P.S. Just hours after the announcement a couple of months ago, I wrote an indictment of the proposed Compaq/Hewlett-Packard merger. This week both founders' families surprised the two PC makers by saying they would vote against the deal. I think the merger is dead, which may not be the best thing for Compaq but it's certainly good for HP.

Now, even some bears
see 4th quarter bottom

Even some of the bears now sense that the semiconductor industry is beginning to pull out of its power dive. Advanced Forecasting, which arguably is one of the more bearish chip market researchers, sees growing indications that the market is bottoming out.

"For most of 2001, our IC Recovery Index depicted a severe and rapid decline in IC sales," notes David Crume, marketing head of Advanced. But now this index, wafer shipments, and DRAM average selling prices "have shown a much slower rate of descent," he says.

These new data points, just as they did in the previous industry recessions in 1996 and 1998, are signaling "an approaching minimum point" or bottom, Crume says.

The Cupertino research firm claims that its quantitative-based forecasting models have accurately predicted turning points in chip markets since the late 1980s, at a time when other analysts missed the mark. The firm says it had discounted signals from Intel and Applied Materials management who expected an industry bottom in the second quarter and a recovery in the third.

Advanced Forecasting maintains it has not changed its outlook since the beginning of 2001--it predicts IC revenues will reach bottom in the fourth quarter. "The events of Sept. 11th seem to have had a relatively limited impact on . . . our quantitative models," Crume says. While the market bottom could have been delayed by the terrorist attacks, he says his company is encouraged by what its quantitative models are now showing.

(See Nov. 6 story.)

Goldman Sachs: IC sales
may fall again in 2002 . . .

The global semiconductor market will fall again next year unless holiday demand for PCs turns out to be as strong as OEM orders have led suppliers to believe.

That's according to Goldman Sachs, which until Sept. 11th had been far more optimistic. It had figured that the release of Windows XP operating system and the holiday season would bring a nice rebound in PCs, and spur a gradual resumption of IC growth next year.

The terrorist attacks "took the wind out of my sails," comments analyst Terry Ragsdale, who now is "worried about PC sell-through in the fourth quarter" due to slumping consumer confidence and delayed corporate PC purchases. If there's only a slight uptick in fourth quarter IC sales followed by a typically flat first quarter, he figures "we're going to need double digit sequential growth Q2 through Q4 to get anything greater than zero growth for the year."

The New York investment firm is also bearish about the fab equipment market. Capital spending will fall another 15% in 2002 after a 30% slide this year, according to chip equipment analyst James Covello.

(See Nov. 8 story.)

. . . But happy-go-lucky SIA
sees global sales going up 6%

For as long as I can remember, analysts and the press have poked fun at the annual chip forecast by the Semiconductor Industry Association. Like most trade associations, they're almost always too bullish.

This year doesn't look any different. According to the new SIA industry consensus forecast, global chip revenues should begin a slow recovery in 2002, growing 6% to $150 billion, then shoot up 21% increase in 2003 to $181 billion, and 21% in 2004 to $218 billion.

The SIA badly miscalled this year's global IC revenues--predicting 21% growth really--but then blamed it on inventory corrections. The SIA's 2001 forecasts would have been closer to being right if the industry had not had the inventory build up that occurred in 2000, says George Scalise, SIA president. He figures inventory corrections should be completed by the end of the fourth quarter. The Sept. 11th terrorist attacks probably damped growth in the fourth quarter by one percentage point, he estimates.

The SIA was far more bullish just five months ago. The trade group predicted that global semiconductor revenues would fall only 14% in 2001, then jump 20.5% in 2002 and shoot up another 25% in 2003.

The new SIA forecast is similar to but a bit more optimistic than the worldwide consensus forecast just released by the World Semiconductor Trade Statistics group. The WSTS sees only a 2.6% increase in chip sales in 2002, 18% growth in 2003, and a 15% rise in 2004.

This year's 31% global debacle is not affecting all the world regions evenly. Hardest hit region is the Americas, which will dive 43%, while the world's other three regions will be down in chip sales by 23%-to-29%, the SIA predicts.

(See Nov. 7 story.)

Future Horizon expects
IC sales decline next year

Another market research firm has changed its mind about next year. London-based Future Horizons had expected the global chip industry to show a slow positive growth next year, but now it expects a 5.5% decline for the 2002 global market.

The researcher also caught up with other industry forecasters by downgrading its 2001 chip forecast, increasing the decline from 25.6% to 33.2% from the previous year.

With the 2002 market now pegged to be $129 billion, a level not seen since 1996, Future Horizons forecasts the IC industry will not now recover to its 2000 level, or $200 billion, until 2004.

"The events of Sept. 11th now mean Q4 will be flat on Q3's already lackluster performance," according to the Brits. "This will delay the market recovery by at least two to three quarters and negative dollar growth in 2002 is inevitable," Future Horizons says. This would be the first time in the history of the semiconductor industry that there were two consecutive negative growth years, according to the firm.

Future does predict modest IC unit growth of 6.9% in 2002, convinced that this year's excess inventory problem is no longer a major issue. But it believes this growth is not strong enough to counter the negative impact that overcapacity will have on average selling prices.

(See Nov. 7 story.)

SIA needs breaks to make
its five-year growth forecast

So what is going to have to happen before the SIA realizes its optimistic five-year forecast for the global chip business? Well, it's going to take a major revival in communications applications, which died in 2001, and solid increases in consumer applications, optical storage and networks, and automotive electronics. The SIA certainly is counting on a lot things falling in place.

Flash memories will be the fastest growing chip segment during this period, the SIA believes, with sales growing at a compound annual growth rate of 22% to $12.4 billion in 2004. Driving this segment will be renewed growth in cellular phones--2.5G handsets in particular--along with digital cameras and flash-based automotive applications, according to the trade group.

Unlike many forecasters, the SIA expects a big turnaround in the DRAM business. Since peaking at more than $40 billion in 1995, DRAM revenues are expected to drop 60% to $12 billion this year, according to the SIA. But the trade group is calling for strong growth in the next few years, with DRAM sales rising 44% in 2003 and 54% in 2004. Wow! That's hard to figure. Other forecasts are looking at flat to slightly lower DRAM sales in 2002 and no strong recovery likely for several years.

Driving DRAMs, according to the SIA, will be new consumer applications such as set-top boxes. It expects DRAMs to show a compound annual growth rate (CAGR) of 7.3% from $20 billion in 1999 to nearly $30 billion in 2004.

The industry isn't expected to get much help from one of the hottest growth drivers in the 1990s. The microprocessor will show a mediocre CAGR of just 2.9% between 1999 and 2004--from $27 billion to $31.4 billion, the SIA says.

Analog semiconductors should register a 10.4% growth rate between 1999 and 2004, rising from $22 billion to $36 billion because of growing communications applications, the SIA predicts. Second fastest IC growth rate will come from optoelectronic devices, which will rise from $6 billion in 1999 to $11 billion in 2004, a 13.7% CAGR. Fueling this segment will be such applications as optical storage systems and optical networks, according to the SIA.

Digital signal processors (DSPs) will also show a good CAGR--growing 13% from $4.4 billion in 1999 to $8.1 billion in 2004. These sales will be driven by wireless, wired communications, and a slew of emerging consumer electronics applications, according to the SIA.

(See Nov. 7 story.)

Dataquest reverses itself, now
sees China market off this year

Some smart market forecasters had believed China's booming chip market, called the industry's new frontier by some, would be immune from the plunging global chip market. In May, for example, Gartner Dataquest predicted that China's chip market would actually grow by 6% this year.

Why not? The China market had shot up 36% last year to $13.1 billion. But unexpectedly--at least to the analysts--the bottom fell out of the soaring China market. It was that old devil, falling prices. While unit shipments continued to soar in China, prices have fallen rapidly during the past few months and this has pushed overall chip revenues into negative territory, according to Dorothy Lai, Dataquest analyst who tracks the chip business in Hong Kong.

As a result, Dataquest has revised its forecast and projects chip sales in China to fall by a surprising 18% this year from 2000, she says. "The market is always growing in terms of units," Lai notes, "but chip prices have significantly fallen in all sectors." Prices for DRAMs and microprocessors are now taking a major beating in China, Dataquest says.

But Dataquest is going back to its bullish ways for China next year. "In 2002, we project single digit growth," Lai adds. "We're looking at 5-to-6% growth." That low rate of growth is still higher than Dataquest's global chip forecast, which it figures will go up 3%. And even that growth is optimistic, when compared with other forecasts.

(See Nov. 9 story.)

Big silicon foundries
signal business upturn

The silicon foundry business continues to pull itself out of the gutter. These fabs were running at as low as 25% of capacity earlier this year. Now it's beginning to look like the two biggest foundries have turned the corner and are headed back up the hill.

United Microelectronics, the world's second largest pure-play chip foundry, reported October sales jumping 12.9% over September's, the third sequential monthly increase in a row.

The No. 1 foundry, Taiwan Semiconductor Manufacturing, had already reported its fourth consecutive sequential monthly sales increase in October, growing 11.1% to $304 million.

But sales were still a far cry from year-ago totals. UMC, for example, was down 56% from year-ago sales. The current downturn is widely viewed as the worst setback for the pure-play silicon foundry business since it was started up in the late 1980s.

(See Nov. 9 story.)

FPGA synthesis tools
will ship in November

Field-programmable gate-array technology is smoking! The way it looks now, FPGAs with 50 million gates, up from 4-to-6 million gates today, will be available by the 2004-to-2005 timeframe.

"FPGAs are moving a lot faster than anyone would have ever expected" toward increasing speed and gate densities, declares Michael Bohn, chief scientist for Mentor Graphic's FPGA Synthesis Group. "What took ASICs about 15 years to cover, we're seeing FPGA technology advancing in just five."

This acceleration is a direct result of programmable logic suppliers having access to leading-edge processes at silicon foundries. Taiwan Semiconductor Manufacturing and United Microelectronics are both partnering with FPGA suppliers to use high-density programmable logic as process-learning curve vehicles, much like DRAMs were used in the 1980s and early 1990s.

Already, as high-density FPGAs become available, system designers are turning to programmable logic for such consumer products as MP3 players and video cameras. This is pressuring design tool suppliers to provide synthesis technology for FPGAs instead of using more traditional ASIC-based concepts.

So Mentor launched a major effort to redefine design software for FPGAs and will now ship in November its first beta versions of knowledge-based, heuristic synthesis tools targeted at next-generation FPGAs. Mentor believes that "synthesis is the lynchpin for FPGAs," Bohn says. "This will open up product development to the garage shops again that have found it more difficult to work with complex system-on-chip designs in the ASIC world," he predicts. "I expect to see a lot more innovation, new startups, and design concepts that were blocked by high costs."

(See Nov. 6 story.)

Is new kind of IC industry
coming over next 10 years?

The integrated circuit industry will be radically different in 40 years and will bear little resemblance to today's technology. That isn't too surprising since scientists say that what the semiconductor industry has accomplished up until now would have been unthinkable back in 1960. I agree. I was reporting on the semiconductor industry back then and one story I recall writing that year was the "transistor industry" would be maturing by 1970.

Thomas N. Theis, a director at IBM's Thomas J. Watson Research Center, laid out the next 40 years at the International CAD conference this week in San Jose. Beginning in about 10 years, he predicts, carbon nanotubes and various self-assembling, human-guided molecular structures will likely start replacing the silicon transistor.

Researchers are excited now about carbon nanotubes as the first replacement for silicon transistors. "Expose hydrocarbon gas to the right temperatures with the right catalyst and you will form these chicken-wire appearing structures," Theis says. "If you give that chicken wire a twist and join the ends, you have a semiconductor."

But these carbon nanotubes are just one of many areas of nanotechnology research that show promise for replacing the silicon transistor, he notes. Researchers are examining self-assembly, which basically tailors or guides atoms into growing into structures that can be used in semiconductors. Self-assembly technology now being researched, he says, includes nanocrystals, cubic superlattice, silicon-germanium quantum dots, and a one-layer-thick pentacene snowflake. What makes them a candidate for replacing the silicon transistor, he points out, is that they may prove that an atom structure can be encouraged to grow in particular configurations.

In 10 years, Theis believes that chemically synthesized nano-building blocks will replace semiconductor logic and memory devices, resulting in an increased need to emphasize redundancy, test and repair, and self-repair. And in 20 to 50 years, he adds, there should be pervasive use of self-assembly.

(See Nov. 6 story.)

Mark this down in
your 2005 calendar

The power semiconductor business will shoot up 50% over the next five years, thanks largely to an end market that's now in the pits. Thanks to cellular phones, these chips will climb from $17.7 billion in 2000 to $26.8 billion by 2005, according to Frost & Sullivan.

"Mobile communication devices provide a tremendous market for the power management and smart power integrated circuits," says analyst Sunderraju Ramachandran. "Power integrated circuits, with much smaller chip/board area requirement provide a natural choice for the portable communication applications such as mobile and cordless telephones."

This kind of long-range prediction is a tough one to make, but what it has going for it is that no one ever remembers the forecast in five years.

(See Nov. 6 story.)

New Hynix called
'unfair competition'

It ain't over, 'til it's over, as far as Hynix Semiconductor is concerned. Indeed, the South Korean government could end up being the controlling power behind the vastly changed DRAM supplier as a result of a $6 billion financial bailout package approved last week.

Several domestic, government-controlled Korean banks are lead institutions in a $2.3 billion debt-for- equity swap in the troubled chip maker. These banks, which hold nearly all of the Hynix debt, are part of an 18-bank group that could end up with stock ownership of 49% or more in the chip maker, far exceeding the 9% now held by former corporate parent Hyundai and its subsidiaries.

As far as its foreign DRAM rivals are concerned, the Hynix bailout smells and will result in unfair competition. Micron Technology is pushing on the U.S. government to protest to the World Trade Organization what it calls unfair government subsidies. The European Union has threatened to file a similar complaint with the WTO as well.

"This is an unfair government subsidization of Hynix to keep a bankrupt company alive that otherwise should not be allowed to exist," declares Jan du Preez, president of Infineon Technologies America. "It distorts the market and puts the other largest DRAM manufacturers at an extreme disadvantage."

But Farhad Tabrizi, Hynix marketing vice president, plays down the role of government-owned banks in the rescue package. Despite their potentially dominant stake in Hynix, the banks have said they will not seek a management role in the company, he claims.

Five of the banks in the equity swap, which hold about 75% of the Hynix debt, have close ties to the Korean government. Because of this, Micron claims these banks are acting as a conduit to funnel taxpayer funds to Hynix. The Idaho chip maker "is watching very closely to see if the big debt-for-equity swap results in the government gaining a controlling voice in Hynix through the banks it also controls."

(See Nov. 5 story.)

We welcome your feedback, comments, criticisms, or questions. E-mail us at bhenkel@aol.com. And remember: God bless America!