But we were still worried. Our little chapel was a lot bigger than it looked, so we didn't think that our little stove would be able to keep the place warm on a zero degree day when a Nor'easter was blowing through. So we also got a portable propane heater and were ready for any howler. Alas, Dec.1st was no test. We had to turn off the portable heater and open the front door--the temperature that day hit 60F!

One way to fix DRAM biz?
Buy Hynix and close its fabs

For months, we've been talking shakeout in the collapsing DRAM business. Once again, there's too many producers and too many fabs. Nevertheless, we were shocked this week to learn that the only U.S. DRAM supplier, Micron Technology, was talking to Hynix Semiconductor about on "a broad array of strategic options" that could end up in a merger that would create the world's largest DRAM supplier. And this is happening just weeks after Micron threatened Hynix with dumping charges.

Within a month, we will know whether a merger is possible," declares Hynix CEO C.S. Park. A Micron official says, "Everything is on the table."

If it does happen, the alliance or merger would create the world's largest memory manufacturer and could reduce the supply in the bloated DRAM market. This dramatic turn of events leaves analysts and industry players wondering how Micron with its $1 billion plus war chest could benefit by investing in Hynix with its several billions of dollars of debt.

They'd better move fast if they want to make a deal. DRAM prices are down more than 80% in the past year, and DRAM revenue is expected to drop by more than 60% this year. Everyone is hemorrhaging.

What Micron CEO Steve Appleton is trying to do is eliminate a competitor," says Dan Hutcheson, president of VLSI Research. "Hynix has been his biggest problem by trashing prices all year because they are willing to sell below variable cost. They'll probably get it for a song and then shut it down," the market researcher says.

Another possible scenario is Micron may want to keep some of the newer Hynix fabs open. But "the No.1 challenge here would be getting those plants upgraded and making them cost competitive," points out Dan Heyler, head of Asia Pacific semiconductor research for Merrill Lynch. "On top of that is the balance sheet and debt issues," he adds. "It would be a challenging integration."

(See Dec. 3 story.)

Four straight months of
chip growth, but it's s-l-o-w

Well, the elevator may have stopped free-falling for the chip business, but it looks increasingly like a very wide, U-shaped bottom. No sharp V here. But global chip sales in October did increase gradually for the fourth straight month.

Chip sales grew a modest 2.5% sequentially to $10.4 billion from $10.18 billion in October, according to the Semiconductor Industry Association. But the industry is still in the toilet when sales are compared to year-ago totals. In October, chip revenues fell 44.1% from year-ago results.

But the SIA continues to expect a modest recovery starting in the fourth quarter. "October sales are another indication that the industry is on track to achieve our forecast of 4.7% growth in the fourth quarter," notes SIA president George Scalise. "All geographic markets with the exception of Japan had sales increases" in October. The U.S. led the way with 5.1% sequential growth.

"Recent data indicates inventory is now largely in balance and prices are rebounding in some product categories," Scalise says. In November, the SIA released its annual autumn forecast and it called for an industry turnaround beginning in the fourth quarter that would end up in industry growth of 6.3% next year.

The Asia Pacific region grew 4% sequentially in October, but chip sales were down 26.7% from year-ago totals. Sales in the Americas were up 3.1% sequentially in October, but were down 57.4% from October 2000. Chip revenues in Japan dropped 2.6% sequentially in October and were 46.1% lower than October a year ago. In Europe, chip sales grew 5.1% sequentially in October, but were off 46.1% from October last year.

(See Dec. 3 story.)

Intel, surprised by rising MPU
sales, doubles Pentium 4 output

I can't write a column this week without discussing Intel's little chat with analysts on Thursday. Turns out, the chip leader was a bit too pessimistic six weeks ago. So it has raised its fourth quarter sales forecast a tad from $6.2-to-$6.8 billion to $6.7-to-$6.9 billion.

Reason? Better-than-expected demand for microprocessors. In fact, the company is now running at capacity not only with the new Pentium 4 but also with its entire logic business. CFO Andy Bryant confirms industry reports of Pentium 4 shortages and points out that Intel expects to double production of its new processor in the fourth quarter over the previous three months.

Intel's new guidance would give it a sequential increase of 3% to 6% over third quarter revenues of $6.5 billion. The improved sales rate comes from seasonal growth patterns taking hold in key chip markets with economic conditions not undermining those sales as much as the company had expected. "The seasonal pattern has asserted itself," Bryant adds.

But only Intel's processor business is doing better than it had predicted earlier. The memory and communications chip businesses "are pretty much doing as we expected" a couple of months ago, Bryant says. Also unchanged is the Intel outlook for gross margins in the fourth quarter.

(See Dec. 6 story.)

SEMI predicts 2 down years
before chip gear upturn . . .

Here is another vote for a very slow recovery. No V-shaped business upturn not even a narrow U-shaped one. Instead, the semiconductor equipment industry turnaround will take two down-years to get started.

SEMI's latest consensus forecast, released in Tokyo at the Semicon Japan trade show, predicts that global chip gear sales will drop again next year before turning up strongly in 2003 and 2004. Sales in 2002 are expected to fall by 3%, after the falls 38% to $29.6 billion this year, the largest annual decline ever recorded in semiconductor capital spending.

The San Jose-based trade group figures the equipment market will grow by 29% in 2003 and 23% in 2004, but that is just like the U.S. Weather Service forecasting the weather beyond the next couple of days. Neither SEMI nor the weather service is likely to call it right. That certainly is the case up here in Maine, and I ought to know--I was a forecaster in the Navy during the Korean War. I do remember, however, that the pilots' hands shook when they handed me their clearance papers to sign before they took off for the China coast. Even if the equipment industry does have back-to-back years like SEMI is forecasting, total semiconductor capital spending in 2004 still won't make it back to last year's record total.

"While expectations are for the market to remain flat, to slightly down in 2002, the majority of industry participants believe that this quarter marks the cyclic trough for equipment orders," says Stanley Myers, SEMI's CEO. Yeah, sure.

Not all regions, however, are going to have falling sales next year. North American equipment sales will go up 2% to $8.5 billion and the rest-of-the-world, which includes fast-growing China, will increase by 6% to $3.5 billion, SEMI says. Europe and Japan are the trailing markets, with sales falling 13% in Japan to $7.1 billion and declining 4% to $3.87 billion in Europe.

(See Dec. 5 story.)

. . . but even SEMI's prediction
for 2002 is too bullish for some

I thought that SEMI's equipment outlook for next year is a bit bearish, particularly for the trade group, but some analysts believe the forecast is way off the mark. Its forecast "implies that 2002 is virtual mirror image of 2001," says Tim Arcuri, analyst for Deutsche Banc Alex. Brown. "We believe this implied slope of recovery is much too bullish and continue to maintain our view that equipment revenue is likely to decline on the order of 15-to-20% next year, with the year being extremely back-end loaded."

Playing a big role in this, he says, are lackluster fab-utilization rates and possible declines in capital spending for 2002. "The last time we had fab-utilization rates so low was in 1990," notes Arcuri.

"We need to see utilization rates at 70-to-75% before the next growth cycle kicks in again," he believes. "The earliest we could expect that is the end of next year." Until then, chip makers will be making just "technology buys" in new chip-equipment procurements, as opposed to the larger "capacity buys," he adds.

Dataquest is also more pessimistic about the 2002 chip gear market than SEMI is. The market researcher expects the semiconductor-equipment industry to show some improvement over 2001, but it still believes this market will decline from 8-to-13% in 2002.

(See Dec. 5 story.)

IBM claims to have solved
double-gate transistor design

Chip designers for some time have eyed the double-gate transistor structure for faster designs but weren't able to overcome such design problems as electrical leakage, high-energy demands, and poor electrical flow.

In such a design, the channel is surrounded by two gates that doubles control of the current. This makes possible significantly smaller, faster, lower-power circuits. The latest International Technology Roadmap for Semiconductors calls double-gate transistors a potential successor to today's CMOS transistor designs.

Now IBM Microelectronics says it can solve the design problems with its new double-gate transistor structure, which it claims already has the potential for actual use in chips. Some 20 technical papers were given by IBM on the new design at the International Electron Device Meeting in Washington, D.C., this week.

There's a big race going on here to come up with a commercial design--several chip makers are working hard on the next-generation transistor structure. Last week Intel showed off its new TeraHertz transistor, which is built on thin-layer silicon-on-insulator (SOI).

IBM says its double-gate transistor will carry twice the electrical current, operate at up to twice the speed, and will be well below today's conventional transistors in size. The breakthrough was made possible by a series of innovations in new device designs and materials, such as SOI, according to IBM.

"SOI is changing the rules in semiconductors," declares Bijan Davari, vice president of semiconductor development at IBM Microelectronics. "The basic transistor has largely gone unchanged for decades, but it has now been shrunk nearly to a point where it will cease to function," he says.

An improved transistor structure such as the double-gate design will be needed in five-to-10 years, IBM predicts.

(See Dec. 3 story.)

Intel, AMD battle it out
for fastest transistor title

Another chip maker is beating the public drums about its next-generation transistor design. Advanced Micro Devices claims it has broken Intel's unofficial record for the world's fastest transistor.

AMD says it has developed a CMOS-based, 15-nanometer transistor (gate length), which can handle switching speeds of 0.3-pico seconds--or 3.33 trillion switches per second. That would mean AMD has beaten Intel's short-lived record by a nose.

In November, Intel claimed to have broken its own record for the world's smallest and fastest transistor, announcing a 15-nm device that can handle switching speeds of 0.38-pico seconds--or 2.63 trillion switches per second.

AMD's 15-nm transistor is aimed at microprocessor designs in the 2009 time frame. By then, AMD claims it will be able to make 30-nm (0.03-micron) chips on 300-mm wafers.

(See Dec. 3 story.)

More heat and noise from
Bluetooth, but still no signal

More hoopla over Bluetooth this week, the over-hyped technology for low cost, wireless connections. After nine months of collaboration, Infineon Technologies and Toshiba breathlessly announced their Bluetooth chips--a radio-frequency transceiver and a baseband device--which are aimed at very high-volume, consumer product applications.

Problem is this technology ain't taking off until these chips get dirt cheap--under $10. But the two chip giants aren't talking pricing yet. There's nary a word either on standards and interference, the twin bugaboos still plaguing this technology. I don't want you to walk into the room and turn off my PC with your cell phone.

But that isn't stopping Infineon and Toshiba. They claim a complete system solution will be available in the first quarter for consumer products with Universal Serial Bus (USB) and Universal Asynchronous Receiver Transmitter (UART) functionality. Customers already have samples, the companies add.

"Our new PMB 6626 ... is a major step towards embedded Bluetooth applications," declares Christian Baumgart, Infineon vice president.

I dunno--I've been hearing about this miracle now for a couple of years. Maybe we'll learn more from the Bluetooth conference in San Francisco, but sources tell EE Times that the standard faces an uncertain future, with one of its key developments in turmoil.

(See Dec. 5 story.)

Curtain goes down on Avanti?
Synopsys wants to buy firm . . .

I often wondered how the Avanti mystery would play out. How the criminal conviction of this CAD software company and its top managers would affect customers, particularly when it seemed that they really liked Avanti products.

Well surprise, surprise. The Fremont, Calif., company is being acquired by Synopsys, which agreed this week to pay nearly $780 million in stock for the controversial company. If this amount holds, it would make the acquisition the largest ever in electronic design automation history.

The move, which shocked some industry observers, aims to combine front-end EDA tools, such as design synthesis, from Synopsys with back-end layout and routing software from Avanti. Synopsys CEO Aart de Geus says that many of his company's largest customers were pushing for the acquisition.

Shareholders of both companies have to pass on the deal, which could be completed in three to six months. At that time, Avanti will become a Synopsys subsidiary.

"The acquisition of Avanti and its key technology offerings ... allows us to assemble and deliver for our customers what we firmly believe is the best IC design flow in the industry," de Geus maintains.

The deal comes after a turbulent year at Avanti, which last summer was ordered by a criminal court to pay $195 million in restitution to Cadence Design Systems in connection with the company's conviction on software theft charges. In the criminal case, the court sentenced six employees--including chairman Gerald Hsu--in connection with the theft of EDA source code and material from rival Cadence.

Because Cadence has filed a civil suit against Avanti in connection with the allegation of source-code theft, Synopsys has obtained $500 million in insurance to deal with litigation as part of its plans to acquire Avanti. The premium for that policy was a whopping $335 million--if the ultimate liability and expenses of the litigation are less than $250 million, Synopsys will be refunded the difference. And if the deal doesn't go through, Avanti will have to pay Synopsys a $45 million termination fee.

(See Dec. 3 story.)

. . . but Cadence slams Synopsys
for buying 'seriously tainted' IP

It seems clear that the Synopsys deal to buy Avanti came as a big surprise to Cadence Design Systems. CEO Ray Bingham blasted Synopsys this week for buying "seriously tainted IP intellectual property."

Some analysts got the idea that a deal was in the works to resolve Cadence's civil suit against Avanti, but Bingham denied it and said that his company may seek well over a $1 billion in the case. "This is only for the tools we already know contain our IP," he says. "We have really only just started discovery on the follow-on products." The "incredibly high" premium paid by Synopsys for a $500 million insurance policy to cover litigation expenses, indicates the civil case would be very hard to defend, he notes.

Bingham says he "very disappointed that a company like Synopsys would take on the risks and get involved with what we believe is seriously tainted IP." He says that he also is disappointed that Synopsys "felt so compelled to do it they are going to pay Gerry Hsu something like $40 million." Such deals are fraught with pitfalls, Bingham says. "Customers and the FTC Federal Trade Commission will and should ask questions about this."

Bingham predicts Synopsys and Avanti will have difficulty merging their different corporate cultures.

(See Dec. 6 story.)

ASML's 248-nm KrF scanner
is first to work at 0.11-micron

ASM Lithography is stretching the 248-nm krypton fluoride (KrF) scanner. Its new Twinscan AT:580B is the first 248 nanometer, step-and-scan tool that's capable of 0.11-micron (110-nm) resolution on 300-mm wafers. The Dutch lithography giant got the improved resolution by developing projection optics with the highest numerical aperture for any KrF system.

"These new high numerical aperture systems are the first to enable the volume production of chips with 110-nm feature sizes using KrF technology," declares Martin van den Brink, EVP at ASML. "By extending production proven KrF technology, we are providing our customers with a powerful cost effective means to shrink existing chip designs and bring the most advanced chips to market faster."

First of the new scanners will ship this month to a major Asian customer.

The dual-stage scanner has a throughput of 95 twelve-inch (300-mm) wafers per hour--the highest in the industry for a 248-nm system, ASML says. The new platform features a dual-stage system, which aligns one wafer for exposure while another wafer is being scanned. It uses a Carl Zeiss Starlith 850 advanced projection optics.

(See Dec. 5 story.)

Fairchild Semi
expands in China . . .

China is still the semiconductor frontier as far as U.S. companies are concerned. Now it's Fairchild Semiconductor's turn. The Maine chip maker is putting up a new manufacturing and warehouse facility in China to expand its internal chip-assembly and testing capabilities.

The China plant is a "cornerstone of our growth strategy," says CEO Kirk Pond. "We plan to more than double sales in China by 2004," he predicts. "We're already the leading global supplier of power components and this plant in Suzhou will play a vital role in supplying our customers with multi-market semiconductors."

Total investment over the next five years will run about $200 million, according to the company. The 800,000-sq.-ft.-plant will be built in three phases for packaging and final testing of logic, discrete, and analog chips for both local customers and export to the rest of the world. It will be located at the Suzhou Industrial Park in the Jiangsu province. Currently, Fairchild outsources over half of its assembly and test volume work.

Early next year, construction will start on the first phase of the new facility--a $10 million, 120,000-sq.-ft., advanced backend facility.

(See Dec. 4 story.)

. . . but Chinese startup
delays big foundry plan

The double whammy of economic downturn and capacity glut apparently has given China's Beijing Huaxia Semiconductor Manufacturing (HSMC) cold feet--at least for now--and the new silicon foundry has quietly put off construction of its first wafer fab. The downturn has hurt all the foundry players, of course, even giants like TSMC, but startups like HSMC were counting heavily on a robust chip market to propel the need for new foundry providers in China.

Even China's own IC market is sagging now. Chip revenues in China are now projected by Dataquest, to decline by a surprising 18% this year from 2000.

But other foundry startups in China are pushing ahead. Semiconductor Manufacturing International recently has moved into production with its first fab, which is located in Shanghai. And Shanghai Grace Semiconductor Manufacturing, another foundry will begin production next year.

HSMC originally was scheduled to break ground for its first fab last spring, with a schedule calling for production to begin production by the third quarter of 2002. The 8-inch, 0.25-micron plant was said to be capable of turning out 45,000 wafers a month.

Now the Chinese foundry reportedly will not break ground on the plant until early 2002--a year's delay., HSMC reportedly has no idea when it will start up production with its proposed fab.

The Chinese start up originally had grandiose plans to become a foundry giant. It had aimed to put up four wafer fabs by 2010, including three 8-inch plants and one 300-mm facility. Indeed, the company originally was going to spend $10 billion for these factories. But HSMC, a joint venture between China steel giant Shougang Group, the Beijing municipal government, two U.S. design houses, Alpha and Omega Semiconductor, and Joshua Semiconductor, ain't talking right now.

(See Dec. 5 story.)

World's fastest SiGe
claimed by Conexant

A new silicon-germanium process developed by Conexant Systems promises a new generation of chips that will run at speeds of 80-gigabits-per-second and at 10 gigabits-per-second ultra-low-power devices.

The Newport Beach, Calif., company claims its new technology is the world's fastest silicon-germanium (SiGe) process--a claim that's hotly contested by IBM Corp., which pioneered SiGe for commercial use. The SiGe200 process is said to support transit frequencies (Ft) equal to 200 gigahertz and power maximum frequencies (Fmax) of 180 GHz.

"Increasing optical communications data rates are driving the need for higher performance process technology," points out Dwight W. Decker, CEO. Conexant's technology evolved from its 0.18-micron, BiCMOS manufacturing process and incorporates a proprietary bipolar transistor design that can be scaled down in size in order to boost overall performance levels in chip products.

Conexant is expected to start building products with the new SiGe process in early-2003.

(See Dec. 4 story.)

3-D chip is neat concept,
but is it a real product?

Back in the late '50s and early '60s, there was a lot of talk about three-dimensional solid-state circuits. The Air Force, if memory serves, sponsored R&D at Westinghouse Electric on something called operational functional blocks. They didn't go anywhere, but it made for fascinating news stories.

Now someone is promoting this concept again. Three-year-old Matrix Semiconductor claims it has made a "fundamental breakthrough" in three-dimensional chip design technology so that it will be able to build devices in the vertical dimension for greater integration and lower cost. It sounds just as good as it did coming from Westinghouse 40 years ago.

But Matrix claims to be close to commercial products--something Westinghouse never achieved. Indeed, Taiwan Semiconductor Manufacturing backed the start up, saying it has been collaborating with the tiny company and building Matrix 3-D memories with a product launch planned for next year. The first products will be nonvolatile memories, which are low enough in cost to be used in place of camera film or audio tape, according to Matrix. What is that old saying, if something sounds too good to be true....

Matrix claims it already has raised more than $80 million from Benchmark Capital, Skymoon Ventures, Microsoft, and several big consumer electronics manufacturers and has already proved that 3-D semiconductors are "both practical and highly manufacturable."

The 3-D memory will be an archival storage device designed to hold information securely for many decades, according to Matrix. Memory cards will be write-once, available in standard flash card form factors, interchangeable with existing flash cards, and compatible with existing memory standards.

(See Dec. 4 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!.)