Many Manhattan firms are reacting the same way--it is really gung ho stuff! We still have a long fight ahead of us, but Americans are up for the job! God bless America!

Book-to-bill for chip tools
surprises, starts falling again

Like Boy Scouts struggling to light a fire by rubbing two sticks together, the chip production equipment industry has been working frantically to move up its book-to-bill ratio from a record low. And for five months, the ratio has been improving--very slowly. Until August, that is.

In August, the book-to-bill slipped for North American suppliers of chip production systems, moving down from 0.63 to 0.61--meaning that the industry was booking $63 worth of orders for every $100 of product shipped. The index hit a record low of 0.44 in April, according to the SEMI trade group.

The new numbers surprised some analysts, who were looking for the ratio to continue going up--or to as high as 0.81. But the new number continues to show a "leveling in the decline of orders and shipments seen in the past several months," insists SEMI CEO Stanley Myers. He is still worried, though. "The effects of recent current events and wavering consumer confidence and spending, however, could further exacerbate the already tenuous situation in regards to chip demand and capacity," Myers says. In other words, the outlook ain't all that good.

Worldwide bookings received by North American suppliers dropped 3% to $742.2 million in August. But orders were down a whopping 75% from the year-ago total, SEMI says. August shipments hit $1.21 billion, flat with July, while shipments ran 50% lower than the year-ago total of $2.4 billion.

(See Sept. 20 story.)

Contrarian Intel doubles its
investment in Shanghai plant

Intel execs--confirming that China is where the action is these days in the global chip business--dropped into Shanghai this week and told locals it will invest $302 million to expand its chip assembly and testing plant there.

The chip leader, one of the very few chip makers investing more money this year in capital spending, will more than double its investment in its plant in Shanghai's free-trade Waigaoqiao zone from an initial $198 million made in 1998 to $500 million.

The expansion will include a new final assembly line and test operations for Intel's new 845 chip sets, which are used with Pentium 4 processors.

(See Sept. 21 story.)

TSMC wants 'open' standard,
but won't include competitors

Taiwan Semiconductor Manufacturing wants to increase its grip on the global foundry business by setting up an "open" standard for the next process generation.

The No. 1 silicon foundry raised a stir in the industry this week by disclosing efforts to set a single, open standard for 0.10-micron process specifications. But these open specs probably won't be open to silicon foundry competitors.

For now, TSMC is trying to reach agreement with semiconductor companies that own and operate wafer fabs but also use third-party foundry services. "Whether or not this becomes something larger is still under discussion," a company spokesman says. "But I cannot imagine why we'd include other foundries. We are in a technology leadership position already and all we would be doing is helping others close that gap." Now that's being honest.

Three of these integrated devices manufacturers already have signed up by TSMC. "We have LSI Logic, National Semiconductor, and NEC . . . and we have other discussions underway as well," a spokesman says.

At least one of TSMC's foundry competitors had tried to standardize at the 0.10-micron (100-nm) technology node. Chartered Semiconductor Manufacturing, world's third largest pure-play silicon foundry supplier, says it floated a similar program a few years ago but the idea never got off the ground.

Common specs for baseline 0.10-micron CMOS digital processes could help reduce costs in design libraries and reusable intellectual property (IP) cores from third-party providers, says Chartered. "Tremendous variations still exist in processes at individual companies," notes John Martin, Chartered's chief technology officer.

Now's the time to do it, Martin says. "It will be just a little under 18 months before each of the three major foundries expect to have 0.10-micron technology in pilot production. We probably have enough time to bring that to together."

(See Sept. 21 story.)

This time it was a typhoon
that creamed chip gear biz

When things are going wrong, it seems like nothing goes right. Take the chip production equipment makers, for example. Struggling with a huge drop in business this year off 39%, so far, from last year they were gathering this past Sunday in Taipei for the Semicon Taiwan trade show.

But the worst typhoon in decades hit the island, forcing the cancellation of the first day of the show and cutting off power to many exhibitors until the last day.

Turnout for the SEMI trade group event, which usually packs the house during its three-day run, was as grim as the outlook is for the equipment industry is this year. With the world economy close to an official recession, there still is no light at the end of the tunnel. "The consensus is that we will reach a trough in the fourth quarter or later--but if the bottom is bathtub shaped, who's to say where the bottom is," says Elizabeth Schumann, SEMI's research director.

There have been some indications in the PC sector that the worst is over, she says, but in communications and networking "the bottom is not visible." That won't come until the middle of next year, Schumann believes.

SEMI is now in the process of revising its forecast for next year in light of the terrorist attacks on the World Trade Center and the Pentagon. The new numbers will be released in December. "It's too early to tell what the overall impact will be on the industry," says SEMI vice president Paul Davis. "Lots of people have been scrambling to get parts and supplies around but that seems to have gotten better now," he notes.

Another indication that tough times are still ahead is that IC unit volumes are now expected to decrease 15% in 2001. "This is very significant," Schumann says, because IC unit sales usually continue to grow even in a revenue recession. "In the last industry recession in 1998, IC unit sales actually grew by half of one percent, she says. There hasn't been a year since 1985 where IC units shipped actually declined from the prior year, she adds.

(See Sept. 19 story.)

Terrorist attack pushes recovery of
Japan's electronics biz to late 2002

The terrorist attack on the U.S. will push the financial recovery of Japan's electronics industry well into next year. It will "prolong the economic recovery," flatly predicts Yoichi Morishita, chairman of Matsushita Electric and acting head of the Japan Electronics and Information Technology Industries Association (JEITA).

"After such a horrible event, and under the circumstance that the word 'war' is used, personal consumption should decrease," he believes. And the "prospect of the recovery occurring in the middle of next year becomes more difficult."

For the first half of 2001, the Japanese electronics sector sold about $102 billion in goods--3.7% less than the year before, according to JEITA. Consumer electronics production fell 8.4% to $7.6 billion, industrial electronics production increased by 2.3% to $53 billion; and electronics devices dropped 9.5% to $42 billion.

Morishita notes that the Japanese trade group's projection early this year that electronics production in Japan would grow by 7% was "completely an illusion. The group now is expecting about a 5-to-6% drop from last year. But Japan "will be lucky if the fall is within that range, " he adds.

(See Sept. 18 story.)

Applied claims it can cut
time in half to install fab

Applied Materials is now pitching a new type of service to its chip-making customers that the equipment giant claims will slash the time from clean room prep to first silicon at a new 300-mm wafer fab by more than 50%.

Applied is trying to position itself between the fab owner and the various tool vendors and contractors needed to bring a new facility online, according to Wang. The integrated approach would put Applied -- not the fab owner -- in logistical control of seeing a fab through to fruition once the shell is complete.

"We have the framework and have tested it in-house at a small pilot line," says Jean Wang, product manager. "For first wafer out, there are a lot of things that we think we can improve to shorten it to less than half the time currently needed." The best ramp time up until now--from facilities prep to first silicon--is 69 days, Wang says, a record set by a Taiwanese client at a 300-mm wafer facility.

Applied's new service--called Ramp Performance Management--is still in the pilot stage but will be used by at least one U.S. client in the next several months, the company says. Applied will coordinate the prep work in clean rooms, such as laying gas lines and installing power sources, and ensure that equipment modules are installed in the most efficient order. One fab director in Taiwan already has expressed interest in the system, but Applied will have to overcome a few key hurdles first. Customers will need to be convinced that the company has the necessary experience to coordinate both the facilities and operations side of bringing a fab online. "Applied has a very good reputation with equipment, but clean room and facilities setup is a very different field of experience," says an insider at one Taiwan's major IC manufacturers.

Another problem will be to get competing tool vendors to accept Applied's supervisory role. Vendors may fear that Applied will learn too much about their systems, the insider says.

Applied is very optimistic about its new service. "We think companies will want to outsource this work," Wang says. "It will eliminate the wasted time and optimize the material flow. Some of the facilities will be able to be built at the same time as tool move-in, not as a traditional model in which it is done in a series and you have to wait until the facilities guys are done before you tell the tool guys to move in."

(See Sept. 19 story.)

Hynix hangs on
by its fingernails . . .

Cash-strapped Hynix Semiconductor is still hanging on by its fingernails. This week it failed to obtain new funding in a much-needed rescue package, but the world's third largest DRAM maker was still able to reach a compromise with its creditors and remain alive, according to reports out of Seoul.

Creditors had failed to agree on providing new credit and funding to Hynix last Saturday, "casting a shadow over its future," according to the Korea Herald. That's putting it mildly.

But all wasn't lost--yet. Hynix's 18 creditor banks agreed to swap Hynix's debt of 3 trillion won ($2.3 billion) for equity and roll over maturing debts, according to the reports. Talks may have continued this week over the rescue package, as observers believe the company's importance to its customer base--and the entire South Korean economy for that matter--will result in it eventually getting new credit and funds to keep it going.

The proposed new funding--Hynix's second rescue package in recent months--is critical if the company is to survive. Already saddled with debts of around 11 trillion won, the DRAM maker is racking up huge losses in the current downturn.

(See Sept. 16 story.)

. . . while rest of Korea's
IC industry suffers as well

Hynix Semiconductor isn't the only thing broken in the South Korean semiconductor industry. It appears that the nation's important chip export business is crashing. Blaming it on America's "war" with terrorist groups and the ongoing chip recession, the South Koreans are now talking about "further contraction" in the last three months of 2001.

Korea's IC industry--which is heavily dependent on the loss-ridden DRAM business--already was on the ropes before last week's terrorists attacks on the World Trade Center and the Pentagon. The Korea Semiconductor Industry Association now says the nation's total chip exports would decline by as much as 35% this year vs. 2000. Earlier this year, the trade group had predicted that South Korean chip exports would rise by 9.5% over last year.

(See Sept. 18 story.)

LSI Logic keeps on
cutting overhead . . .

It seems that more and more chip makers and production equipment suppliers are giving up on any market turnaround this year. LSI Logic, Applied Materials, and Philips all laid off more workers this direction this week.

LSI Logic didn't change its third-quarter forecast , but it decided this week to eliminate another 600 jobs, or 8% of its workforce, as part of its factory restructuring and the transfer of process R&D from Santa Clara, Calif., to Gresham, Ore. The Santa Clara R&D facility will be shutdown next month. Employment will fall to 6,900 employees after this new round of layoffs.

The company continues to see third-quarter revenues declining sequentially by 10-to-15% from $465 million in the second quarter. And it continues to anticipate a loss of 31 cents per share before charges and goodwill amortization.

In April, LSI Logic closed its 8-inch wafer fab in Colorado Springs, Colo., and cutting its workforce by 500 jobs, or by 7%, as part of a plan to consolidate manufacturing capacity at its two major sites in Gresham, Ore., and Tsukuba, Japan.

(See Sept. 19 story.)

. . . Applied lays off
2,000 more workers . . .

The equipment marketplace is not going as expected, so Applied Materials is laying off 2,000 employees--10% of its worldwide workforce. Some 700 jobs are being eliminated by Applied in Silicon Valley and another 500 positions are being cut at the company's operation in Austin, where it builds fab tools. The rest of the jobs being cut are located outside the U.S.

"We implemented a series of aggressive cost-reduction programs in an effort to avoid reductions in our work force," says CEO James Morgan. "Unfortunately," he says, "the length and severity of the downturn and lack of near-term visibility have made it necessary for us to take additional action to align our operations with the current level of business."

(See Sept. 20 story.)

. . . As Philips Semi takes
new $100 million charge

The deepening chip recession has forced Royal Philips Electronics to take another $100-million-plus charge for layoffs and restructuring of its semiconductor operations in the last half of 2001.

The Dutch company now expects semiconductor revenues to drop 17-to-18% sequentially in the third quarter from second-quarter chip sales of $1.04 billion.

"We will take charges for restructuring and headcount reduction costs in the third and fourth quarters of 20 million and 90 million euros, respectively, as we continue to reduce our cost base and rationalize capacity," explains Arthur van der Poel, member of Philips' board of management.

Even so, Van der Poel expects factory utilization rates in the fourth quarter to be higher than the current period.

(See Sept. 20 story.)

Intel speeds introduction
of 0.13-micron Prestonia

Intel is doing what General MacArthur did so successfully in World War II in the Pacific--island hopping. In the chip maker's case, it is skipping a process generation to get ahead of the competition.

This week Intel dropped the 0.18-micron version of its Xeon microprocessor line for dual-processor servers and accelerated the introduction of 0.13-micron chips targeted at the same computer segment. Making it possible is Intel's fast and successful ramp of its 0.13-micron technology, which is "is healthy and going well," the company says.

Instead of rolling out the 0.18-micron Xeon, the microprocessor giant is accelerating the introduction of its Prestonia chip series, which has a 2.2-gigahertz Xeon central processor with 512 kilobits of cache memory. Intel decided to scrap the 0.18-micron Xeon because its introduction fell too close to that of Prestonia's.

Also based on the Pentium 4 platform, the Prestonia central processor unit is a 2.2-GHz Xeon chip, with 512-kilobytes of cache memory. Geared for servers, Prestonia CPU will be rolled out in early first quarter of 2002, confirmed a spokesman for Intel. Originally, the processor was supposed to be out by the latter part of 1Q of 2002, the spokesman said.

Following up Prestonia will be the Gallatin, a 0.13-micron, Pentium-4 platform designed for use in multiprocessor applications. It should be out in the second half of 2002.

(See Sept. 19 story.)

Terrorist attacks don't change
Microchip's improving outlook

Microchip Technology isn't changing its forecast for the quarter ending Sept. 30 because of the uncertainties created by last week's terrorist attacks on the World Trade Center and the Pentagon.

"We believe that the rate of decline in distributor inventories observed at this stage should allow Microchip's channel inventory to be at or close to targeted levels by the end of the quarter," says CEO Steve Sanghi.

Microchip is anticipating sequential revenue growth in its microcontroller and analog product lines this quarter, he predicts. And he notes that the overall rate of design activities "continues to be positive."

Microchip is leaving unchanged its revenue estimate for its second fiscal quarter ending Sept. 30--or between $136 million and $141 million. This would result in something between a sequential decline of 2% to a rise of 1.5% from the $138.9 million in sales for the prior quarter.

For the second fiscal quarter ending Sept. 30, Microchip expects its gross margins to fall within its previous guidance of 50%. Earnings for the second fiscal quarter are projected to run between 15 cents to 17 cents a share, or flat with the prior quarter's 16 cents a share.

(See Sept. 17 story.)

Intel signals that it
still supports Rambus

Looks like Intel is going to keep supporting Rambus after all, despite the problems with the Rambus DRAM. This week the two companies signed a patent cross-license agreement that supersedes a prior pact.

Intel has been supporting Rambus, which has tried to establish its own high-speed Direct Rambus DRAM format for wide-bandwidth memories in competition with DDR. Questions have been raised over Intel's support of Rambus memories and the Los Altos company, but this pact appears to be aimed at squashing any such doubts.

The deal provides complete patent coverage to Intel for all patents and gives Rambus license coverage necessary for its high-speed interface businesses. Intel CEO Craig Barrett says he looks "forward to continued cooperation with Rambus in the further development of RDRAM-compatible chip sets and communications chips as well as the company's support of InfiniBand and future initiatives."

Last year Rambus launched a controversial patent-enforcement campaign aimed at collecting royalties for high-speed interface technologies in synchronous DRAMs and double-data rate (DDR) chips. Rambus also has been accused of trying to trick the industry by claiming exclusive rights to interface technologies that have been debated and adopted by industry standards groups. Rambus denies that it has done anything wrong.

(See Sept. 17 story.)

Can Motorola Semi
do what it promised?

Motorola Semi launched a major PR campaign this week to demonstrate it was deadly serious about marketing its cellular phone chips to the outside world and no longer restricting them to in-house products. Some observers, including me, question whether Moto can win over cell-phone competitors as chip customers.

Motorola has started talking about its first complete solution for 2.5-generation cellular phones, including a highly integrated four-component chip set that reduces electronics parts count in 2.5G handsets from more than 250 devices, including both passive and active parts, to as low as 125.

First details of the new platform come just nine weeks after the company said it would no longer protect any of its cellular phone technologies for use only in its own handsets. The strategy shift was questioned by industry observers because the plan hinges on Motorola's ability to win the confidence of cell-phone competitors to become dependent on its technology for new platforms for 2.5G and next-generation 3G handsets. Old hands in the business still question Motorola's promise to make this technology available to outside chip customers at the same time it is handed over to Motorola's own cellular phone division.

But Motorola is convinced it has a strong story here. "The market will be receiving a proven, mature software stack and customers won't have to worry about qualifying and testing it on their own," says Ed Valdez, a marketing director for Motorola Semi. "No other semiconductor supplier in the world--that we know of--makes as complete of an embedded solution and also ships phones for a living," he says.

I guess we'll just have to wait to the end of this year to see if Moto can prove its new strategy is working--that is announce design wins as promised.

(See Sept. 17 story.)

Is global chip business due
to sell more in China now?

I wish I had a dollar for every story I've written or edited on China's push to join the World Trade Organization (WTO). It's been a long 15 years of arguing and fighting, but it is finally happening. And this move just might be good news for entire global chip industry as well as for the Chinese.

This week the WTO put the finishing touches on a 900-page legal document for Beijing's formal acceptance into the group on Monday. China will become a full-fledged WTO member one month after its government officially accepts.

WTO membership is expected to open the Chinese market and liberalize its trade rules by gradually eliminating tariffs on semiconductors and other high-tech goods. The deal is also expected to transform the way that China deals with intellectual property rights through its implementation of the Trade-Related Aspects of Intellectual Property Rights accord.

It certainly could be good news for global chip vendors and chip equipment suppliers. The deal lays out a schedule for the eventual elimination of tariffs on semiconductors, computers, and software. China's 6%-to-10% tariffs on ICs are due to be eliminated in 2002, while tariffs that now run as high as 35% on semiconductor production equipment and materials should be eliminated by 2005.

"By opening, once and for all, the large Chinese market and subjecting China to the rules of the international trading system, China's WTO accession should create substantial new opportunities for our industry's exports and investments," declares William Archey, president of the AEA electronics trade group.

Should is the operative word. You just never know with the Chinese.

(See Sept. 17 story.)

IBM and Conexant duke it out
over who has best SiGe process

Look for a fight to begin in the next couple of months, when David--also known as Conexant Systems--takes on Goliath, better known as IBM, in silicon germanium (SiGe) process technology.

Conexant officials are not at all shy and claim their process is better than IBM's technology, and that includes IBM's groundbreaking transistor announced in June. IBM's device, a working chip with a transit frequency of 210 gigahertz that makes clock speeds of 100 gigahertz possible, could be ready for products as early as 2003.

But Conexant officials say those devices will be handicapped and that their own SiGe process, while producing more modest numbers, has a more realistic shot at the high end.

The two companies will collide head-on at an IEEE meeting in early October and then will present their future SiGe plans at the International Electronic Devices Meeting during the first week of December.

It's difficult to determine which process is faster. "There are so many subtleties . . . That's why I refuse to get into the numbers," says Bernard Meyerson, VP of IBM's communications research.

IBM measures speed by the highest wavelength at which a transistor can show a gain in current, but Conexant figures a more accurate measure is the highest wavelength for a power gain--power equaling current multiplied by voltage. But Theodore Zhu, Conexant's head of communications foundry services, contends that IBM's process has run out of steam. IBM's Meyerson disputes Conexant's numbers.

Zhu says it's easy to build one device with extremely high frequency the way IBM measures it, but Conexant's approach reflects more system-level conditions such as base resistance, base capacitance and system noise, making it a better measure of process speed. But IBM doesn't use Conexant's yardstick because it's wildly variable.

Fred Zieber, president of Pathfinder Research says both side are right about the shortcomings of both benchmarks. But he believes that IBM's 210-gigahertz SiGe is stronger than Conexant's Zhu infers. "IBM doesn't announce something until it's there or almost there."

(See Sept. 20 story.)

We welcome your feedback, comments, criticisms, or questions. E-mail us at bhenkel@aol.com. And remember: God bless America!

(Click here for last week's Semiconductor Alert!.)