the end of our fiscal year in October," comments James L. MacDonald, president of Photronics.
As a result, the Jupiter, Fla., photomask maker has cut its revenue estimate for the fiscal quarter ending July 31 to between $80-million and $85 million--15%-to-20% lower than the $100.6 million recorded in the prior quarter.
"We now believe that the balance of fiscal 2001 will continue to be impacted by these conditions and do not expect that fourth quarter revenues and earnings will exceed those we report for the third quarter," MacDonald says.
(See June 26 story.)
A cool new way to
fabricate ICs on plastic
I've always been a sucker for new companies introducing an exciting new technology. But I'm often disappointed because these startups usually underestimate the amount of money and marketing effort it takes to launch a new business in competition with the existing corporate giants. Having said that, here's another rookie stepping up to the plate.
This week FlexICs opened up a pilot facility in Milpitas, Calif., to fabricate chips on plastic substrates. Making it feasible is an ultra-low temperature process developed by the company that needs less than 100 degrees C to fabricate ICs on plastic.
The year-old company says it has developed several proprietary techniques for high-performance gate dielectrics at low temperature and for photolithography on flexible substrates. FlexICs is hoping that the technology will make possible rugged electronic circuitry as well as thin, shatterproof, and flexible displays.
It will aim the plastic-based ICs at portable systems, memory modules, optical component integration, and imaging applications. "Our technology will enable these products to be lighter, more durable, and thinner," claims CEO Magnus Ryde.
(See June 26 story.)
Like most of Japan's chip makers, South Korea's Samsung Electronics spent countless millions in an attempt to carve out a piece of the lucrative global microprocessor market. But now it looks like Samsung will end up failing just like the Japanese did.
Samsung and Compaq together had invested $500 million in Alpha Processor, now called API NetWorks. The U.S. joint venture was designed to help Samsung enter the high-end processor market and expand its non-memory chip business. Compaq had acquired the Alpha processor and its technology as part of its $9 billion acquisition of Digital Equipment three years ago.
Samsung had high hopes for the 1998 joint venture. It tried hard to push the Alpha processor in the merchant market. The deal called for Compaq to develop the Alpha architecture and Samsung to be the foundry. Alpha's ambitious sales target was $100 million in 1999 and $1.5 billion over the next five years. But Samsung fell far short of its original expectations, analysts say. It has only a few foundry customers for the chip, most notably Compaq.
Then this week Samsung's Alpha processor received a body blow when Intel took over Compaq's Alpha processor technology. As part of the deal, Compaq will move all of its 64-bit server and workstations to Intel IA-64 bit processors by 2004. While Samsung will continue to make the Alpha chip on a foundry basis for Compaq and others, the processor is doomed in the long run, analysts say. It's not clear just how the deal will help Intel's 64-bit processor ramp, but it does remove at least one 64-bit MPU competitor.
(See June 25 and June 26 stories.)
Halla sees China
dominating IC design
We've been saying for some time that China has become the new frontier of the global semiconductor industry. Now some industry leaders see this nation gaining an increasingly dominant role in chip design as well.
Its booming business in PCs, wireless and landline telecom gear place China in a class by itself as other regional markets sag. That the number of cell phone subscribers in China will surpass the U.S. this year is evidence of that nation's growing clout, says Brian Halla, CEO of National Semiconductor. "It's possible that in not too many years the entire global semiconductor roadmap will be driven by China," he suggests in an interview this week with EBN.
China now has 14 new fabs under construction or about to break ground, Hall points out, far more than any other region in the world. He wonders what would happen if a significant portion of this new capacity went unsold, causing China to begin dumping chips on the world market. "If the SIA thought Japanese dumping years ago was a problem, try fighting with China on this issue," he warns. "SIA would look like an ant trying to attack an elephant."
Because of China's growing role in chip design, National is hoping to expand its early joint ventures there. National has a joint venture to sell chip sets for third-generation cellular phones, and another with Legend Computer, China's largest PC maker, to incorporate National's Geode processor in set-top TV boxes for Internet access.
(See June 29.)
Silicon lives! IBM promises
100 Ghz. chips in 2 years . . .
Big technology breakthroughs are exciting, but it always seems to me to take many years for them to move into commercial products. IBM Semiconductor just may have changed my thinking here. This week, the chip maker described a new silicon-germanium (SiGe) transistor that will be able to reach speeds of 210 gigahertz and still draw only a milliamp of current.
IBM, which has been the leading developer and promoter of SiGe technology, claims the new transistor is the world's fastest silicon-based device announced so far--five times faster than any other announced silicon based-transistor. It also claims that the design will be used in production communications ICs to get speeds of 100 gigahertz in less than two years, which would be four years sooner than recently announced competitive approaches (read Intel).
"Just as aircraft were once believed incapable of breaking an imaginary 'sound barrier,' silicon-based transistors were once thought incapable of breaking a 200-gigahertz speed barrier," notes Bernard Meyerson, IBM vice president. "Now, makers of high-performance electronics like networking gear will be no longer forced to use chips made of exotic and expensive materials to reach these speeds."
IBM started out in this technology direction 12 years ago by add germanium to the basic silicon transistor to speed up the electrical flow in the device and at the same time reduce power consumption. In the new transistor, IBM built an improved transistor design in SiGe material that shortens the electrical path to speed up the device. The SiGe device is based on a "heterojunction bipolar transistor" design where the electrical flow is vertical instead of horizontal like it is in standard transistors.
(See June 25 story.)
. . . But IBM's new SOI
may be even better
IBM's potency in R&D was demonstrated to me this week when researchers at IBM Microelectronics argued over what new IBM technology was better. Some engineers believe the low-power, 130-nanometer silicon-on-insulator (SOI) technology that IBM unveiled recently at the VLSI Technology Symposium in Kyoto may prove to be superior to the firm's latest silicon-germanium technology announced this week.
The company has developed SOI for both digital and RF applications. A low-voltage version for logic and SRAM works in the 0.6- to 1.2-volt range, while a higher-performance version works at up to 1.5 volts. That IBM can successfully move SOI into digital and RF is further proof of how trouble-free scaling SOI-plus-copper technology into new applications has become, declares Richard Wachnik, senior manager at IBM's technology integration division.
"Copper is easy to make and so is SOI, really," he says. "Those that back SOI technology--Motorola, AMD, Compaq and so on--they all constitute evidence of market acceptance of this fact." IBM continues to pour considerable resources into SiGe technology, which offers comparable performance but is more resistant to integration, he notes. Wachnik says that IBM could put its SOI technology into limited use "in a year or two."
Moving SOI research into RF applications could lead to more highly integrated devices and superior passives, Wachnik believes. This technology also has the potential to better solve the longstanding wireless and communication integration problems, he adds. "Once upon a time SOI was an exotic compound," Wachnik says. "We already had silicon germanium. Now, here is an alternative. SOI will allow us to combine and solve the RF applications and attach the digital."
(See June 27 story.)
Site prep to begin at Intel's
second 12-in. fab in Oregon
Intel isn't letting any grass grow under its feet, as far as building 300-mm fabs is concerned. It filed permit applications this week to build a $2 billion development fab capable of making chips with 0.07-micron design rules. The Hillsboro, Ore., fab would be Intel's second 300-mm wafer plant at this site.
The word had gotten out in May that site work would begin on a second 300-mm development fab named D1D, which would be used to develop next-generation processes for 300-mm wafers. Construction of the fab will begin in 2002, with completion slated for late 2003. Intel will probably install 157-nanometer wavelength lithography in the new fab to get to a 0.07-micron processing level.
The D1D fab will supplement Intel's current D1C fab at the site, which is developing 0.13-micron processing on 300-mm wafers using 193-nm wavelength lithography. D1C will push development down to the 0.10-micron level, overlapping the next-generation D1D fab.
(See June 27 story.)
Now, even Taiwanese are
making Rambus DRAMs
Now even the Taiwanese are building Rambus DRAMs--can price cutting be far behind?
Winbond Electronics has started ramping volume production of Rambus DRAMs and samples are now available of 128- and 256-megabit memories. Winbond is the first Taiwan chip company to launch Rambus DRAM production.
Just recently, Winbond indicated it was thinking of getting into the RDRAM business, but that an ample supply was then available from Samsung, Toshiba, and Elpida, the NEC-Hitachi joint venture. But now it says that demand is growing now for RDRAM memories from PC makers. "Rambus RDRAMs will be a strategic part of our product offerings this year," says president C.C. Chang.
Winbond is Taiwan's largest DRAM maker and ranked in the Top 10 DRAM suppliers last year, according to IDC. Taiwan chip makers presently account for 12% of the world's memory market. Winbond will not have to make significant capital investments to get into the RDRAM market because it will be supported by Toshiba, which currently supplies Rambus DRAMs to Sony for use in its Playstation 2 game consoles.
(See June 27 story.)
Journalist salaries don't keep
up with electrical engineers'
When I walked across the stage to get my diploma from the University of Nebraska a few years ago (actually it was in 1956), the dean told the audience that I already had been hired as an intern at The Wall Street Journal at a salary that was higher than that of a starting electrical engineer. I was making the grand sum of $100 a week, or far in excess of what starting EEs then earned--I think they got $4,000 annually.
Well, today's engineers can stop bellyaching about starting salaries--they're certainly higher than those of starting reporters. According to the latest salary survey by IEEE-USA, salaries climbed at an unprecedented rate in 2000, particularly for high earners. The median salary last year rose 13.5% to $93,100 from the $82,000 median income in 1998. Total median income, including salary and other compensation, rose to $99,000 in 2000, up from the $87,200 in 1999.
The latest survey also shows a continued widening of the pay differential between the top and bottom 10% of engineering professionals. Engineers reported a $97,000 split between those in the upper and lower 10%, the highest decile had a median of $162,000, while those in the lowest 10% had a median of $65,329. "That is a bigger difference than there has been in the past," says Dick Ellis, consultant who put together the survey for IEEE-USA. It shows that the top rung has gone up, he says. "Employers seem to be willing more and more to jack up what they pay for people who are really good."
Solid-state circuit designers also earned more than their bosses, reporting $112,000 in annual salaries. Engineers who worked in industrial application and systems and control had the lowest wages of those surveyed. In the industrial area, the median salary was $84,300, while systems and control engineers earned $84,150.
As one might guess, wages were higher on the east and west coasts. San Francisco salaries led the way, where the median was $122,000. Other cities with high medians were Boston and Baltimore with $104,000 and San Diego with $102,000. In contrast, the median in St. Louis was $78,000 and Tampa was $79,500.
Double Es with less than two years of experience earned a median of $70,000. The survey's typical respondent had 20 years of experience and earned a median salary of $98,685. Unemployment for EEs in 2000 was at its lowest level in memory, the IEEE-USA reports. Only 0.6% of those participating in the survey said they were unemployed.
(See June 28 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!.)