Commentary & analysis of week's chip news, Aug. 20-24, 2001
Greetings from Down-East Maine, where we are busily clearing and cleaning up an 80-year old apple orchard that adjoins a 180-year-old home we recently acquired along with a 125-acre "greenbelt" of forest surrounding our home. Old timers tell us the house hasn't been lived in for 50 years--it's great fun restoring historic homes since so many of them are torn down in this country. Weather is cool, but still no rain to speak of.
No sign of improvement
seen yet in chip business
Don't look for any recovery now, as the chip industry has not yet approached the bottom of the current semiconductor recession. So says maverick researcher Advanced Forecasting, which doesn't see any signs of improvement yet despite the increasingly optimistic comments coming from some chip makers.
The Cupertino, Calif., researcher maintains the latest data still doesn't suggest that business will bottom out in the third quarter. The researcher is "not encouraged by the downward direction" of its IC Recovery Index. "The decline in actual sales of ICs and semiconductor production systems support" its negative outlook, says Advanced Forecasting.
Recent suggestions that sequential growth is returning in the chip business--from such leaders as Intel and Applied Materials--are based on "vague signals" in weak markets, the market researcher says. Advanced has long claimed quantitative based models that accurately predicted inflection points in up and down semiconductor business cycles.
"Unfortunately, as with actual IC revenue, our IC Recovery Index continues its downward trend without any signs of the model slowing its descent," explains David Crume, marketing director at Advanced. "As we had expected, the month of May did not represent a minimum for IC sales."
"Actually, the IC industry broke a record in June," he points out. "The rate of decline in June of IC revenues reached 32.3% below their level a year earlier, surpassing the record of 1985, 30.9%."
(See Aug. 20 story.)
Motorola is thinking about
selling its chip business . . .
Hey, those rumors floating around the industry that suggest the unthinkable are more accurate than I thought. The word was that Motorola was thinking about selling its chip business. Now that's a stunner.
And this week officials from Motorola surprised me by indicating the electronics giant would indeed consider shedding its troubled semiconductor business--if the unit doesn't turn itself around.
A Motorola official confirms the corporation will consider a sale or spinoff of the Semiconductor Products Sector if it does not stem the flow of red ink. This shocking revelation follows a long string of recent setbacks for the chip operation, including major plant closures, big layoffs, and a second quarter operating loss of $381 million.
(See Aug. 21 story.)
. . . chip giant won't not rule out sale
or spin-off, but doesn't expect either
As word spread rapidly this week that Motorola might sell its struggling semiconductor business, COO Bob Growney tried to take the heat off by e-mailing semiconductor employees that media stories were a "purposeful misinterpretation" of remarks he made earlier this week to Wall Street analysts. Yep, the messenger is always blamed.
Asked by analysts what could happen if the chip unit couldn't improve its financial performance, Growney told semiconductor employees that "while a sale or spin-off could not be ruled out, we did not expect this to occur. He added that there "are no sacred cows and no company segment is immune." But he added, "our patience is not 'infinite' with all our business units."
Not surprisingly, the possibility of selling or spinning off the Moto chip unit is causing turmoil in that division. "I want to assure you that semiconductor is a valued and valuable part of Motorola," Growney reassured them. And he added that "we are confident that semiconductor can return to financial health in a reasonable period of time and continue its role as an essential part of Motorola."
Industry analysts believe that Motorola's position isn't all that much different from other major chip makers right now. "The industry has reached a point where vertically integrated companies are finding it challenging to be competitive," says Mark Edelstone, chip analyst for Morgan Stanley.
Some observers believe the chip unit may be looking down the road at the massive expenditure needed to build a 300-mm fab, which could be pushing corporate Motorola executives to embrace the idea of a spin-off. The company doesn't have a facility capable of producing 300-mm wafers and will have to invest upwards of $2 billion to build one. One observer noted that the company will likely have to at least start that process within the next 12-to-24 months if it wants to retain its role as a world-class chip company.
Industry insiders regard Growney's comments as most likely a carefully crafted message to Wall Street. "You have a bunch of financial analysts who want to know how many billions of dollars Motorola will have to spend to keep this capital-intensive business afloat," notes one former Motorola executive. "So they let it be known that they won't have any cash problems because they can always sell the business and generate cash. At the same time," he says, "this is a good way to let it be known that the business could be for sale, maybe as a way of creating a bidding war." Poor old Motorola Semi.
(See Aug. 24 story.)
Agilent still sees no upturn,
so lays off 9% of workforce
No positive news came out of Agilent Technologies this week, but there was plenty of negative info. The Hewlett-Packard spin-off is "not seeing any definitive signs of an upturn," complains CEO Ned Barnholt.
"Earlier this year, we thought we would see a recovery by this point," he says, "but the downturn is deeper and longer than what we thought. This is by far the worst industry downturn I've seen in my 34 years with the company," Barnholt says.
As a result, the Palo Alto supplier of test systems, semiconductors, and instruments is cutting more costs to restore profitability "as soon as possible." For one thing, it is cutting 9% of its work force, or 4,000 jobs.
Agilent reported a 33% sequential drop in revenues to $1.8 billion in the quarter ended July 31, and losing $219 million. The company recorded $240 million in cancelled orders, somewhat better than the staggering $500 million cancelled in the previous quarter.
There's no telling when Agilent will go in the black. Break-even for the company is quarterly sales of $1.9 billion. "I think we will hit $1.9 billion sometime in 2002," Barnholt says. "I can't tell if it's Q1, Q2, or Q3."
(See Aug. 21 story.)
Now analysts say it could be 2003
before chips return to full health
I wonder if the dynamics of chip market forecasting are similar to those in calling a stock market turnaround? You know the old saw: If analysts agree that stocks have hit their low and are about to go up, look for more months of falling stock prices.
At the beginning of this year, only a handful of us were saying the chip business wouldn't turn back up in early 2001. On Wednesday, industry analysts at a SEMI trade group panel in San Jose agreed that it may take as long as another 18 months for the ailing IC industry to get back to full health. That's even longer than my pessimistic predictions.
Taking a contrarian view, maybe this new unanimity means good news may be closer than I had believed.
The analyst panel was definitely negative about a turnaround next year. "We look at 2002 to be the healing year, but not a great year," says veteran chip analyst Bill McClean, president of IC Insights. "In 2003 or 2004, the industry will be rolling again." Wow! That's a long downturn for this business.
Average selling prices and unit demand for chips "are still going down," he warns. "We are not seeing a kick, because there is too much capacity in the market." IC Insights is predicting that the semiconductor industry will end up with worst year ever in 2001.
However, "we think we're at the bottom," says Carl Johnson, president of Infrastructure. "But we also think we will be bouncing there for a while."
The silicon foundry business remains extremely weak, says analyst G. Dan Hutcheson, president of VLSI Research. "If you look at the new foundries in Malaysia, they are running at only 10% to 20% in fab utilization," he notes. That kind of production really is shipping a $100 bill with every wafer they make.
"The semiconductor equipment market remains deep in the tank and will not recover at least until the second half of next year," Hutcheson predicts.
(See Aug. 23 story.)
EDA sales go up
Electronic design automation tools didn't do all that badly in the first quarter of this year. EDA sales hit $987 million, a 14% increase over the first quarter last year but a 6% decline from the previous quarter, according to data this week from the EDA Consortium.
Semiconductor intellectual property--IC design cores used in designing systems-on-chips--dropped 27% in the first quarter from a year ago to $26 million and 11.3% sequentially from the fourth quarter of 2000.
The strongest design automation software sales came in IC layout tools, which grew 42% on a year-to-year basis to $246 million.
The EDA industry's largest category, computer-aided engineering (CAE), generated $469 million in revenues during the quarter, an 8% increase over the same quarter last year. But sequentially, CAE revenues dropped 7.5% from the fourth quarter of 2000.
CAE's first-quarter growth was led by formal and functional verification tools, which increased 35% to $35 million over year-ago sales.
(See Aug. 23 story.)
Is that a bright spot at end
of chip equipment tunnel?
Hey, I found a tiny bright spot in the generally dark July business report from the SEMI trade group for North American semiconductor production equipment suppliers. Orders were up 5% in July from the previous month.
Equipment vendors received $764 million in orders in July, but they were still a whopping 74% lower than July last year.
Shipments were another story, falling 12% from June, SEMI says. This did enable the book-to-bill ratio to rise a tiny bit from 0.56 in June to 0.67 in July, meaning that $67 worth of new orders were received by equipment vendors for $100 of products they shipped. This was the third consecutive monthly rise since the book-to-bill hit its lowest point ever of 0.44 in April.
SEMI was optimistic as usual. Bookings provide "some indication that capital equipment orders may have reached bottom," says SEMI research director Elizabeth Schumann.
(See Aug. 21 story.)
No, that's not a bright spot,
I can't recall worse order rate
I don't remember order rates for chip equipment as bad as they were in the second quarter of this year. Worldwide chip-equipment orders totaled $4.29 billion in the second quarter, down a big 73% from the same quarter last year and 34% drop from the first quarter, according to the SEMI trade group.
And for the first half, the numbers were just as bad. "Net bookings of semiconductor manufacturing equipment in the first half of 2001 are 63% below the same period last year," says Elizabeth Schumann, SEMI research director.
"Although shipments in the first half are down only 16% from the same period last year, the continuing decline in orders . . . supports our expectation that the overall equipment market will decline to approximately $31 billion this year," she predicts.
(See Aug. 24 story.)
Fabs ran only 72.7% of capacity
in second quarter--a 20-year low
Would you believe that wafer fabs are more under utilized today than at anytime in the past two decades? Well it's true. And it ain't good with all that money they owe on their chip plants.
The latest numbers show that the industry's wafer fab utilization rate plunged to 72.7% of installed capacity in the second quarter, down from 83.8% in the first quarter and below the record low of 80.5% recorded three years ago, according to the quarterly Semiconductor International Capacity Statistics (SICAS).
And believe it or not, wafer processing capacity was added, not removed, during the quarter. The world's installed fab capacity grew sequentially 3.6% to 1.32 million eight-inch equivalent wafers per week, reversing a 0.5% cutback in the installed capacity during the first quarter.
Total MOS capacity grew 3.7% sequentially to 1.2 million eight-inch equivalent wafers per week and 17.6% over the industry's processing capacity a year ago. Capacity utilization fell to 73.1% from 84.2% in the first quarter.
Bipolar processing capacity also grew sequentially by 2.4% to 306,200 five-inch equivalent wafers, as bipolar utilization plummeted to 68.3% in the second quarter from 80.1% in the first quarter.
(See Aug. 14 story.)
Fujitsu begins massive cutbacks
to get back in profit column . . .
Fujitsu is making massive cutbacks to get back in the profit column. The Japanese giant will cut 16,400 jobs, close at least five semiconductor plants, and take a $2.5 billion restructuring charge. This radical surgery, company executives hope, will enable Fujitsu to produce an operating profit of $3.33 billion in fiscal 2003.
The company will stop making and selling hard-disk drives for PCs, and curtail printer production. As part of this pullout, it will lay off 4,200 employees in the Philippines, Thailand, and Vietnam. It will continue to build hard drives for servers and notebook computers.
Fujitsu also will close down fabs in Mie, Iwate, and Aizu-Wakamatsu, reducing its front-ends from 12 to nine. It will also cut two back-end and test lines in Japan, leaving five operational. The line closures will cost about $1.04 billion.
The company also will spend $375 million to streamline and downsize its legacy telecommunications equipment businesses, and will now focus only on photonics, Internet Protocol, and 3G mobile equipment. It will take a $250 million charge to consolidate its server and storage businesses and $250 million writeoff to restructure other units to focus on higher value-added businesses.
Fujitsu figures it is facing "an extremely elongated U-shaped recovery." And with no economic remedies around the corner, its only option was to reduce fixed costs. The layoff represents one of the largest job cuts in Japanese corporate history, touching nearly 10% of the work force. About 5,000 jobs will be lost in Japan and 11,400 will be cut overseas.
(See Aug. 20 story.)
. . . but forms venture with AMD
instead of closing Oregon flash fab
But Fujitsu's massive cutbacks didn't include its fab turning out flash memory in Gresham, Ore. Rumors had circulated that the company would sell or mothball the facility. Instead, Fujitsu will form a joint venture with its flash memory partner, Advanced Micro Devices, to "govern the fab."
"This year the flash and system LSI business has been extremely poor," says Kazunari Shirai, Fujitsu EVP. "After several months of discussion with AMD, we have decided to form a 50-50 joint venture with AMD and we will sign a memorandum of understanding with them either next month or the month after."
He denies there will be any immediate job cuts at the Oregon plant. Fujitsu chip executives were torn between selling the Gresham facility or closing it down, according to one informed source. "There is an indecisive situation in Fujitsu management," he says. "It is also clear that AMD just doesn't have the money to buy the fab, even if they wanted to."
The decision to form a new joint venture represents an amenable half-way solution for Fujitsu and AMD to split costs and share risks, says Michito Kimura, analyst with International Data Corp. in Japan. "They seem to be looking at a risk-hedging strategy," he adds.
(See Aug. 20 story.)
How cell-phone could
change DRAM market
What do you do when your primary customer has run head-on into a falling market for his products and you're awash in inventory in a product that represents most of your sales? Go all out to find new customers, of course.
And that's exactly what the largest producer of DRAMs has done. South Korea's Samsung Electronics signed a deal to furnish low-power versions of DRAMs to cell phone giant Nokia. This agreement puts Samsung out in front in a race to sell DRAMs into a new market estimated to hit more than one billion units annually by 2004.
The other Big Five DRAM vendors also are hustling to sign up cell phone OEMs for their low-power DRAMs. Everyone is hoping to improve their margins by catching this trend early and partially offsetting their heavy losses in the PC market where DRAMs are now selling at or below their production costs.
What's driving this emerging DRAM market is the need to provide sufficient storage capacity to handle the huge volume of images and data that will flow next-generation Internet-enabled handsets.
Cell phones equipped with interim 2.5G service usually have sufficient extra bandwidth to provide adequate Internet service, but when carriers launch 3G wireless service in the next few years, cell phones will have to process far more information, including streaming video. That will require memory densities far in excess of what SRAM, which stores data in most cell phones today, will be able to offer.
While these new-generation cell phones need a less DRAM than a PC, the new market's unit volume will be several times larger. This year alone, 175 million low-power DRAM chips will be shipped to the handset sector, accounting for 4.7% of the total DRAM market, according to Sherry Garber, analyst at Semico Research. She expects that total to increase to 1.17 billion units, or 18% of the DRAM market, in 2004, and 2 billion units, or 28%, in 2005.
The low power DRAM could change the current DRAM business significantly. Unlike commodity PC memory, for example, the new DRAM will become more customized to meet the different requirements of handset OEMs. In fact, cell phones may be the catalyst for a wave of application-specific DRAMs that will fragment the market, says Mike Despotes, president of Elpida Memory.
(See Aug. 20 story.)
More people are thinking
about 'ditching Bluetooth'
For a year or more, I have been critical of the many pie-in-the-sky forecasts calling for a multi-billion market to develop in just a few years for Bluetooth chips.
Support for the low-powered, RF networking technology may now be waning, as speculation builds that the highly touted Bluetooth may not be able to compete with the faster 802.11b on the high end or the simpler, low-power alternatives on the low end.
This speculation was reinforced this week by Bluetooth being notable by its absence at the IEEE Rawcon conference in Boston, which focuses on all things wireless. A mere two papers were presented on Bluetooth.
"Despite the hype and expectations, Bluetooth clearly hasn't lived up to expectations," says Craig Mathias, principal at the Farpoint Group. "We're telling clients to not pay a whole lot of attention to it."
Ditching Bluetooth is "definitely something people are thinking about," acknowledges analyst Navin Sabharwal with Allied Business Intelligence. But he cautions against writing off the technology. "There's still a lot of development happening on the Bluetooth front, so I think it's very premature to call Bluetooth dead," Sabharwal says.
But Mathias argues that Bluetooth will not cost less than the 802.11b WLAN technology that's already shipping and that the window has closed on Bluetooth. "Its only real advantage is power consumption--and if that's important to you, just turn the power down on the 802.11b radio."
But Allied's Sabharwal "has a very hard time seeing 802.11b being integrated into cell phones and cordless headsets." He still gives Bluetooth the edge in form factor and pricing.
Bluetooth was never meant to be a high-speed network connection in the realm of 802.11b, just a simple cable-replacement technology for personal-area networks. As such, its most highly touted application remains cell phones and other portable devices.
But on the cable-replacement side for headsets, Bluetooth is under pressure from established, lower-power, lower-cost alternatives such as magnetic-field-based connections. Aura Communications, a leading developer of such devices, is about to sample its LibertyLink programmable single-chip solution for point-to-point and point-to-multipoint connectivity. Aura claims it offers a 10-to-30-times power-consumption improvement over Bluetooth's RF solution at a lower cost. Stay tuned--the battle ain't over.
(See Aug. 20 story.)
Two China fabs compete to be
first to fab 8-inch wafers in 2001
Here's more evidence that China's chip making industry is on the move. Semiconductor Manufacturing International (SMIC) is starting to move production equipment into its fabrication facility at the Zhangjiang Science Park outside Shanghai in a drive to become that nation's first eight-inch wafer foundry.
SMIC wants the fab to open in November and by year's end turn out a few thousand wafers per month using a 0.25-micron process. The $1.4 billion first-phase provides for two wafer fabs and a support fab, which will produce photo masks and handle metalization. At full capacity, the main fabs will be capable of churning out 45,000 wafers per month.
Another construction project is underway on a second advanced wafer foundry in China. Grace Semiconductor Manufacturing also hopes to be turning out wafers before the end of the year.
(See Aug. 20 story.)
Intel goes public with latest
CPU push outside PC world
Intel has started going public with its latest effort to move into computing markets outside the PC. The PC processor leader stumbled in a previous attempt to crack the server market, but now is ready to try again.
While Intel is making some headway now in high-end computing with its Xeon and Itanium processor lines, it has lacked a platform to turn such CPUs into multiprocessing systems. Intel's latest answer is the 870 chip set, which will allow OEMs to build 15 processors into a single system. The company is expected to roll out the set early next year for use with a new version of its Xeon 32-bit processor and for the next-generation IA-64 processor, or McKinley.
"It's another step by Intel to get into computing markets outside of the pure PC space" says Dean McCarron, analyst with Mercury Research. "Intel is doing the kind of engineering that was once done only in server and high-end computing. They're significantly lowering the barrier to enterprise computing."
Learning from its Rambus DRAM experience, Intel has loosened its grip over server design engineering and is giving OEMs wide latitude to make their own design decisions. Intel's willingness to license its key I/O switching technology is one example of this newfound openness. "What we're doing is providing the ability to design building blocks so you can expand by bringing in different topologies," says Fay Briggs, Intel's director of chip set architecture.
Intel last year ran into problems with a server chip set design and decided to cancel the project. After licking its wounds, Intel is now trying to make up for lost ground with its homegrown 870.
(See Aug. 23 story.)
We welcome your feedback, comments, criticisms, or questions. E-mail us at firstname.lastname@example.org. Have a great weekend!
(Click here for last week's Semiconductor Alert!.)