Greetings from Down-East Maine, which just happens to be one of the last bastions of pure democracy. And the best remaining example of this is the annual New England "town meeting." Here, town residents gather to vote on every, and I mean every, budget line item. I thought I'd tell you about my town meeting because it's not only a moving experience, it's a heckuva lot of fun.
Our tiny village of Robbinston 450 souls is located on the Passamaquoddy Bay overlooking New Brunswick. It held its town meeting this week at the grade school gym to elect town officers and set this year's budget. It all started with a town supper--ham, turkey, several varieties of baked beans, dozens of side dishes, and a 10-foot long table loaded with two dozen kinds of homemade pies and cakes only $6 per person. The proceeds, which came to $1,600, originally were going to help pay for the 8th grade trip to Boston. But the students voted to give the money instead to a local family whose 15-year-old home burned to the ground earlier this week. Good kids, we'll find the money to send them on their class trip.
There wasn't much arguing or yelling at the town meeting this year. There were no contested elections for the three selectmen, the tax collector or the town clerk. In less than two hours, 120 residents 1/3rd of the town's adult population--not a bad turnout voted on 36 articles. Last year the town spent a total of $1.2 million, $750,000 of which paid for our local grade school. We got $460,000 from the state for education and $60,000 in excise taxes, meaning we had to raise $250,000 in real estate taxes.
I got involved in one warrant because I wanted to help our local volunteer fire department. As some of you may remember, despite my age I am an active member of the department. I had been upset since last year after the department decided to buy a 1983 Ford pumper as a second truck. But the firemen had to finance the purchase themselves by giving up their fire department stipend for two years. I thought that wasn't fair. But we needed this second pumper and it was in great shape, only 12,000 miles on it, and the price was just $11,000 A new one would cost $125,000!.
So this week the fire department asked for $20,750 for their 2002 budget, but they would still have to finish paying for the truck by again donating their $5,600 annual stipend. Now it's getting harder and harder to get people to volunteer because of the amount of training they have to dot. We're shorthanded; we only have 12 volunteers and last year, we put out seven forest fires, two chimney fires, two structure fires, and assisted in 10 automobile accidents and vehicle fires.
So I jumped up at the town meeting and made an impassioned speech saying it was unfair for the firemen to have to pay for the truck out of their own pockets. I made a motion to raise the fire department budget by $5,600 to pay off the truck loan. It was a 25% increase, but surprisingly the town passed the bigger department budget only a couple of old timers voted no. The fire chief, who usually works most weekends at the firehouse, was the most surprised the bigger budget passed. My good deed for the day.
Biggest budget item for the town was spending $367,250 to replace the roof and upgrade the structure of our 25-year-old grade school. It actually wasn't that big a pocketbook deal, since the state would pick up two thirds of the cost and the village has five years to pay off the remaining, no interest loan. Once our town conservatives found out it would only cost the average home owner about $25 a year, the warrant passed easily.
We also voted to spend $45,000 for snowplowing, sanding, and repairing town roads, $6,825 for ambulance service, and $1,300 for the street lights--all 15 of them. The town shelved a request for it to pave and maintain a road leading to our deep-water port--a "developer" wants to develop the site for bringing cargo in here. It would have cost us $15,000. So goes life on the Maine coast.
TSMC suddenly boosts
capital spending by $1B
This move is either a positive sign the global chip business is recovering or that the Taiwanese are real gamblers. I think it's a bit of both.
Taiwan Semiconductor Manufacturing's chairman Morris Chang surprised me and probably a lot of other industry watchers Thursday when he revealed his company is sharply increasing its capital spending this year.
TSMC, largest pure-play foundry in the world, is set to boost capital spending to $2.57 billion, 56% more than the $1.65 billion target forecast it made just months ago. This is even 15% more than the foundry spent in 2001.
Chang also told Taiwan's premier Yu Shyi-kun that TSMC would keep its promise to spend $20 billion to build five of 300-mm wafer fabs in Taiwan. But he didn't give a timetable. Chang met with the premier just days before Yu Shyi-kun is expected to announce his decision on lifting the ban on Taiwanese chip makers from building 200-mm wafer fabs in China.
TSMC and its biggest competitor, United Microelectronics, have been working hard to get Taiwan to ease its restriction so they can participate in a market that many see growing faster than any other over the next few years. "China will represent 20% of the global semiconductor market in the next 20 years," Chang says. "It's time to start getting foothold there."
TSMC and UMC already are feeling the competition from two Chinese startup fabs. "TSMC and UMC will do fine for the next one to two years," believes Alfred Yin, research head of BNP Paribas Peregrine Securities in Hong Kong. But "the longer it takes for the government to remove the ban, the less leading edge they will have when they actually invest there."
(See March 28 story.)
AMD hits top 10 IC makers
as STMicro jumps to No. 2 . . .
I'm always a sucker for census-like data on the semiconductor industry. Any industry analysis or ranking for that matter. iSuppli has just come up with a fresh set of numbers for last year and there were some surprises at least to me.
"The seismic forces that rocked the industry last year reshaped the competitive landscape, bringing major changes in the rankings of the leading semiconductor suppliers," pronounces Dale Ford, director of market intelligence services at iSuppli and coauthor of the report. "Only three of the top 10 semiconductor suppliers and just six of the top 30," he says, "retained the same ranking in 2001 that they had in 2000."
The research firm found that only 27 out of 172 companies covered in its study experienced revenue growth in 2001.
Only one semiconductor market segment grew out of 49 subcategories last year, iSuppli says, it was 32-bit microcontrollers. And only five of the 50 chip companies with revenues exceeding $500 million ended up growing in 2001. Just three of those five chip companies grew without having to rely on acquisitions, according to the El Segundo researcher.
By iSuppli's count, global semiconductor revenues dropped 31.7% in 2001 to $150.5 billion from $220.5 billion in 2000. The median "growth rate" in 2001 for the 172 companies covered by iSuppli was a minus 26.8%. Hardest-hit last year were the big DRAM vendors such as Toshiba and Samsung where chip revenues fell by more than 41%.
There were a couple of surprises in the top 10 rankings too. STMicroelectronics jumped over four companies to climb into the No. 2 position--NEC, Samsung, Texas Instruments, and Toshiba. Another winner in a bad year was Advanced Micro Devices, which moved from back in the pack in16th place last year into the No. 10 spot. It did this by holding its revenues losses to just 11.2%.
Micron Technology, which got killed by the DRAM debacle last year along with Toshiba and Samsung , fell out of the Top 10 all the way down to No. 18 in 2001. Intel kept its first place ranking, while the companies in the 2nd to 5th positions--Toshiba, TI and Samsung--all dropped one notch as STMicro flashed by. Motorola actually climbed one place to No. 6, NEC dropped two places to No. 7, while Infineon and Philips maintained their 8th and 9th place positions from the previous year.
(See March 27 story.)
. . . and U.S. increases its
IC market share to 53.3%
Three cheers for the red, white, and blue! U.S. chip makers were able to increase their share of the global semiconductor market in a down year despite the industry downturn.
U.S.-based chip suppliers grew their market share two full percentage points, from 51.3% in 2000 to 53.3% in 02001, according to a new study from iSuppli.
Also gaining market share (surprise) was the European chip suppliers that saw their market share climb from 10.8% in 2000 to an 11.8% share in 2001, according to iSuppli.
But chip makers based in Japan and in the Asia Pacific region were on the losing end of share in the global market. The Japanese share slipped from 27.9% in 2000 to 26.3% last year, and Asian Pacific vendors fell from 10% in 2000 to just 8.6% share in 2001, iSuppli says.
In iSuppli's Top 30 global chip suppliers last year, 14 of them were U.S.-based chip companies. One company really stood out. Nvidia climbed from No. 51 in 2000 to No. 30 last year, shooting up 81.2% from $713 million to $1.29 billion. Wow!
(See March 27 story.)
Is DRAM market morphing
away from commodity parts?
The prediction may be a little self-serving, but it does raise a highly interesting possibility.
The memory market is rapidly changing and is moving away from its lowly, commodity-product status, claims Jon Kang, senior vice president for product planning at Samsung Electronics' memory division. A new class of digital applications, he says, will "digest" and propel the need for next-generation devices.
"I think we will see a paradigm shift," Kang predicts. "Semiconductor memories have been looked at as commodities. But now, memories are becoming more divergent. I think we will see a lot more application-specific memories in the marketplace," he says.
In this new market, the personal computer industry will no longer be the key driver for DRAMs, Kang says. A new class of consumer, mobile and networking products will emerge and cause a "proliferation of new memory types" in the market, such as fast-cycle DRAMs and reduced-latency DRAMs. The shift in DRAM technologies and applications will cause what Kang calls a "digital digestion." For example, the speeds of networking backbones are moving from 2.5- to 10-gigabits-per-second and faster.
(See March 25 story.)
This new chip uses light
instead of wire to move data
Here is my pick for the hot technology story of the week. It's a fascinating new development that might actually do away with the connections that link a chip into a system. A team at Johns Hopkins University has developed a prototype chip that uses light instead of wires.
"We've developed a very fast and cost-effective way of getting data on and off a chip without using wire," says professor Andreas Andreou, laboratory director. "It really promises to revolutionize how computer systems for homes and businesses are put together." Pretty strong words for a scientist.
The researchers have added a layer of synthetic sapphire to a semiconductor wafer and bonded microdetecters on to the sapphire to handle the light signal. An optical receiver attached to the sapphire layer converts the light back into an electrical signal that's compatible with the chip's electrical circuits.
To beam out an electrical signal from the chip, the data is transformed into light by a tiny laser and sent through the sapphire substrate out to a high-speed optical receiver circuit that converts the stream of photons into a stream of electrons that can be understood by other system components.
By using optical instead of electrical signals for interchip communications, the researchers believe that a signal could move 100 times faster than it does along a metal wire. The optical interface circuits also need far less power because the sapphire substrate is an insulating material, not a semiconductor.
This kind of interface could significantly improve the internal transfer rate among chips in computers, the researchers say. The prototype chip can move data on and off the chip at 1 gigabit per second, they say. And they say that the next generation of the new chips will be able to run up to 5 gigabits per second.
(See March 25 story.)
Next-generation DDR-II set
to replace today's SDRAM . . .
The big thing in memory technology these days is the second-generation, double-data-rate (DDR) SDRAM, which is now being called DDR-II. Every memory maker is scrambling now to develop the new DRAM, and first parts are due to be sampled by late this year or early 2003.
This week for the first time, the first details and specs came out. The proposed DDR-II standard has been in the works for several months and is aimed at replacing the existing DDR SDRAMs, which are now being called DDR-I.
The mainstream memory market is expected to migrate from today's 200- and 266-megahertz DDR SDRAMs to the 333- and 400-megahertz DDR-I chips, then move on to the DDR-II, according to some memory makers.
The DDR-II market is expected to move into the mainstream by 2004, predicts Katsuyuki Sato, a deputy general manager at Japan's Elpida Memory, the joint DRAM venture between Hitachi and NEC.
The DDR-II specification should be finished up in March by the JEDEC Solid State Technology Association, formerly known as the Joint Electron Device Engineering Council. According to the JEDEC roadmap, DDR-II memories will come in 440-, 533-, and 667-MHz versions.
(See March 25 story.)
. . . and it may come faster,
eliminating the DDR-400 . . .
Some DRAM vendors don't believe there is a real market for the 400-megaherz version of DDR-I, which is called DDR-400. They expect the mainstream memory market to jump directly from the DDR-333 products to the new DDR-II.
In fact, the JEDEC Solid State Technology Association, "may or may not" ratify a much-anticipated standard for the 400-megahetz DDR-I SDRAMs, says Bill Gervasi, technology director at Transmeta and a JEDEC representative. He implies that DDR-II would become the next DRAM standard after DDR-333. "I don't see DDR-400 as a mainstream product," he maintains. "I see it as a boutique memory," where the DDR-400 would be geared for more specialized applications.
Other DRAM vendors, however, are bound to be unhappy over this lack of support. Micron Technology, Samsung Electronics, and other DRAM makers already are rolling out the first engineering samples of memories based on proposed DDR-400 standards.
Micron already has shipped its first samples of DDR-400 memories to Taiwan's Silicon Integrated Systems for qualification testing, and Samsung says it has begun delivering DDR-400 samples to several chip set makers.
For now, the 333-megahertz version (DDR-333) has been ratified by JEDEC and is expected to become the new, mainstream DRAM technology for high-volume PCs and other products, according to Bill Gervasi, technology director at Transmeta and a JEDEC representative. The DDR-333 is being pushing by several vendors, including Taiwan's Silicon Integrated Systems and Via Technologies.
But it will take a while to establish the DDR-333 market. The chip set rollouts are only in the early stages, while systems manufacturers seem to be reluctant to adopt the new memories over the lower cost DDR-266 parts.
(See March 25 story.)
. . .so 3 DRAM makers
won't offer DDR-400
Three major DRAM makers have decided not to market the 400-megahertz DDR-I--Elpida Memory, Hynix Semiconductor, and Infineon Technologies.
Samsung Electronics and Micron Technology have just decided to introduce DDR-400 chips with the current DDR-I architecture.
But trying to achieve 400-megahertz with all its high speed challenges in the existing DDR-I standard "doesn't make sense," says Joo Sun Choi, director of application marketing for Hynix Semiconductor America.
And even if DRAM makers can make 400-megahertz DDR-I devices, it isn't clear that module vendors or motherboard supplies would support such a memory, says Katsuyuki Sato, a deputy general manager for Elpida.
(See March 26 story.)
Xilinx test program reduces costs
by up to 80% for FPGAs in volume
Xilinx has come up with a new program that it says slashes the cost of FPGAs in volume production.
These so-called EasyPath devices are identical in every way to a FPGA prototype, the company maintains, but are tested by Xilinx to meet a customer's specific design specifications. Because they're much less expensive to manufacture, Xilinx claims EasyPath provides the customer with a total cost management solution with none of the problems of custom design conversion.
Making the program possible are increasing FPGA densities that make it possible to apply a customer test program at substantially reduced costs, the company claims.
"The gate array vendors only test the gates that have been routed, not every possible gate and line," says Babak Hedayati, senior director of product solutions marketing at Xilinx. "And we're not testing every single logic resource and interconnect. We're testing only the ones that are needed."
Under the new EasyPath program, Xilinx generates a custom test program for a customer's FPGA design. The company promises to deliver first silicon to the customer within two months after finishing the design. Minimum order is 5,000 devices for a non-recurring engineering fee of $150,000 to $300,000.
The customized test program reduces per-piece costs from 30-to-80%, according to Xilinx. By testing just one design, any defects outside the design no long matter and the wafer can yield more good die. The higher the device density, the higher the cost savings percentage, the company claims.
EasyPath is not an FPGA-to-ASIC conversion program, Xilinx says, because it does not involve reducing die size or mask programming. Xilinx will use the same mask set, process technology, and package as it would for a standard FPGA.
(See March 25 story.)
ICs pick up steam in China;
Advance Semi builds new fab . . .
China's Advanced Semiconductor Manufacturing is moving quickly from the aging world of 5- and 6-inch fabs to bigger wafers. In a major move to upgrade its foundry business, the Shanghai company is beginning construction this month on its first eight-inch fab.
One of the first chip makers in China, ASMC is a joint venture between the Chinese government and Holland's Philips Electronics. It plans to start production sometime next year.
The foundry says it will not compete head-on against the foundry giants from Singapore and Taiwan. Instead, it plans to focus on making analog chips and power ICs.
Up until now, ASMC had mainly built chips for Philips. But it recently signed a major foundry alliance with California Micro Devices under which it will make analog semiconductors and application specific integrated passive parts for the U.S. company.
(See March 26 story.)
. . . while Grace Semi's fab
okay for 2003 production . . .
Silicon foundry startup Grace Semiconductor Manufacturing originally planned to move into pilot production early this year with its new 8-inch fab. But last year after a series of delays, the Shanghai company pushed out the completion of the fab and its production start up to early 2003.
And this week Grace confirmed that it is still on schedule to begin production in March 2003.
Located in the Zhangjiang Hi-Tech Park in Shanghai's Pudong development zone, Grace is spending $1.6 billion fab for the 8-inch, 0.25-micron plant. Grace reportedly has licensed the sub-micron IC technology from Japan's Oki Electric Industry. Management of the well-connected company includes the sons of China President Jiang Zemin and Taiwanese plastics tycoon Y.C. Wang.
(See March 26 story.)
. . . and ChipMOS is building
Shanghai test, assembly plant
Shanghai is getting more chip production infrastructure. Taiwan's ChipMOS Technologies plans to build an IC test and assembly plant in the Qingpu Industrial Zone to serve the nation's growing manufacturing base for integrated circuits.
ChipMOS intends to set up two operational facilities in the industrial zone with full-scale production expected to begin by the third quarter of 2005.
Planned capacity for the backend assembly and testing facilities will be 50 million pieces per month, according to the company, which was set up in 1986 for Mosel Vitelic. In 1997, ChipMOS was spun out as a joint venture between Mosel Vitelic, Siliconware Precision Instruments, and several other companies.
This is the latest move by a Taiwan chip company to expand into China. China also has become a hotbed of activity for backend chip assemblers, which are now racing to set up and expand operations on the Mainland.
ChipMOS also will set up module and subsystem manufacturing at its Shanghai plant. The factory will be capable of producing finished memory modules, ASICs, liquid crystal modules (LCM), liquid crystal on silicon (LCOS) microdisplays, LCD driver ICs, and ink-jet print head modules.
(See March 27 story.)
ESEC dumps CEO; taking over
is ex-Siemens chip boss Knorr
Veteran European chip executive Jürgen Knorr has taken over as the new boss at ESEC Group, the struggling chip-assembly tool supplier in Cham, Switzerland. Head of Siemens Semiconductor in the '80s and '90s, Knorr has been a member of the ESEC board since 2000. He was also chairman of Europe's Medea R&D program, which was trying to boost the competitiveness of the European Semiconductor industry.
ESEC's board of directors replaced Felix Bagdasarjanz as CEO this week after a disagreement between him and the board over the strategies to turnaround the company. Knorr immediately took over as CEO.
Like most chip production equipment makers, ESEC had a lousy year in 2001. So it instituted a major restructuring, taking a special charge of 64 million Swiss francs and placing all of its operations into two divisions--die attach and wire bonder.
"Far-reaching measures" had to be taken, the company says, to increase its flexibility and to reduce its breakeven point. Even so, low visibility continues to hamper ESEC's ability to forecast a recovery in capital spending for backend chip equipment. The Swiss company "assumes" the bottom of the slump has been reached, but it still expects weakness to continue "far into 2002."
Its operating loss last year was 108 million Swiss francs ($65 million). Revenues fell to 189 Swisss francs ($114 million), down 64% from the record high of $320 million in 2000.
(See March 25 story.)
Infineon's all-out drive
to add DRAM capacity
Infineon Technologies is trying frantically to expand its DRAM base as this market begins to consolidate.
"Infineon is getting more anxious than anybody else in the DRAM industry" comments Helen Huang, analyst with ABN AMRO in Taipei. "It has to do something to expand market share and increase pricing power, and Taiwan is the right place to go because it's attractive as a partner."
Taiwan's Nanya Technology confirmed this week that it is now in talks with the German DRAM maker regarding "possible cooperation." It wouldn't elaborate, but Taipei's Economic Daily News reports that Infineon is teaming up with Nanya to build two 300mm wafer fabs to produce DRAMs in Taiwan.
Infineon's efforts come as Micron Technology gets closer to acquiring Hynix Semiconductor's DRAM fabs. Nanya is probably the island's biggest maker of double data rate (DDR) DRAMs.
If Micron is successful here, Micron would end up with a 35% share of the global DRAM market. That would pose a serious threat to its two closest DRAM rivals, Samsung Electronics and Infineon. The German company would face the biggest challenge since it has only about 13% of the market, while Samsung is the current global leader.
Earlier in March, Infineon signed a non-binding memorandum of understanding with Winbond and agreements with Mosel-Vitelic, with the goal of increasing its output by more than 20,000 wafers a month. Infineon already has started getting 48% of the output from ProMos, up from 38%.
Nanya plans to break ground on a 12-inch wafer fab in the second quarter with production slated to start in 2004. It now operates two 8-inch wafer fabs, giving it a monthly capacity of 63,000 wafers.
(See March 26 story.)
We want your feedback, comments, criticisms, or questions. E-mail us at bhenkel@aol.com.
(Click here for last week's Semiconductor Alert!.)
