The cost could be dropped that low, Halla claims, because the CMOS sensor chip "can already be produced at costs between $20 and $100 depending on the output resolution. Without explaining my math, I don't call any camera disposable when it has a retail price of more than $100.

Analysts worry over no growth
in chip gear biz they should be

We've got a growing group of observers who can't figure out what's happening in the chip production equipment business. Analysts are now saying they're puzzled by the lack of growth in new chip-making gear--especially in wafer-processing, frontend systems. The experts had been expecting a turnaround because of the optimistic comments now being made by such big suppliers as Applied Materials.

Analysts also are dismayed by the SEMI trade group revising recent order bookings downward when orders for January were essentially flat. The book-to-bill did rise a little in January--from 0.77 in December to 0.81--because billings dropped by 4.3% as bookings went up only 1.3%. This put sales of North American-based tool suppliers at their lowest monthly rate since the mid-90s.

I've always thought that 2002 was going to be a sluggish sales year for an industry famous for its roller-coaster sales.

The frontend tool segment was surprisingly weak, according to analysts, as January orders slipped 6.4% to $507 million from December. Overall equipment orders edged up to $637 million in January, but were still a whopping 66% below the year-ago monthly total of $1.85 billion.

The frontend book-to-bill was lower than the backend ratio in January, which was just the opposite of the way they were in last year's downturn and in late 2000. Frontend tool shipments dropped 5.4% in January to $649 million, while backend systems were up 1.3% to $135 million.

Backend assembly and chip-packaging tools in January showed bookings surging 50% to $130 million, up from a pitiful $87 million in December.

(See Feb. 20 story.)

TI starts paying more attention
to logic as market share dips . . .

For as long as I can remember, Texas Instruments has been known for its logic families. Product lines like TTL were cash cows for decades. But this traditional business slipped in recent years at the same time that the Dallas chip maker was concentrating on its more glamorous, faster growing digital signal processors and analog ICs.

For the second straight year, TI's logic business took it on the chin in 2001. And this is compelling the company to pay closer attention to this market as competitors begin to encroach on its market share. The chip giant has long led in logic sales, but now it is facing a growing challenge from Fairchild Semiconductor and perhaps even ON Semiconductor.

TI's logic revenue plummeted from $884 million in 2000 to $400 million last year. But the company is still strongly committed to logic, maintains Steven Hanke, marketing director for its Standard Linear and Logic business unit.

"TI supports almost every logic product line, and we have either market leadership or a very strong position in each," the marketer claims. TI has supplied logic devices for 25 years and currently offers about 20 families.

"The overall logic market took a beating in 2001 and will likely be about half of 2000, down from $3.2 billion to around $1.6 billion," he says. "It really hurt us, and we definitely lost share. It's definitely got the attention of TI."

. . . while Fairchild Semi
breathes down TI's neck

Texas Instruments owned 37% of the logic market in 2000, but is now losing market share as Fairchild Semiconductor, its main competitor, closes in.

Fairchild claims to have boosted its market share from 15% in 2000 to around 20% last year. "Our goal is to be number one in logic," declares Brent Rowe, Fairchild's logic market head. "When we broke out of National Semiconductor in 1997, we had about 13% share in logic, so the data would indicate we're on the right path."

But "you'd be hard pressed to find any other logic supplier that's investing as much in these product families" as TI is, according to David Hoover, head of logic products marketing. In fact, he says, TI is committed to "being the last logic supplier in each of these families."

But the logic market does seem headed for a shakeout. Both TI and Fairchild point to ON Semiconductor, which last year discontinued its LS logic family and has since been mired in financial difficulties. "I think there's going to be consolidation going on and both Fairchild and TI are the guys that are really putting some emphasis on this stuff and winning," claims Fairchild's Rowe

But ON Semiconductor isn't buying that. "We are in the logic business to stay," declares VP Charlotte Diener, who heads ON's standard component division. ON claims more than an 80% market share in ECL devices and is increasing its investment in mini-gates, VHC, and LCX, she says.

Logic is one market that requires significant investment to maintain individual families over decades, TI's Hoover points out. "I'm routinely asked how long is the average lifecycle of a product family, and my answer is always, 'When the first one dies out, I'll let you know.'" TTL has been around for 30 years, he says, and some TTL devices are "still some of the highest-volume products."

(See Feb. 18 story.)

Is Fairchild Semi leading
industry out of downturn?

Here's more optimistic news from Fairchild Semiconductor. The Maine-based chip maker has upgraded its outlook for the first quarter. Improving orders have changed its outlook to flat sales vs. the fourth quarter instead of a 3-to-5% drop it had previously expected. As a result, Fairchild this week canceled a planned temporary shutdown of its wafer fab in South Portland.

"Normally orders in the first quarter don't accelerate until mid-February, but we've actually had fairly good bookings since the third week of January," points out Fairchild CEO Kirk Pond. "Overall, our book-to-bill ratio continues to run well above 1:1."

"We're particularly encouraged," Pond says, "by the continued demand from the personal computing and consumer end markets, in what is usually a seasonally soft period for these segments. Demand from wireline communications, networking, and power supply segments have maintained their slow and steady rebound, while wireless handset demand has remained seasonally weak," he adds.

Fairchild reported fourth quarter revenues of $324.6 million, or flat with third quarter sales.

"Demand has been strongest for our power analog and power discrete products, especially our low voltage power MOSFETs in advanced surface mount packages," notes Pond. "Historically, demand for our newest products has rebounded first as we come out of industry downturns," he says. "The current cycle seems to be following this pattern. Order rates for our more mature standard logic, discrete, and standard linear products continue to lag our newer products."

(See Feb. 20 story.)

Xilinx is one more IC maker
to bump up its sales forecast

Another chip maker that sees its business improving is Xilinx, which this week bumped up its revenue estimates for the current quarter. The FPGA supplier now expects revenues to be 10% higher sequentially over the $229 million that it reported in the three months ended Dec. 31.

Earlier, Xilinx had expected sequential growth only in the low single-digit percentage range. Business is stronger than anticipated, the company says, because of the higher demand for Virtex-E and Virtex-II field-programmable gate arrays.

The San Jose company also predicts a couple more points of gross margin--it now expects to report a gross margin of 56%, up from its earlier estimate of 54-to-56%.

Also showing new strength, according to Xilinx, were market applications in storage and networking for enterprise and third-generation wireless systems.

(See Feb. 21 story.)

Game playing continues
in Micron-Hynix talks

The game playing was going on big time this week as the South Koreans tried to get out of their Hynix hole. But it still looks to me that Micron will end up buying Hynix Semiconductor's DRAM business. What this will mean, I suspect, is that Micron will become the world's No.1 DRAM supplier and DRAM capacity will edge into balance with demand with some memory producers actually even making money.

Right now, though, the deal is still stalled as negotiations have twisted and turned since early December. This week the South Korean government once again was trying to push the debt-ridden Hynix into a strategic marketing and development alliance with Samsung Electronics, the world's largest DRAM maker. But Samsung still isn't interested. Analysts view this as just one more of many attempts by the government, creditors, and Hynix to push Micron into an acceptable deal for Hynix and its lenders.

Hynix contributed to the hype this week by declaring that its board of directors was "seriously discussing a stand-alone business model with full support from creditors." Yeah, in your dreams.

Hynix creditors, owed more than $6 billion, are expected to meet this weekend to make a counter-offer to the U.S. DRAM supplier's $4 billion bid for what amounts to 70% of Hynix's business. The word is they plan to fly to Idaho as early as this weekend to make their counter offer to Micron. The Micron purchase has to be approved by the financial institutions that rescued Hynix with a series of multi-billion dollar bail-outs last year.

Micron wants to hear from Hynix and its creditors by next week on whether they will accept the $4 billion offer. Meanwhile, the spot price for 128 megabit DRAMs was $4 this week, up 16% in February and more than four times higher than the low last November.

(See Feb. 20 story.)

New European group working
on next-generation lithography

Like the cat with nine lives, conventional optical lithography continues to outlive all the forecasts on how long it could continue to build ever-denser chips. Now it looks like optical lithography will be good for at least the rest of this decade, which takes the pressure off the companies that are racing to develop NGL, or next-generation-lithography, tools. It also opens up the competition for other players.

Several NGL programs, in fact, will be discussed at the SPIE Microlithography conference next month in Santa Clara. And one of them will be a bold new effort called Extatic that's getting underway by group that includes ASML, Zeiss, Sagem, and Xenocs. The new European consortium is developing NGL tools and production systems based on extreme ultraviolet (EUV) technology, which is now favored to take over for optical lithography in chip making.

Extatic will address the main system aspects and the basic optical system for using extreme ultraviolet radiation for lithographic applications, according to an ASML official. Goal is to demonstrate that EUVL is the lithography technology of choice for the 50-nm node and beyond, she says.

That's not the only NGL effort in which ASML is involved. The Dutch company is expected to be the systems integrator for a U.S.-led consortium that's developing a EUV tool for turning out devices with feature sizes of 0.07-micron and below. It also will develop a commercial tool built around this EUV technology.

The U.S. consortium, called Extreme Ultraviolet LLC, revealed last year that it had developed and demonstrated the world's first exposure tool based on EUV. New optics and other upgrades to the Alpha tool will be described at next month's SPIE meeting.

Production-capable EUV tools are not expected to go on sale until 2007 or later, but some chip makers like Intel want to their hands on a EUV production tool before 2005.

While Japan's Canon and Nikon are each developing their own EUV tools, Nikon is making a stronger push in a competing technology called electron-beam projection lithography (EPL). It is now developing an EPL exposure tool for making ICs with feature sizes 0.07-micron and below. Nikon's alpha EPL system, called the EB stepper, should ship in late 2004. The Japanese lithography giant is building the tool platform, while its partner IBM is developing the optics.

(See Feb. 19 story.)

There's light at end of tunnel
for besieged ATE business . . .

Last year couldn't get much worse for the global automatic test equipment business. Sales fell by a staggering 64.2% from the previous year, or from $6.7 billion in 2000 all the way down to $2.4 billion in 2001, according to VLSI Research.

But believe it or not, optimism is growing about a turnaround. While "2001 was a tough year," says Jean-Luc Pelissier, vice president and head of Schlumberger Semiconductor Solutions is convinced that a new wave of chip designs will help pump new life into the besieged IC test market as early as this year.

Pelissier says that the downturn now appears to be at its end, and he predicts the shell-shocked ATE market will show some positive growth in 2002. "2003 and 2004 also look promising," he says. "A lot of devices will require new test capabilities and this is going to drive demand for ATE."

A new class of high-speed microprocessors, memories, chip sets, and graphics ICs will drive the ATE market beginning this year, he points out. Another emerging market for ATE vendors will be system-on-a-chip products and ICs based on emerging I/O standards, such as 3GIO and HyperTransport, he adds.

. . . but ATE vendors first will
have to reduce the cost of test

While the automatic test equipment business sees new opportunities beginning to resuscitate this market, it is also facing a host of new challenges. "The biggest challenge is to reduce the cost of test," points out Jean-Luc Pelissier, VP and head of Schlumberger Semiconductor Solutions.

One of the biggest challenges will be to test the new wave of communications devices, including wireline circuits for 10- and 40-gigabit-per-second applications, says Schlumberger VP Doug Cutsforth.

Intel, reportedly one of Schlumberger's biggest ATE customers, has declared war on chip-testing costs, which have surged 25 times for complex microprocessors and ICs in recent years. To combat this staggering cost increase, Intel is moving from a functional to a "distributed test" or structural approach to chip testing. Schlumberger rolled out its new low-cost structural tester last year, and one of the first customers reportedly was--you guessed it--Intel.

The San Jose test systems supplier is part of Schlumberger, the French oil field service giant, but last year the parent revealed plans to divest its ATE operations so that the company could focus on its core businesses. So far, the ATE unit has sold off two of its business units--a division making handlers and IC thermal conditioning systems to Cohu and its electron-beam wafer inspection business to Applied Materials. Planned for later this year is a spin-off of the test operation and other key groups into a new independent company.

(See Feb. 21 story.)

Here's powerful trio
building 300-mm fab . . .

Word was leaking out this week about another 300-mm wafer fab joint venture. The three companies involved here already have fab joint projects underway. Royal Philips Electronics and STMicroelectronics reportedly are planning to build a next-generation fab with foundry giant Taiwan Semiconductor Manufacturing. The announcement is due out in March.

TSMC and Philips Semiconductors already have formed a $1.2 billion foundry venture in Singapore, called Systems on Silicon Manufacturing. This venture opened its 200-mm fab last spring with tool sets and technology for volume production of system-on-chip products. It will initially employ 0.25- and 0.18-micron processes but expects to move to 0.15- and 0.12-micron technology later this year.

STMicro already has been working with Philips on 300-mm development and pilot production at its plant at Crolles, France

(See Feb. 21 story.)

. . . but these two
decide to split up

Something must have gone wrong bigtime between Hitachi and United Microelectronics.

The Japanese giant and Taiwan silicon foundry are discontinuing their 300-mm wafer fab, just two years after forming Trecenti Technologies to get a head-start in producing the new wafer generation. Hitachi will acquire UMC's 40% share in Trecenti and turn the big fab into a wholly-owned subsidiary.

Earlier in February, UMC officials were seeking to gain management control of Trecenti, but those efforts apparently failed. And there were earlier reports the two companies were negotiating a restructuring of the joint-venture 300-mm fab in Hitachinaka City, Japan.

Both companies claim the breakup was amicable and blame the chip downturn for their change in strategy to share 300-mm wafer production. Trecenti was described as the first 12-inch fab to start mass production using single-wafer processing tools.

Both of them say they will focus on separate 300-mm efforts. UMC, world's second largest pure-play foundry, will concentrate its 300-mm production at its Fab 12A facility in Hsinchu, Taiwan, and at its Singapore subsidiary, which houses its UMCi joint venture with Infineon Technologies and a planned joint-venture with Advanced Micro Devices.

(See Feb. 19 story.)

Can lightning strike twice
for Intel and Microsoft?

Can the "Wintel" team repeat its magic and dominate another major product family by setting standards like it did in the 1980s for the PC? I don't think so, but I'd never bet against them even this late in the game.

Intel and Microsoft apparently want to set platform standards for wireless handsets and personal digital assistants in the same way that they did with the PC. As part of a new joint initiative, the two giants will develop reference design platforms for next-generation PDAs and smart phones.

Both companies figure they can speed up market growth here because they believe the lack of industry standards in wireless communications systems is holding back the spread of smart phones and next-generation PDAs.

The deal, disclosed at the big cell-phone meeting in Cannes this week, calls for the reference designs to be based on the Microsoft Windows Powered Pocket PC and Smartphone 2002 software platforms and the Intel Personal Internet Client Architecture, known as Intel PCA. Microsoft will support the PCA as a "development blueprint" for building wireless handheld devices that combine voice communications and Internet-access capabilities.

Microsoft also is working with other chip suppliers for cellular phone platforms. The software giant and Texas Instruments has unveiled design platforms for smart phones based on TI's OMAP digital signal processor.

The Intel-Microsoft partnership intends to provide large and small manufacturers with a "foundation for quickly bringing powerful, Windows Powered mobile devices to market," says Microsoft vice president Ben Waldman. "We're striving to bring down the barriers to entry for device manufacturers and paving the way for a market explosion of innovative, smart devices at low cost."

The launch of the Wintel initiative comes as other chip makers and systems software suppliers team up to bundle reference designs for next-generation mobile phones and handheld terminals. Cellular phone giant Nokia and DSP leader TI disclosed their collaboration on "complete solutions" for 2.5- and 3G cellular phones that are based on the OMAP DSP engine, chip sets, and software from Nokia.

(See Feb. 19 story.)

Bell Labs laser produces
first broadband output

Here's the technology story of the week. Up until now, attempts to produce broadband laser output, a long sought goal, have failed. But Bell Labs scientists now claim to have developed the first reliable ultra-broadband semiconductor laser.

This development, which was disclosed in the latest issue of Nature, may open the door for new applications in optical communications, sensitive chemical detectors, and other future systems, researchers believe. The semiconductor laser is the first to emit light continuously and reliably over a broad spectrum of infrared wavelengths, according to Bell Labs scientists.

Current semiconductor lasers are compact, rugged, and often portable, but they are typically narrowband devices that emit light of a single color at a characteristic wavelength, according to researchers. Ultra-broadband lasers offer major advantages in that they allow for sampling of a wide swath of wavelengths at the same time, they point out. This opens the possibility of producing a semiconductor laser that's reliable and works under a wide variety of operating conditions, according to Bell Labs.

"An ultra-broadband semiconductor laser could be used to make an extremely sensitive and versatile detector that can detect minute traces of pollutants in the atmosphere," says Claire Gmachl, a Bell Labs physicist. "It could also be used to produce new medical diagnostic tools such as breath analyzers."

To fabricate the new laser, Bell Labs scientists stacked more than 650 different layers of standard semiconductor materials used in photonics on top of one another. The new device belongs to a class of high-performance semiconductor lasers called quantum cascade lasers, which were invented at Bell Labs in 1994.

The new devices belong to a class of high-performance semiconductor lasers called quantum cascade lasers, which were invented at Bell Labs in 1994. The QC laser operates much like an "electronic waterfall," with electrical current flowing through the device causing electrons to cascade down "an energy staircase," Bell Labs says. The ultra-broadband lasers emit 1.3 watts at peak power over the mid-infrared range of 6-to-8 micrometers, researchers say.

(See Feb. 21 story.)

Fabless ASIC startup
gets off to good start

Who says the chip startups are dead in the Valley? Take a look at eSilicon, a fabless supplier of ASICs. It got $6 million in funding this week from a knowledgeable bunch of investors that includes Fremont Ventures, Crosspoint Venture Partners, and Texas Pacific Group.

Recent customer announcements plus this new investment confirms that eSilicon "is charting a new course for the ASIC industry," brags Jack Harding, CEO of the startup. "We made remarkable strides in 2001, despite the worst semiconductor downturn in history," says Harding, who previously was CEO of Cadence Design Systems.

In January, eSilicon garnered 10 design wins, and sometime during the next quarter it will begin shipping its first chips in volume, according to the CEO. The company's first designs include a special delay-locked loop circuit for a digital camera designed for Eastman Kodak, an MP3 encoder/decoder for PortalPlayer, a home gateway device designed for 2Wire, and a 1 million-gate device for a set-top maker. Not bad.

(See Feb. 21 story.)

We want your feedback, comments, criticisms, or questions. E-mail us at bhenkel@aol.com.

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