Commentary & analysis of week's chip news
Greetings from Down-East Maine, where the vacation season is in full swing. The fairs, yard sales, bean suppers, and town celebrations are in full swing, and the Maine State Fair is right around the corner. Our local historical society's talent show goes on tonight with 17 acts. Come see it--admission is only $5 or $12 for the entire family--we expect a sellout. It's being held in a 190-year-old former Methodist church.
Has battered chip gear
industry reached bottom?
Semicon West certainly wasn't much fun this year for chip production equipment makers. Rarely if ever has any industry gone from an 80% growth year in 2000 to what increasingly looks like a 40% down year! Amazing, huh, even for the feast-or-famine chip gear business.
Most executives attending the giant San Francisco exhibit didn't even want to predict the 2002 market because they still don't have a clue as to what will happen in the second half of this year. About the only thing they could agree on was that current business just plain stinks.
"There is no capacity being added for 8-inch fabs at all anywhere except for China," declares Arthur W. Zafiropoulo, CEO of Ultratech Stepper. But, unlike some, he believes the industry has "reached the bottom." But he adds quickly that the business is still "going to bump along for a while and we won't have any visibility until October or November."
But that didn't stop the SEMI trade group from coming out with a new downgraded forecast. Its midyear consensus forecast, based on a survey of its worldwide members, now calls for a 35% decline in revenues for semiconductor equipment this year--or falling to $31 billion vs. $47.7 billion in 2000.
The typically optimistic forecast shows a slow, but steady recovery over the next three years with semiconductor capital equipment spending reaching $42.4 billion by 2003--but that's still 11% below last year's $47.7 billion blowout.
Many executives believe that forecasts will drop again from most market researchers before the end of this year--and I agree.
(See July 18 story.)
Don't take market
forecasts too seriously
Market forecasters have got to get used to journalists going back to check their old "clip" files. A year ago at Semicon West, Dataquest forecast that worldwide semiconductor capital spending would increase 68% in 2001. By the end of last year, the Gartner unit had changed its mind and predicted an increase of just 10%. By May, its equipment forecast dropped off a cliff--to a 24% decline this year.
This week Dataquest dropped its global forecast once again. It now projects the worldwide semiconductor capital spending at $47.3 billion this year, a 26% decline from last year.
Yup, let's see if I got that right. Instead of a 68% increase, Dataquest now says global capital spending in 2001 will be down 26%. Let's see--the market researcher changed their forecast by only 100 percentage points over the past year. Morale of the story: don't take market forecasts too seriously.
(See July 18 story.)
New delay in shipping
193-nm tool may hurt Intel
Intel may be getting into a real pickle with its timetable to move to 130-nanometer process technology.
ASML USA, formerly Silicon Valley Group, once again has delayed the initial shipment of its latest tools, which are based on 193-nm, argon-fluoride (ArF) technology. The latest delay will be for at least three to four months, a slippage that most likely impacts Intel and its 130-nm and below timetable.
The company originally was supposed to deliver its Micrascan 193 scanner in April. But it delayed shipment until July. Now the company has confirmed reports that it will not ship the Micrascan 193 until October.
For months, Intel has been counting on the Micracan 193 tool to process chips at the 100-nm mode and the first model reportedly was headed to its 300-mm development fab in Hillsboro, Ore.
Nikon reportedly has already shipped a 193-nm tool to Intel and both 193-nm models reportedly will be involved in an elaborate "mix-and-match"' manufacturing process. Intel flatly denies the delays would impact its push to 130-nm technology, but analysts believe that Intel may be running out of patience with ASML, prompting the chip giant possibly to evaluate competitive scanners.
While Intel is waiting for the delayed SVGL tool, it can use what's called hard phase shift masks with its current lithography systems to reach 0.13-micron feature sizes on chips, according to sources close to the Intel litho situation. However, unless Intel has been developing phase shift mask processing all along, it will take additional time for the chip maker to qualify this alternate approach. Phase shift masks, by the way, are more expensive to use and will only run up Intel's already high Pentium 4 production costs.
(See July 18 story.)
ASML doesn't expect to see
upturn until 2nd half of 2002
ASM Lithography has grown increasingly pessimistic about business this year and now doesn't expect to see any upturn before the second half of next year.
The Dutch chip production equipment maker's CEO, Doug Dunn, won't even make any estimates on how business will go for the rest of this year. "The uncertainties and constantly changing dynamics in the industry make it very difficult to predict customer's order behavior," he complains. During the first half, he says that "customers continually eroded their capital requirements with a sequence of cancellations and push-outs."
While his June 30 backlog looks "reasonably robust under the current market circumstances," Dunn points out that most of this backlog represents orders for system deliveries in 2002. "That is why we have implemented aggressive and substantial cost saving measures, including a reduction in the number of employees."
This week ASML reported its first-half results and they generally followed the lithography systems supplier's warning last week of a big loss and shrinking tool deliveries and backlogs. ASML had net sales of $875 million for the six months ended June 30, down 24% from $1.52 billion in the first half of 2000.
Some 250 units were shipped in the first half, including 120 lithography systems. This compares to 322 units last year, which included 207 lithography systems.
Net income collapsed from $128 million in the first half of 2000 to a net loss of $83 million in the first half of this year.
(See July 18 story.)
Most folks seems to accept
Intel's lousy quarter but me
While Wall Street seemed happy that Intel's second quarter results met everyone's reduced expectations, I just couldn't get past the fact that the world's largest chip maker had a lousy quarter. Revenues dropped 24% from the year-ago period, net income declined 76%, and earning when acquisition costs were included collapsed by 94%. Those are big numbers.
But CEO Craig Barrett told shareholders that "our microprocessor business performed better than expected, with sequential growth in units, while our communications and flash businesses remained soft."
He expects third-quarter revenues to be in a range of $6.2-to-$6.8 billion; that would be something between a drop of 1.5% and an increase of 7.9% from the previous quarter. Intel estimates that its gross margin percentage in the third quarter would be flat -- about 47% plus or minus a couple of points--as compared to 48% in the second quarter.
The Santa Clara company also is sticking with its capital spending plans of $7.5 billion in 2001.
(See July 17 story.)
Wow! Conexant losses
exceed its quarterly sales
Just how bad the chip business is getting was demonstrated dramatically this week when Conexant Systems reported a net loss that exceeded its sales in the June 30th quarter. The Newport Beach, Calif., chip maker had revenues of $200.1 million for its fiscal third quarter and a pro forma net loss of $220.4 million.
Sales fell 20% from previous quarter and were slightly below the lower end of guidance given by the company in April.
The pro forma net loss for the third quarter was $220.4 million, or 89 cents per share, using a tax rate of zero. This compares with a pro forma net loss in the prior quarter of $187.5 million, which used a 38% tax rate. The June 30th quarter loss included $45.1 million charged for inventory and other items.
In perhaps the understatement of the week, CEO Dwight Decker says "the market environment continues to be challenging." But he does "believe that the personal networking business has found its bottom, with a stabilization in demand for many of our products in the mobile communications, personal computer and PC peripheral market segments." From now on, he says, Conexant anticipates "relatively flat sequential revenues for this business."
(See July 19 story.)
Will foundry business
bottom out this summer?
The foundries are really getting creamed this summer and Singapore's Chartered Semiconductor may be getting hit the hardest. Chartered is now warning that its fab capacity utilization during this quarter probably will fall to the mid-20% range, down from 31% in the second quarter and 107% in late 2000.
That means Chartered's fab utilization rate is running about half the industry rate. The global rate will run in the mid-40% range in the third quarter, estimates Dataquest analyst James Hines. But he figures the global foundry business will "most likely" hit the bottom of the current recession in the current quarter.
This chip recession is the worst ever in the 14-year history of the silicon foundry business, Hines says. "We have seen average foundry utilization rates plummet from nearly 100% in late 2000 to below 50% in just two quarters," he says. "Clearly it has been the worst oversupply situation that the industry has yet experienced."
Because of Chartered's deepening problems, it is slashing its capital expenditures budget by another 30% and has decided to postpone the start up of its 300-mm fab by one full year to 2003. This delay comes as some equipment industry leaders are beginning to worry that several 300-mm fab projects could fall victim to the chip recession.
Chartered is cutting capital spending 2001 to $700 million, down from $1 billion. Earlier this year, it had cut its original 2001 capital-spending budget by 20% to about $1.2 billion.
"We don't believe we'll need the 300-mm Fab 7 until 2003 because we have significant leading-edge capacity in our fabs 5 and 6," maintains CEO Barry Waite. It could take a year before the foundry business completely recovers, he predicts. Even by the end of next year, the average foundry capacity utilization rates are not likely to be above 80%, he adds. "The next 12 months will be the most difficult year ever faced by the foundries," Waite says.
This week Chartered reported its net sales fell sequentially 63% to $100.7 million in the second quarter from $271.4 million in Q1 of this year. The Singapore company is also bracing for another sequential drop in its foundry revenues in the third quarter. It now expects sales to be 15% lower than the second quarter's $100.7 million.
(See July 20 story.)
Would you believe NO
turnaround until 2010?
If you still think I've been too pessimistic about when the industry will turnaround, check this out. According to an industry analyst at Deutsche Bank, growth in the global semiconductor industry will remain weak until at least 2010. That's a decade away!
"The long cycle we have entered is one of weak growth until at least 2010. Any rebound between now and December would in fact be no more than a micro-cycle of small scope and short duration," analyst Masaru Koshita was quoted this week by the French daily Les Echos.
A Deutsche Bank official denied the bank had issued a formal report, saying that the "study" referred to by the financial daily may have been "a stray comment" by one of its Japanese analysts, speaking in his native language at a Paris conference on Friday.
According to Les Echos, the analyst said that chip companies refused to acknowledge that "the real problem" affecting this market was demand. "Many firms have invested very heavily in this equipment planning on continuous growth in these sectors . . . but a lot of this equipment, particularly PCs, has proven far too high-performance compared to the market's real needs," Koshita was quoted as saying.
"From now on," the Deutsche bank analyst said, "the dynamic will be one of replacing existing equipment, which translates into longer cycles."
(See July 16 story.)
Few changes seen in next
chip technology roadmap
This technology document is perhaps the most significant industry guidebook around--most chip makers use it to do their product planning. And this week the first draft of the 2001 edition of the Technology Roadmap for Semiconductors was released.
But the changes proposed may be fewer than were made in past years. "Microprocessor and logic parameters may end up being moved up slightly, but the 1999 roadmap on DRAMs is likely to remain unchanged," says Robert Doering, Texas Instruments senior fellow who is deputy leader of the industry working group revising the roadmap. Final version is due to be released on Nov. 29.
The existing 1999 roadmap shows microprocessor frequencies hitting 2.1-gigahertz clock rates this year--a milestone that is being reached by MPU makers. But the first draft of the 2001 roadmap calls for the die shrinks with finer device geometries may accelerate in the future. That could mean speeds would move forward in the next roadmap. The 1999 roadmap now shows clock frequencies increasing to 2.5 gigahertz in 2002, 2.9 GHz in 2003, 3.5 GHz in 2004, and 4.15 GHz in 2005.
But the first draft shows DRAM technologies on the same schedule laid out in 1999. That schedule has 2.5-gigabit DRAMs introduced in 2002, 2.9-gigabit memories in 2003, 3.5-gigabit DRAMs in 2004, and 4.15-gigabit chips in 2005.
Experts believe the pace of DRAM technology won't accelerate in this decade, partly because of the big decline in DRAM sales. Market conditions and lack of profits are expected to douse efforts to accelerate DRAM technologies. The initial draft of the 1999 DRAM roadmap had DRAM technology racing ahead to the 1-terabit devices by 2014, but it was cut out of the final version.
The 2001 roadmap also will retain most of technical roadblocks listed in the 1999 roadmap, says Doering. "In some cases they have been moved out another year or more." Only a few of the "red" problem areas listed in 1999 have been solved, he adds.
(See July 19 story.)
Guess what vendor claims lead
for DSPs in Internet products?
Texas Instruments isn't letting the chip recession slow their market efforts. The Dallas chip maker is beating the drums now for its sales lead in digital signal processors for Internet audio-enabled products. TI claims its monthly shipments of these DSPs surpassed 3 million in June, more than any other programmable semiconductor maker.
TI also reports a bunch of new design wins using its chips with Compaq Computer, Clarion, Olympus, Pioneer, and Pontis. And current customers Thomson Multimedia and Digisette have just ordered additional products, it says.
TI isn't alone here. "Growth remains strong in today's digital audio hardware market, as consumers continue to spend on high-quality products," points out Mike Paxton, analyst at the In-Stat Group. "TI remains on the leading edge of this market due to their consistent technological advances, key partnerships with industry leaders and a growing number of successful design-wins." Hmm.
TI expects to see increasingly strong consumer demand for products with Internet audio capabilities such as cell phones, PDAs, Internet appliances, home stereos, and car audio players.
(See July 18 story.)
Hewlett-Packard patent could result
in molecular-scale circuits by 2005
Here's my technology of the week story. The full impact of this R&D may be a decade away, but this work could have major impact on the semiconductor business. Hewlett-Packard has just received a patent that it believes may have eliminated a major obstacle to building a molecular-scale computer.
The patent--issued this month by the U.S. Patent Office to HP Labs scientists Phil Kuekes and Stan Williams--comes up with a way to solve the problem of connecting molecular-scale devices to today's much larger integrated circuits.
"We have a strategy to reinvent the integrated circuit with molecular rather than semiconductor components," points out Williams. "We've received two key patents and have several more pending that we believe will eventually enable computers to be millions of times more efficient than they are today."
HP's roadmap calls for molecular electronics, which it regards as a brand new technology, to augment silicon-based integrated circuits within the decade and eventually replace traditional solid-state memories. Most experts believe that silicon technology will reach its key physical and economic limits by about 2012.
"Once you've built a circuit from molecular-scale devices--something the size of a bacterium--the question is how you get data into and out of it," says Kuekes. "To that, you have to bridge the size gap between current technology, which is a hundred times bigger, and molecular-scale wires and devices, which are about the size of a bacterium."
Since it would be just about impossible to make precise connections between molecular-scale wires and today's ICs, the new patent proposes making connections randomly using a chemical process.
Researchers are now working on fabricating circuits from these components. They expect to be able to fabricate a 16-kilobit memory using this approach by 2005.
(See July 17 story.)
Applied is latest to cut forecast
for this year's chip gear market
Is there no end to the parade of industry sales forecasts being revised downward? Nope, and I'm afraid it will continue for a while longer.
Latest big name to lower its forecast for global semiconductor equipment purchases is Applied Materials, which says that the industry's sharp slump is still showing signs of slowing in a "second phase" of capital spending cuts.
Applied now estimates that chip-processing tool purchases by chip makers will fall 30% from $33 billion to around $23 billion. Based on brand new estimates, all capital spending by chip manufacturers will drop 27%, while chip revenues will be down 23%, predicts CFO Joseph R. Bronson.
The CFO notes that forecasts for industry capital spending and fab tool purchases have been dropping continually every month recently. He said the technology sector now has fallen behind the rest of the industry sectors and no clear driver is yet in sight in the information technology markets. Foundries are suffering through their worst recession ever, and the DRAM market is experiencing an unexpected wave of additional price cutting. All of these developments are setting up a second phase of the downturn, says Bronson.
What's beginning to worry Applied is the continuing downturn could end up hurting the momentum of the 300-mm movement, which was delayed once before by the 1998 semiconductor downturn. Most fab budgets for 300-mm wafer production gear are being spared the budget ax--so far, says Bronson. But worries that a prolonged downturn and a continuation of low semiconductor profits could cause some chip makers to push out or cancel major 300-mm projects.
Applied is still forecasting that 300-mm tool demand in 2001 will represent about 25% of customer "technology purchases"--meaning frontends aimed mostly at R&D and pilot production of new processes. About 65% of total tool orders are for these technology buys, he estimates.
Bronson estimates that 15 fabs globally will be in volume production with 300-mm wafers by 2003, while nine more 300-mm R&D and pilot fabs will running that year. If those projections hold up, Applied figures about $16 billion will be spent for 300-mm fabs in 2003.
(See July 18 story.)
Cypress lays off 650,
as sales keep tumbling
Cypress Semiconductor is taking drastic measures to "right-size" the company as sales continued to drop precipitously. Sales in the second quarter ended July 1 tumbled 29% from the previous quarter, the same drop they took in the first quarter.
"Our company's quarterly revenue has effectively been cut in half in six months," reports CEO T.J. Rodgers. "We are proud to have endured this precipitous decline in revenue without losing money."
Cypress eked out a profit of $787,000 for the second quarter, before acquisition-related costs and non-recurring items. Revenue amounted to $186 million, down 38% from a year ago when revenue was $301 million. The San Jose chipmaker hit record revenue of $370 million in the fourth quarter of 2000.
"The absence of a sizable short-term recovery in the datacom segment--despite promising signs in the computation and consumer segments--along with pressure on average selling prices will cause us to be unprofitable for the rest of 2001, unless we take action," Rodgers points out.
Cypress will take a restructuring charge of $140-to-$180 million in the third quarter, mostly to write down excess manufacturing equipment and to reduce its manufacturing headcount by 500. There also will be a 10% across-the-board reduction in force in non-manufacturing jobs, bringing the total reduction in force to 650, he says.
(See July 19 story.)
Applied launches new line
for 0.10-micron designs
What better way to get a handle on new trends in semiconductor production equipment than to check out the Semicon West trade show in San Francisco this week. The exhibit certainly is Applied Materials' top showcase, and the market leader didn't disappoint--even in this recessionary year.
The equipment giant unveiled the first products in a family of tools that it says will pave the way to what it calls the "nanochip" era. The new line, which will make possible chips with design rules of 100-nanometer (0.10 micron) and below, "will significantly advance the future of chipmaking," says CEO James C. Morgan.
An atomic-layer-deposition (ALD) chamber for its chemical vapor deposition (CVD) gear in the new line was exhibited. First applications will be depositing thin, conformal high-purity films of titanium nitride and tungsten nucleation layers.
(See July 16 story.)
Will Hynix's Oregon fab
be upgraded, or what?
Is this the beginning of the end for a DRAM factory in Eugene, Ore.? Hynix Semiconductor, better known by its old name--Hyundai Electronics Industries--is shutting down its fab in Eugene, Ore., for six months.
The big DRAM vendor has denied reports that its aging fab is up for sale. And now they say that plant will be upgraded during the shutdown to make it competitive in the current memory market. A new 0.13-micron process will be installed that will be able to turn out 256-megabit chips. Currently the plant can make only 64-mgeabit SDRAMs with its 0.22-micron process.
But upgrading takes money, and Hynix currently is working through a major financial crunch as it reschedules more than $2 billion n short term debt that's coming due shortly. So what's going to happen? Some 600 employees, who are now idled by the shutdown, would sure like to know.
(See July 18 story.)
To get Pentium 4 rolling, Intel
speeds up Brookdale intro . . .
Intel is pulling out all stops to light a fire under its struggling Pentium 4 microprocessor. As part of this drive, it is pushing up the launch of its Brookdale SDRAM-enabled PC chipset to late July.
The speeded up introduction of the Intel 845 chipset is part of an accelerated roadmap to turn the Pentium 4 into the firm's dominant mainstream processor by the end of the year, according EVP Paul Otellini.
However, the version of the 845 that will be introduced this month will ship with the double-data-rate capability disabled since Intel will still not roll out a DDR SDRAM-enabled version of the Brookdale until the first quarter of next year.
Some analysts now expect the venerable Pentium III family to all but disappear from Intel's desktop processor roadmap. But it will be retained in the notebook market for the time being, they figured. Intel plans to introduce a Tualatin-class Pentium III built with its 0.13-micron process within two weeks. A Pentium 4 notebook processor isn't expected until 2002.
(See July 18 story.)
. . . But Intel needs to cut prices
to maintain lead in notebooks
Despite the growing competition from the other three MPU vendors, Intel should be able to maintain its "dominant" marketshare in the notebook-based microprocessor business, according to a new report by In-Stat/MDR, a Cahners research unit just formed by the merger of In-Stat and MicroDesign Resources.
But to solidify its leadership position, Intel will be forced to lower the average selling prices (ASPs) for its notebook-enabled Pentium III microprocessor lines, predicts analyst Kevin Krewell. Although prices of desktop Pentium 4 and Pentium III processors have been slashed recently by up to 51%, mobile Pentium III prices are expected to decline only slightly during 2001.
And it will be the second half of next year before they are cut substantially, InStat/MDR forecasts. Intel's ASPs for its mobile MPU will fall by 20% from the 4th quarter of 2001 to the 4th quarter of 2002 as a result of increased competition and continued reliance" on the aging Pentium III, Krewell predicts.
"AMD, Transmeta, and Via will continue to nibble at Intel, offering slightly different combinations of performance, power, and price," he says. "But considering that Intel's present mobile volume is roughly the same as AMD's total CPU shipments, including desktop, the alternative vendors have a long way to go to supplant Intel's sheer volume and manufacturing capacity," Krewell says.
Nevertheless, Intel's notebook-based processor business is slowing down. "Despite the present economic slowdown, we project that a cyclical recovery in the fourth quarter of 2001 will allow Intel's mobile unit volume to grow 5% in 2001," he says. In 2002, the notebook PC-based processor market will return to a normal growth rate--rising 26% over 2001, according to In-Stat/MDR. Making this happen will be pent-up demand and a migration to WindowsXP, he says.
(See July 18 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!.)