My next door neighbors' 8th grader scored 7 points to lead his team. Counting the cheerleaders and both teams, 34 boys and girls participated, which is 60% of the kids going to our grade school! I'll keep you posted as to our progress.

Is reduced overcapacity, higher
prices ahead for DRAM business?

Maybe this year's DRAM market isn't going to be the chaotic place it was last year. Why? Fewer big players, for one thing.

Hynix Semiconductor and Micron Technology are close to combining their DRAM businesses, which would make the combined operation the world's largest DRAM supplier, nosing out Samsung Electronics. They apparently agreed this week on the price the U.S. DRAM supplier will pay for the South Korean firm's seven DRAM fabs.

The deal would allow Micron to cool down the hot competitive battle that has pushed DRAM prices to record lows during the past year and to reduce the huge overcapacity that hangs over the memory market. That could mean higher average-selling prices, and maybe even some profits for the bloodied DRAM business.

A Hynix spokesman told the South Korean press yesterday that the No. 2 and No. 3 DRAM suppliers have "narrowed pricing differences to around US$4 billion and are near conclusion. But we still need consent from our creditors and board of directors." That's what we figured all along would happen.

The two companies could sign a memorandum of understanding as early as next week. Under the deal, Micron would take over the Hynix workforce and its memory fabs for $4 billion. A preliminary agreement could be signed as early as next week, as Hynix's creditor banks meet to go over the tentative deal.

The U.S. chip maker had confirmed in January that its talks with Hynix had broken off and no further discussions were scheduled. Micron reportedly had offered $2.5-to-$3 billion for the fabs, but several of the largest Korean creditors of Hynix were holding out for more.

Then Hynix moved into discussions with Infineon Technologies over a possible alliance, but this week the German chip maker pulled out of these talks, because synergies between the two companies "were not easily reached and were only realizable with high risk." That probably means that Infineon wasn't willing to spend the money that Micron was offering to take over Hynix's memory business.

(See Feb. 13 story.)

Applied says chip gear market
coming back--but very slowly

Applied Materials, the global equipment giant, was talking a bit more positively this week. This year is still going to be a bad one for the chip production equipment business, but Applied believes that the industry may be on its way back up--albeit slowly.

"We are somewhat encouraged by orders for the new technologies," says CEO James C. Morgan. New orders totaled $1.12 billion in the quarter ended Jan. 27th, a slight increase over the $1.1 billion reported in the previous quarter. But these orders were still off 54% from the $2.43 billion in the year-ago quarter.

Sales are still falling--that's no surprise. Applied showed a 21% sequential drop to $1 billion in the quarter, and a 58% fall from the $2.36 billion reported in the year-ago three months.

The world's largest supplier of semiconductor equipment did manage to beat Wall Street estimates of on-going earnings, staying slightly profitable with a profit of $15 million instead of a predicted slight loss. But the company ended up with a quarterly net loss of $45 million after including a charge for eliminating 1,700 jobs and for acquiring in-process R&D.

Orders will increase sequentially by 10-to-15% and revenues will be flat in the current quarter ending in April, Applied predicts, but says it will take another quarter or two before it is clear that a sustained recovery is underway in the end markets for chips.

Global semiconductor capital spending will fall by 25% this year, Applied predicts, but will increase dramatically for the latest generation of equipment. Leading edge processes are now running at 95% of capacity or more--or essentially sold out. As a result, $10 billion will be spend on 300-mm wafer fab equipment in 2002, predicts CFO Joseph Bronson, up two-thirds from the $6 billion spent on 300-mm systems last year.

But it's going to be a while before Applied revenues get back up to the $3 billion range per quarter as they were before the current downturn. That won't happen until capacity "buys" start up again. As a rule of thumb, these tool purchases for volume production require a fab utilization rate of at least 80%, and the global industry is running now at only about 60%.

(See Feb. 12 story.)

Teradyne, Lam Research pushed
off Top 10 list of chip gear makers

VLSI Research came out this week with its annual Top 10 suppliers of chip production equipment and the list makes for interesting reading.

Surprisingly at least for me, lithography system suppliers "performed better than other equipment manufacturers in 2001," according to the San Jose market research firm. Their combined shipments decreased only 24% from 2000, which compare to a 57.3% collapse by test gear makers and a 34.7% fall by other wafer processing equipment manufacturers.

The automatic test equipment vendors got hammered the hardest. Boston-based Teradyne, for example, fell completely out of the Top 10, all the way down from its No. 4 spot last year. Japan's Advantest slipped from No.7 to No. 10 as its sales plunged 50.5% in 2001.

Being one of the 10 biggest suppliers didn't hurt though. Overall, the Top 10 semiconductor equipment companies did better than the entire industry of chip-making tool suppliers, according to VLSI Research. Revenues of the Top 10 dropped 31% in 2001, while the entire industry declined 38%, the researcher says.

Revenues of Applied Materials, again the largest supplier by far of semiconductor tools, declined 37.3%, which was the steepest decline outside the ATE vendors. Tokyo Electron repeated as No. 2 with a 30.8% drop to $3.56 billion, and lithography giant Nikon was again No. 3 with a 26.5% fall in sales.

KLA-Tencor had only a 13.7% decline in revenues last year--lowest rate of decline among the Top 10 and enough to move the metrology supplier from No.6 to No.4 in VLSI Research's Top 10 list.

Lithography supplier ASML held on to the No.5 spot, partly because of its acquisition of Silicon Valley Group. Its revenues, including SVG sales, dropped 23.5%. Canon, another lithography giant, climbed from No.6 to No.9 because its sales dropped only 20% last year.

Teradyne wasn't the only company to get pushed out of the Top 10 last year. Lam Research also fell out of the ranking--dropping from No.8 spot in 2000, says

(See Feb. 11 story.)

Surprising PDA shipments
shoot up 58% in 4th quarter

By last October, it looked as if the 2001 market for personal digital assistant market was going to be a loser. But in a surprising shift, PDA sales came roaring back in the fourth quarter to produce another record year.

For the year, an estimated 13.1 million PDAs were shipped by manufacturers worldwide last year, an 18% increase over the 10.08 million sold in 2000, according to preliminary figures from Dataquest.

"The PDA industry performed surprisingly well in the fourth quarter, considering the current economic conditions and the PDA market's weakness shown in the previous quarter," says Dataquest analyst Todd Kort.

PDA shipments in the fourth quarter soared 58% over unit volumes in the previous quarter, although the number of systems dropped 2% from the year-ago quarter. "Strong holiday season purchases were a big contributor to the sales upturn," Kort says.

But the market leaders Palm and Handspring saw their average selling price fall to $165, he notes, "as their price cuts continued and older inventory was pushed through the retail channel at low prices."

"Today, only about 20% of the market is willing to spend $400 or more on a PDA," Kort believes. "Microsoft, Intel, Compaq, and HP understand this," he says, and "they are also the only vendors making decent profits in this market today."

Palm, the PDA leader, saw its global market share decline from 50.4% in 2000 to 38.6% last year, Kort estimates, with total shipments dropping from 5.58 million PDAs to 5.06 million.

No. 2 supplier Handspring grew its market share slightly from 12.4% in 2000 to 12.6% last year. It boosted shipments from 1.37 million units in 2000 to 1.65 million units last year, according to Dataquest.

Compaq Computer's market share doubled last year, jumping from 4.2% in 2000 to 9.8%. Shipments moved up from 466,000 units to 1.28 million.

(See Feb. 13 story.)

Altera hopes new PLD design
will overtake Xilinx's Virtex

For the past year, Altera has been going all out to develop a new programmable-logic architecture that it's counting on to reclaim market share lost to archrival Xilinx's Virtex line. Its current Apex line of products has been losing key design wins to Xilinx, according to analysts.

Xilinx certainly has been getting high marks with Virtex. "It has proved to be a solid product that is good at integrating IP at a rapid rate," says Bryan Lewis, a Dataquest analyst. He also hears that "Virtex tools were superior--and that's what got designers interested."

But Altera claims its new architecture, called Stratix, will overcome earlier deficiencies and ultimately overtake its biggest rival. Indeed, CEO John Daane claims that Stratix-based products will generate $150 million in revenue next year.

"If he can pull that off, that would be a very quick ramp," comments Dataquest's Lewis. "By making that statement, Daane is betting that it will be one of the best--if not the best--ramps that Altera has ever had."

Stratix architecture aims to increase logic density and speed and still ease the design process so that designers can take advantage of more modular, reusable intellectual-property blocks. "No matter where you are on the chip it always has the same access to the logic elements, so it's friendly for block-based designs," Daane says. Altera also claims Stratix will provide 40% better performance than its current-generation devices, thanks largely to a fourfold increase in routing resources.

Stratix could also help Altera gain more design wins in 3G basestations, which often rely on programmable logic devices to do chip-rate processing in parallel. This is an area where Xilinx has an edge, according to Will Strauss, president of Forward Concepts. "Altera has a greater depth of DSP algorithms, cores and software than Xilinx, but Xilinx has a bigger share of the market because they were first."

(See Feb. 4 story.)

Advantest plans to grow in '02
even if ATE market doesn't . . .

Japan's Advantest figures 2002 will be another tough year for automatic test equipment and it is expecting no better than a flat market. However, the ATE giant expects to grow faster than the industry and gain market share in 2002.

Sales are predicted to grow 10% this year. "Inventories are going down and there is also a need for more technology buys in the market," says an optimistic Nicholas Konidaris, CEO of Advantest America.

After very little growth in the first half, the ATE industry "will see an acceleration of growth in the second half," he predicts. In the long term, ATE will continue to be a tough business, where competition among vendors is expected to be "fierce," he says.

The worldwide ATE market suffered its worst downturn in its history last year, falling a stunning 57% from $6.5 billion in 2000 to just $2.8 billion, according to Kondaris. The fall was so great, in fact, that he doesn't expect to see global ATE sales matching or exceeding the record 2001 revenues even by 2004.

(See Feb. 12 story.)

. . . as it opens U.S. design center
to make up for lost ground in SoC

The key to future growth at Advantest lies in the VLSI logic/system-on-chip market. But the Japanese ATE giant is currently "making up for lost ground from the 1990s, when it fell behind U.S. makers due to slow software development," according to Noriko Oki, analyst at Morgan Stanley Dean Witter's Tokyo office.

So Advantest decided late last year that it would open a design center to jump start overall efforts in ATE products as part of its bold-new "rebirth"" strategy. This week it revealed that it has launched the new SoC design center at its U.S. headquarters in Santa Clara, Calif., to define a new ATE architecture for a new line of testers for SoC products.

The company, which is the world's leading supplier of memory testers, is still behind Teradyne and other U.S. rivals in VLSI logic testers. "Advantest's medium-term goals are to maintain its current market share for memory test, where it holds 60% of the global market, while nearly doubling its share in the logic test market to 20%," explains analyst Noriko Oki at Morgan Stanley Dean Witter's Tokyo office.

"The company should have relatively little trouble in maintaining its market share in memory test, given the decline in surviving players in the DRAM industry," the analyst says. But "the logic ATE market is much larger than the memory market," he adds.

(See Feb. 12 story.)

Will Intel's Xscale
goose PDA market

Intel rolled out the first processor using its Xscale design, a microarchitecture touted for low power and high performance. The hope is that it will charge a PDA market that's now suffering from a lack of differentiation, poor wireless connectivity, low horsepower, and a dearth of true productivity applications.

The new Intel part, which builds on its StrongARM design and is backward compatible, features dynamic power and frequency management, and scaling.

"Instead of everyone being in the same range, Xscale will give you a wide range of processing power in PDAs," says Todd Kort, Dataquest analyst. "That will enable more powerful machines, maybe even able to do voice recognition."

Palm, which has 58% of the PDA market, is hampered by a severe lack of processing power, he notes. Built around Motorola's Dragonball processor, "the Palm only went from 16 to 33 megahertz over five years," he says. Palm has now moved to Texas Instrument's Omap architecture, but Kort believes that some of Palm's licensees, maybe Handspring or Sony, might opt for Xscale. "It now seems likely that Intel will be the major player in PDAs," Kort says.

PDA shipment growth has been slowing down, according to Neil Struthers, In-Stat analyst. "A little over 6.8 million PDAs shipped in 2000, vs. 8 million in 2001," he says. That is only a 17% growth rate, down from 36% earlier. Strothers expects growth this year to improve a bit to 18%, driven by better wireless connectivity such as with Bluetooth, better applications, and lower unit prices.

(See Feb. 11 story.)

GaAs chips due to be built
with new X-ray litho tool

Work continues on a novel X-ray lithography tool for producing gallium-arsenide devices. JMAR Technologies, a San Diego firm that's been developing this new technology, got another $34.5 million from Defense Advanced Research Projects Agency to deliver "with all diligence" its 1-nanometer, point-source X-ray lithography system.

JMAR figures that this money will enable it to "clearly demonstrate the superiority of XRL for producing high-performance compound semiconductors, such as those based on GaAs or gallium nitride, more rapidly and more cost-effectively than other lithography processes," claims CEO John S. Martinez.

JMAR plans this spring to combine the point-source system with an upgraded Model 5 stepper that it produces. The integrated system then will be installed at a GaAs microcircuit production facility, where it will be used for system demonstrations and manufacturing process evaluation.

Critical GaAs components, such as high-power monolithic microwave integrated circuits (MMICs), will probably be built with the new integrated system. It also will serve as a platform for developing future X-ray litho systems.

(See Feb. 12 story.)

More exciting stuff
from Carver Mead

I can remember the first time I interviewed Carver Mead. It was a warm spring day back in 1964 in Pasadena, when I parked my Triumph TR-4 at Cal Tech and walked to his classroom. He was already doing landmark research in integrated circuit design. Now here we are 38 years later and Carver is still doing awesome work.

This week his latest start-up company, Foveon, made headlines in many newspapers, including even the Bangor Daily News. Most of them "discovered" the new X3 sensor that had been developed for digital cameras, but the industry has known about the chip for more than two years. Now it's going into production at National Semiconductor, an investor, in its South Portland, Maine fab.

The new silicon image sensor, which is claimed to have resolution equal to film, is different from any current sensor in that each pixel captures all three primary colors by determining how deeply the photons penetrate the X3 sensor. Present CCD and CMOS sensors can detect only one color at each pixel, limiting the range and accuracy of the color captured.

The New York Times was lavish in its praise of the new sensor, but as neat a technology as the Foveon sensor is, National has to be able to mass produce it, and camera makers will have to be persuaded to switch to a new technology. That's never very easy, even with great new technology.

While the new sensor has the potential to reduce the complexity and cost of a high-quality digital camera, the first model using the Foveon sensor will come from Japan's Sigma, not known as a digital camera leader, and will be priced at about $3,000. I know there already are digital cameras on the market that cost that much, but I can't imagine spending that much for a camera. I guess I'll wait, even if the Foveon sensor can produce images that can be enlarged up to 30-inches wide, exceeding the quality of even 35-mm film. What ever happens, it is great new technology--congratulations Carver.

(See Feb. 12 story.)

Intel to launch server chip set
to revive languishing business

Intel will be pushing chip sets at the Intel Developer Forum, Feb. 25-to-28. It will introduce a new chip set for PC servers as part of a new effort to revive its languishing business in this sector.

The drive is seen as a direct challenge to the market leader, ServerWorks, a Broadcom subsidiary, and as opening up a new front in an interface battle between 3GIO and PCI-X.

Intel is expected to talk about its Plumas, a chip set for its Pentium 4 Xeon processors, which uses a 400-megahertz front-side bus. The set has a PCI-X bridge, links to an Infiniband bridge chip, and supports two processors and two banks of double-data-rate SDRAM. A low-cost version, Plumas LE, comes without the PCI-X bridge and supports only one bank of DDR memory. Both sets are slated to go into production late in the second quarter.

The chip giant may also detail its Cayuse follow-on chip set that will be Intel's first server core logic to support the 3GIO interconnect. Cayuse will support Xeon processors with a 533-megahertz front-side bus.

Analysts and OEMs alike tend to think that Intel will find its reentry into the server chip set market slow going. Although it already offers several server chip sets--including the NX that supports SDRAM and the 840 that supports Rambus DRAM, the company has not rolled out any products aimed at mainstream servers using DDR memory.

One reason why Intel is getting back into this business may be a desire to plow the way for 3GIO, a follow-on to the PCI bus that Intel is currently spearheading through a definition phase. ServerWorks, which claims 80% of the market for server core logic and boasts design wins in as many as 200 shipping servers, plans to support PCI-X 2.0 for upcoming 533-megahertz front-side-bus processors.

The computer industry is tending to divide between 3GIO and PCI-X. In fact, Compaq Computer says that it has no plans to support 3GIO in its servers and has led the charge to define PCI-X as a more evolutionary path.

"The test over the next six to 12 months will be whether Intel can regain customers in the server chip set business. I don't know if Intel will find it that easy to get back into this market," declares Nathan Brookwood, analyst with Insight64.

Intel's new play in this market "will be challenging for ServerWorks, but they have a good position as an incumbent," comments Kevin Krewell, analyst for The Microprocessor Report.

One OEM familiar with both companies' product plans believes the new Intel parts will not represent a strong technical challenge to ServerWorks. "ServerWorks has pretty comparable parts," he notes.

Raju Vegesna, founder of ServerWorks, shrugs off the Intel challenge. His company's proprietary interconnect, called IMB, has better throughput and latency than 3GIO and is already shipping in 3 million chip sets. "We will be here for a long time," he claims.

(See Feb. 13 story.)

Optical systems biz to grow 8%
this year after collapsing in '01

Two years ago, or back in the glory days, the booming market for optical transmission systems was expected to explode by 33% in 2001 because of the high-bandwidth Internet demand. But the horrible slump in communications markets and weak economic conditions pulled the rug out from under this market.

Instead of hitting another record, revenue for optical transmission systems fell 10% in 2001 to $30.3 billion, according to Dataquest. These sales are expected to turnaround this year--going up 7.6% to $32.6 billion, the research firm predicts.

Worldwide revenue won't even make it back to the record high that it hit in 2000 until next year. Dataquest predicts it will increase by 10.4% to $36 billion in 2003.

"While the market in 2003 will be similar in size to the market in 2000, its nature will have changed significantly," notes Dataquest analyst Peter Kjeldsen. "Consolidation in the market will have changed the number of vendors and operators, and their relative strengths are likely to have changed as well."

"Drivers for this change include a continued increase in the demand for bandwidth, introduction of an optical layer to handle DWDM dense wavelength division multiplexing channels intelligently and a continued migration of the fiber optic technology towards the edge of the network--most noticeably in metro networks," Kjeldsen says.

While North America is the dominant region for optical transmission system sales, the bright spot in regional sales was the Asia Pacific market, which grew 18.8% last year--thanks to a 27.3% rise in Chinese sales. "The Chinese market is the obvious attraction right now," Kjedsen says. "But because it is so obvious an opportunity, many players will try to address this market, so the competition will heat up."

(See Feb. 14 story.)

Oki beats Intel by 5 years
in using fully-depleted SOI

Here's my technology story of the week. While industry leader Intel plans to use depleted silicon-on-insulator designs in its processors in about five years, Oki Semicoductor has already beaten them to market with high-integration chips using fully depleted SOI.

One problem with the thin channels needed in fully depleted SOI is that they tend to force up the resistance through the source and drain areas, which in turn slows the device down. But Oki got around this by finding a way to build transistors on silicon layers just 50 nanometers thick with low-resistance source and drain areas.

The SOI wafers, which came from France's Soitec, were used to build devices for watches powered by solar cells. Power consumption in the chips has been cut by 75%, largely due to lower junction capacitance.

Oki started SOI production in 1998, which was based on a partially depleted technology and uses a thicker silicon layer under the transistor. In a year from now, the Japanese company plans to use the fully depleted process in future short-range RF devices, using the high-resistance substrate to create passive devices on-chip with higher Q factors than conventional silicon.

(See Feb. 15 story.)

Philips slashes 'uneconomical'
chip business, laying off 4,000

I've been wondering what's been happening in Philips' chip business, which tends to get buried in the Dutch electronic giant. Well I found out this week and it ain't good. "Several uneconomical semiconductor and component facilities were closed, or are in the process of being closed," the firm says.

Philips is shutting down a 6-inch, 0.7-micron bipolar and BiCMOS line in Caen, France, and moving an 8-inch 0.5-micron BiCMOS line from Albuquerque to Fishkill, N.Y.

Most of the 18,600 jobs Philips got rid of this past year were from the component, semiconductor, and consumer electronics operations. Some 4,000 workers, or 12% of its staff, were laid off in the semiconductor division, which lost money in the fourth quarter. For the year, the Dutch firm had sales of about $28 billion, down 15% from 2000 and a loss of about $2.6 billion.

Philips says the severe declines of the information technology and telecommunication markets have stopped now and excess inventories have mostly disappeared. But it adds that "we do not see markets strengthening," so the company is maintaining "a cautious stand on costs, capital spending, working capital, and employment in 2002."

(See Feb. 15 story.)

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

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