But I wouldn't mind a mild winter for a change, particularly since spring is still more than two months away. I also wouldn't mind seeing a good growth year for the semiconductor industry, but I'm convinced more than ever that it is going to be a slow recovery at best.

Researchers don't agree
on 2002 chip market ...

I admit my memory ain't what it used to be, but I can't remember when analysts and market research firms disagree so much on the semiconductor outlook for the coming year. This disparity certainly showed up in the numbers coming out of this week's annual Industry Strategy Symposium.

The gaggle of market researchers that showed up in Pebble Beach did agree on one thing--they all believe that IC markets had bottomed out and would start improving this year.

Let's start with the happy one. Veteran analyst Risto Puhakka of VLSI Research sees the worldwide IC market climbing a healthy 20.6% this year after last year's record decline of 31%. That's rebound I personally don't see.

Neither does analyst Bill McClean, president of IC Insights. VLSI Research is "looking for the market to skyrocket in 2002," he comments. "We don't see it that way." He expects the global chip business will grow by a miniscule 1% in 2002. He figures the market fell 32% last year.

"The recovery will be slow in 2002," agrees Moshe Handelsman, president of Advanced Forecasting and one of the more controversial chip forecasters. "We don't expect the recovery to be a 'V' shape," he adds. Advanced claims to be the world's only chip research house to have predicted the current downturn.

. . . or fab capacity utilization and
chip gear market for that matter

Talk about the 2002 outlook for fab capacity utilization and the semiconductor equipment market, and once again the big optimist at Pebble Beach was analyst Risto Puhakka of VLSI Research. He must have had a great holiday season. I did too, but it's hard to figure out where he's coming from here.

Puhakka expects the global fab utilization rate, which is now running at 73%, to rebound this year. Global chip assembly capacity is now running at 70%, he notes, while test floors are running at 73% capacity. "All three of these manufacturing areas should cross the 95% point in the fourth quarter this year," he predicted.

That was also too optimistic for analyst Bill McClean, president of IC Insights. "I don't agree that fab utilization rates are going to be 90%. It's not going to be that kind of year," he says. Instead, he predicts the global fab utilization rate will rise from the current level of 60% to 80-to-85% by the end of this year.

That outlook nearly matched what Dataquest is saying in its first-of-the-year forecast. "We're seeing fab utilization rates of about 60%,"says Klaus Dieter Rinnen, who heads Dataquest's semiconductor manufacturing research. "We will see that appreciate and reach 80% by the end of this year," he predicts.

But that kind of improvement in fab utilization will take a strong recovery. "Going from below 60% utilization in the third quarter of 2001 to 80% utilization requires a 30-to-33% increase in demand for silicon run rates," Rinnen says. The run rate will not have to go up quite that much, of course, if chip makers retire some of their older capacity.

VLSI Research's Puhakka believes his projected improvement in capacity utilization could provide a much-needed boost to the semiconductor equipment business, but "the recovery for equipment will be modest in the second half of 2002."

Overall, he sees the global chip equipment market declining 5% this year from 2001, when tool revenues plunged nearly 38%. That is still too optimistic for me and other market researchers--Dataquest predicts that spending for chip production systems this year will fall another 19%.

(See Jan. 7 story.)

Micron, Hynix get closer
to DRAM joint venture

Surprisingly at least to me, Micron Technology and Hynix Semiconductor seem close to making a deal. I would have given two-to-one odds it wouldn't happen after the U.S. DRAM maker grabbed Toshiba's Virginia fab in December.

But this week, Micron reportedly has proposed setting up a DRAM joint venture to take over up to seven Hynix Semiconductor memory fabs as well as taking a 20% minority interest in the remaining non-memory business of the Korean chip maker.

Micron is offering stock and possibly a little cash as its share in the joint venture and the minority stake in the rest of Hynix, according to reliable sources. That amounts to a $1.5-to-$2 billion deal, they say.

This value is based on Micron executives' statements that their offer is based on an earlier independent valuation by Arthur Anderson that Hynix currently is worth 25 cents on the dollar. Micron also says that its offer to Hynix is in line with what it will pay Toshiba for its Dominion Semiconductor fab in Manassas, Va. That "is a quarter of what it would cost us to build a similar new fab from the ground up," Micron executives say.

Hynix is expected to use the Micron stock to pay off much of its huge debt to domestic and foreign creditors. The lenders are expected to take the stock, not only because it is the best deal they can get, but also because they're counting on a big jump in Micron's stock price to recover money that they can't get from Hynix. There's precedent for that, of course. Both Intel and Texas Instruments, both of which ending up with Micron stock several years ago in separate deals, were able to double their money when their Micron shares subsequently shot up in price.

Before responding to Micron's offer, however, Hynix must sound out its creditors, many of which have conflicting interests. However, the sudden quiet that has descended over the negotiations has indicated to some observers that Hynix and its creditors are seriously considering the offer, and could respond shortly. Micron CEO Steve Appleton, who originally was schedule to return to the US. after making the offer, decided to extend his stay in Seoul, sources say.

The seven fabs that would become part of the proposed new joint venture are: a new leading-edge fab in Chongju; two older fabs in Chongju, two DRAM fabs in Ichon, the re-opened fab in Eugene, Ore., and a flash and SRAM fab in Ichon. That would leave six older Hynix fabs in Ichon, Chongju, and Gumi that have been, or are in the process of being converted to logic and foundry chip production.

Another surprise this week was Micron's reported offer to take a 20% minority stake in what would be left of Hynix after its DRAM business goes into the joint venture. But Micron has working for several years to expand its own logic chip business and participating here could aid its diversification efforts.

(See Jan. 9 story.)

North Carolina firm may move
gallium nitride into mainstream

Gallium nitride is being used to make blue light-emitting diodes and laser diodes, but defect problems have prevented it from moving into high-power or wireless circuits, where it potentially could set records.

But that could change. Developers at a tiny company in Raleigh, N.C., have come up with a way to grow high-purity gallium nitride on silicon that could be the breakthrough that will allow GaN transistors to play a mainstream role in the production of high-performance radio-frequency devices.

Nitronex, a three-year-old company started up by students at North Carolina State, is based on their graduate work in growing GaN on any substrate. This so-called Pendeo process is said to significantly reduce the number of defects found in GaN crystals. Using a new platform technology for growing GaN on 4- and 6-inch silicon wafers, the company is now building and sampling discrete power amplifiers for wireless base stations.

Commercializing the process meant starting from scratch on the process technology as well as having to build the equipment. "No one who had commercial GaN crystal growth equipment could grow GaN on 4-inch wafers, let alone silicon," says CEO Bob Lynch.

The advantage of GaN is that it can run extremely high amounts of power. The power density is at least six times that of silicon, so the same-size device in GaN produces six times the power of a silicon device. "Conversely," Lynch says, "you can make a much smaller device and get the equivalent amount of power out of it."

The Nitronex CEO is talking 100-watt or 200-watt devices. "The nice thing about GaN is that it has a great amount of power . . . It's a circuit designer's dream," he says. "People looked at GaN for years and said that if anyone could figure out how to grow this crystal . . . it would be the elixir for power devices in the RF space."

The initial product focus for the new process is a discrete power transistor based on a high-electron mobility transistor structure. The company has samples out to customers who are now going through reliability testing. It expects to generate serious revenue here by the end of this year.

Next step will be to move into other high-frequency RF products and more diverse applications. "We view gallium nitride as a platform semiconductor technology," Lynch says. "Besides RF power transistors for base stations," he adds, "there're families of products such as high-voltage switching for power grids."

(See Jan. 7 story.)

Bluetooth shows true colors
at Las Vegas: a niche product

Those folks flogging Bluetooth so hard in recent months should have gone to the Consumer Electronics Show in Vegas this week. They would have found what is looking more and more like a niche product for new PC peripheral and telematic applications. But certainly not a mass market technology.

Motorola introduced a Bluetooth-enabled hands-free car kit that includes a wireless headset interfaced with a handset that can process voice activated numbers. Palm and Compaq demonstrated Bluetooth-based handheld devices that interface with PCs and Bluetooth-enabled handsets, now offered by Nokia and Motorola.

What these applications demonstrate is that Bluetooth is finally settling into a comfortable market niche , comments Jack Quinn, analyst with Micrologic Research. "As to the claims that Bluetooth did not live up to its 'promise,' I think it would be more accurate to say that it did not live up to its early hype," he says. I would argue that it has not lived up to a lot of its current hype.

Quinn maintains that Bluetooth was never designed to be high-speed technology and was optimized for low cost and low power consumption. "It was never designed to be a computer network and to compete with 802.11," Quinn claims. "Bluetooth is designed to add wireless connectivity to devices such as PDAs whose batteries are too small to power an 802.11 node." Gosh, that's not what I've been hearing.

"Some proponents got caught up in the hype like everyone else did, and they extended their mission that exceeded the ability of the technology," acknowledges Kevin Duffy, vice president of Siemens' Home Networking Division in Austin.

(See Jan. 10 story.)

Automotive electronics hyped
at Consumer Electronics Show

Here's another "exploding" market for chips that has been hyped for as long as I can remember. In a desperate attempt to grow new markets, many of the exhibitors at this week's Consumer Electronics Show in Vegas were going after the largely untapped vehicle telematics market.

Makers of satellite radio systems, digital video disk players, rear-seat videos, navigation systems, and wireless transceivers are stepping up their efforts in the automotive arena. Analysts view the emphasis on automotive electronics as an "untapped frontier" and were saying that it "could signal the beginning of a new era." I think I first heard such hype back in the 1970s.

"Companies such as Intel, Hewlett-Packard, Sun, and Microsoft see automotive electronics as the one big market that has not yet been bombarded by PDAs, cell phones and video," gushes Thilo Koslowski, lead automotive analyst for Dataquest. "Some automakers believe it will blossom into a $10 billion to $20 billion market in the next four years, with virtually every car being equipped with a cell phone and many using in-car information and navigation services."

"This is good for the consumer electronics industry," Koslowski says. "The consumer home market is approaching some level of saturation, and the business market is still suffering from a poor economy. The vehicle is the only outpost left that's not approaching saturation."

Some observers even believe that telematics is so strong that it ultimately could become the tail that wags the automotive dog, with customers trading in their old vehicles in order to buy a new, telematics-equipped model. Give me a break.

(See Jan. 11 story.)

NEC's joint venture expects to be first
Chinese fab to build 018-micron chips

Denser, smaller-featured ICs may be coming from Chinese fabs faster than I had expected. The Shanghai-based joint foundry venture of Japan's NEC and China's Hua Hong expects to become China's first IC maker to make chips using 0.18-micron process technology.

Shanghai Hua Hong NEC plans to be using the technology at its 8-inch fab by June, according to Toshio Ohta, who runs the company. The company already is making chips using 0.35-to-0.25-micron processes for NEC and foundry customers.

Only one other Chinese chip maker has revealed plans to make 0.18-micron ICs. Singapore's Chartered Semiconductor Manufacturing is taking an equity stake in Semiconductor Manufacturing International (SMIC) in exchange for licensing its 0.18-micron technology to the silicon foundry startup. The schedule calls for 0.18-micron production to begin by the end of this year.

(See Jan. 7 story.)

How about a wafer fab that
will cost a cool $40 billion?

Long term forecasting is certainly fun, but lets face it: It's almost always off-the-mark. Would you believe that a panel of senior managers in the chip production equipment industry kicked around this week what wafer fabs would look like in 2025?

To begin with, the projected costs were staggering. A single wafer fab could cost an astounding $40 billion by then, predicts Kenneth Levy, chairman of KLA-Tencor. And based on current trends, he believes that a single lithography system could cost as much as $300 million. If that does happen, then there will be even fewer chip makers than there are car makers today.

As far as process technologies are concerned, Levy predicts that chip makers will be processing wafers with device geometries of only "10-to-20 angstroms" in 2025. And for memories, he can see "a terabit-on-a-chip."

Incredible numbers to be sure. But one prediction that I didn't find incredible at all is that the IC industry may be maturing and won't grow as much over the next 25 years as it did in the past 25 years. I don't think anything could.

The global semiconductor industry has grown 35% annually over the past 25 years--from $6 billion in 1975 to $210 billion in 2000. If the industry grew at that same rate over the next 23 years, the chip industry would hit an unbelievable $7 trillion in revenues by 2025.

But is it possible that the chip industry will reach that level of sales in 23 years? Of course not! It's highly unlikely, agrees James W. Bagley, CEO of Lam Research. "If the IC industry grows to $7 trillion by 2025, then the semiconductor business could be 25 times larger than the electronics industry," Bagley notes. "I don't think that will happen." Neither do I, of course.

(See Jan. 8 story.)

Motorola Semi moves quickly
to shut down four chip plants

This week Motorola Semi did what it said last month it was going to do--but it still comes as a bit of a shock. The chip giant told employees this week that it was closing its advanced IC assembly and test operation in Austin along with shutting down a backend chip-packaging plant and an eight-inch wafer fab in Sendai, Japan. It's all part of Motorola's "asset light" strategy to slash costs.

Earlier this week, the chip maker revealed it was closing its Hong Kong chip assembly and testing facility. China was always the crown jewel and represented the future at Motorola Semiconductor--or so I thought. The division had invested hundreds of millions of dollars there to achieve its strong position in that fast-growing market. This plant closing alone will result in the loss of 700 to 800 jobs. The company has run the chip assembly plant at Silicon Harbor outside Hong Kong for 33 years.

Under its new "asset light" strategy, Motorola Semi plans to outsource 50% of its CMOS products. The four chip plants in Texas, Japan, and Hong Kong are all that are scheduled to be closed in 2002. They represent about 2,000 of the 4,000 jobs being eliminated this year by the division, the other half will come from operations other than manufacturing.

A big chunk of Motorola's chip making capacity already is gone. It plans to operate just eight wafer fabs next year, down from nine today and 29 fabs in 1997. Now that's downsizing.

(See Jan. 10 story.)

Budget cut forces Sematech
to slash 10% of its workforce

International Sematech's 2002 budget was cut $4 million to $136 million from last year's $140 million, so the research consortium decided to take the hit by slashing 10% of its workforce, or about 60 full-time jobs.

The Austin operation will make this cut without cutting any R&D or sending any of the 120 assignees back to their member companies. Sounds to me like the little guys are the only ones getting hurt--or maybe the consortium was over-staffed to begin with at the clerical and support levels. Sematech revealed the staff cutback internally just before Christmas Why do companies always pick this terrible time to lay off people? and is still working on the cutback details. Employees to be laid off will be told in February.

The 13 member companies, who fund Sematech to identify and develop next-generation chip production technologies, were told that the consortium "intends to keep the fabric of Sematech in place," a spokesman tells SBN.

(See Jan. 8 story.)

TSMC wins round in
December from UMC

The fight between the world's two largest silicon foundries ended the year with United Microelectronics (UMC) losing the December round to Taiwan Semiconductor Manufacturing (TSMC).

TSMC reported its sixth sequential increase in monthly revenues as December sales jumped 6.1% from the previous month, while UMC sales fell 8.2% from November.

UMC had also been growing sequentially since July, but that string of month-to-month gains ended when December sales hit only $130 million, down from $141 million in November. For all of 2001, UMC revenues dropped nearly 39% to $1.9 billion from $3.1 billion in 2000.

TSMC sales in December hit $345 million vs. $325 million in November. For the year, TSMC revenues dropped 24% from $4.9 billion in 2000 to $3.7 billion.

(See Jan. 9 story.)

Hynix restarts its Oregon fab;
does this fit into Micron deal?

Here's what the world really needs--more DRAM capacity. And it's coming from Hynix Semiconductor, troubled South Korean supplier that's now talking to Micron Technology about merging some of their DRAM operations.

Hynix says it's starting up full production next week of leading-edge 256-megabit, synchronous DRAMs at its revamped fab in Eugene, Ore. The factory had been closed since last summer so its production equipment could be upgraded to 0.14-micron wafer processing.

Some of the 300 employees laid off when the fab was shut down temporarily will come back to work next week. A pilot line has been producing the 256-megabit SDRAMs, and has now been qualified to resume full production.

The Oregon fab reopening comes at a time when final negotiations are being held in South Korea between Hynix and Micron. The U.S. DRAM vendor reportedly is offering to take over as many as seven of Hynix's major DRAM fabs for a joint venture between the two firms.

The Eugene fab could be a complicating factor here, since it was financed separately with a $1 billion loan from a foreign bank group lead by Chase Manhattan and Societe Generale. This group may think differently about any merger than do the Korean domestic bank creditors, who are government-owned or controlled and may get government help on the side.

(See Jan. 9 story.)

Motorola will lay off
120 VP-level executives

You don't see this every day. Motorola is eliminating 20% of its 600 corporate officers--at the vice president level and above. Things must really be tough.

CEO Chris Galvin will make the final cut by identifying the 120 officers whose jobs will be eliminated. The chip division couldn't say exactly just how many of the company's corporate offices work there.

Until now, all of the company's layoffs had involved only lower-ranked employees, and top management, a spokesman says, felt "a clear responsibility to eliminate management layers."

Motorola expects to have 100,000 employees by the end of 2002, down from 150,000 at its peak in August 2000. These job cuts are continuing, with some 4,000 positions to be eliminated from the chip division this year.

(See Jan. 9 story.)

MEMC readies plant to begin
production of 300-mm wafers

This past year was a lousy one for blank wafer suppliers. MEMC Electronic Materials, for example, warned last summer that it was running out of cash. But a U.S. investment group saved the St. Peters, Mo., company by acquiring a 72% stake and restructuring its debt.

Now MEMC has completed the transfer of production technology for 300-mm blank silicon wafers to its first full-scale manufacturing site, located in Japan. "MEMC's 300-mm processes are sufficiently developed and ready to be implemented in a manufacturing environment," declares CEO Klaus von Horde. "MEMC's improved financial position enables us to increase 300-mm spending in 2002."

The first production site in Utsunomiya, Japan was established as a "center of excellence" for large-diameter substrates back in 1997 and initially provided 300-mm engineering samples to the industry. This factory has enough capacity to meet the industry's estimated demand for 300-mm wafers during the next several years, according to MEMC, the world's third largest supplier of silicon wafer materials.

(See Jan. 10 story.)

IDT closes obsolete fab, unifies
production in Hillsboro plant

Integrated Device Technology is closing its obsolete six-inch wafer fab in Salinas, Calif. The 17-year-old plant is running a 0.35 process and isn't able to support new product designs with feature sizes smaller than that, according to the company.

The shutdown will cut costs as well as accelerate the migration of IDT's products to more advanced process technologies. Salinas production will be moved to the company's five-year-old, eight-inch fab in Hillsboro, Ore., which this year will move from 0.18-micron processing to 0.14-micron production.

The Salinas plant will close in the second half of 2002, eliminating 260 jobs in manufacturing and support positions. This will reduce IDT's total head count by nearly 7%. But 90 people will continue to work in Salinas at IDT business groups handling SRAMs, multi-port devices, and Internet Protocol co-processors.

Consolidation of production at one facility is expected to improve IDT's financial performance this year without impacting its revenues, says CFO Alan Krock. "We also believe the non-recurring costs associated with the facility closure will be fully offset by gains realized by IDT from the sale of other investments during the quarter ended December 2001," he says.

(See Jan. 10 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!.)