Commentary & analysis of week's chip news
Greetings from Down-East Maine. Yep, it's even been hot in Maine, just like the rest of the Eastern U.S. But down here on the beach, the temp usually runs 10F-to-15F cooler, thanks to the afternoon sea breezes. Yesterday, for example, we hit a high of 70F, and today it maxed at 78F. Our problem is rain--we need some. The lawns here look as brown as Arizona's. But the two streams that pass through our fields on the way to the sea are still running strong. And our artesian well never dries up--water out of the tap is still ice cold
It's getting close to the time when some optimistic chip makers and analysts expect sales to turn up. Are July foundry sales the first indicator? I still don't see any real turnaround until later next year--same prediction I've made for the past eight months. Remember I'm talking about orders, not semiconductor stock prices, which don't seem to bear any semblance to reality.
TSMC gain may mean
fall in foundry sales is over
Is the worst plunge ever in foundry revenues finally starting to pull up?
For the first time since early this year, revenue grew on a month-to-month basis at the world's largest foundry. It ain't much, but Taiwan Semiconductor Manufacturing Co. showed a 1% increase in July sales to $249 million--up from $246 million in June.
While sales at United Microelectronics are still dropping, the decline is slowing. July sales at the No. 2 pure-play foundry dropped 8.6% to $113 million from the previous month; in June sales fell by 14.4%.
These latest results support comments by some market observers that foundry revenues have finally hit the bottom in the current slump. Foundry fab utilization rates are now running in the 25-to-35% range in this quarter, and Dataquest figures they will gradually start recovering now.
But the market researcher believes that fab utilization by foundries will only make it to the 60% rate by the second quarter of next year. Dataquest's latest forecast says global foundry revenues will drop more than 25% to about $10.1 billion this year, down more than 25% from last year's $13.7 billion.
TSMC revenues apparently hit the bottom in the current downturn in May and June, a spokesman says. Its book-to-bill ratio (new orders vs. processed wafer shipments) moved above 1.0 in July, reinforcing the company's belief that business will steadily improve during the second half.
(See Aug. 9 story.)
A shortage to worry about now?
Not enough 200-mm substrates
While most folks in the industry fret over chip order rates and when they'll turn up again, some people are actually worried now about a potentially big problem that could pop up as a result of a recovery.
The problem is whether chip makers can get enough blank 200-mm silicon wafers when demand picks up. Until recently, market planners were counting on the initial build up of 300-mm wafers to take care of soaring demand. But now that scenario looks dicey as chip makers push out startup dates for their multi-billion dollar 300-mm fabs.
Dataquest figures that demand for the 8-inch substrates could surge unexpectedly at the same time that 300-mm wafer shipments are ramping at a slower rate than chip makers had predicted. The market researcher's leading scenario now is that demand for 200-mm blanks will reach a peak volume of from 7.5-to-7.8 million substrates a month in the next recovery cycle, says analyst Takashi Ogawa.
It appears that suppliers of blank silicon wafers appear to have the ability to deliver 6-to-6.5 million substrates per month, if they increase output by making minor improvements in their plant efficiencies, he says. That forecast assumes that no more major 200-mm wafer fabs will be built, but the delay in starting up the next-generation 300-mm fabs could mean that new 200-mm frontend lines will have to be built.
Dataquest believes that wafer substrate suppliers will face a difficult decision shortly: Do they start setting up 300-mm substrate production lines or do they add new 200-mm capacity?
The San Jose market researcher is now saying that there's a strong chance that a shortage of 200-mm wafers will hit in the 2004-2005 timeframe because of the lack of blank substrate capacity.
(See Aug. 7 story.)
Blank wafer vendors put
investment plans on hold . . .
One big reason why it might be difficult for silicon substrate manufacturers to respond to a market upturn is the collapse of their business this year.
Global shipments of blank silicon wafers for chip production fell 21% in the second quarter from the previous three months and 28% from the same period last year, according to the SEMI trade group.
"Silicon demand has nose-dived this year, causing the silicon suppliers to implement cost-cutting in the form of layoffs or frequent plant shut-downs," says Stanley Myers, CEO of SEMI.
As a result, he says, "virtually all 200-mm investment plans have been put on hold." Myers says that "the question remains as to how quickly silicon suppliers could respond to a demand increase."
And the substrate suppliers' current financial problems "threaten the industry's ability to provide next-generation 300-mm diameter products in quantity and quality as required by the market," he adds.
(See Aug. 7 story.)
. . . as MEMC now worries
about running out of cash
The nose-dive in the blank wafer market has hit some suppliers especially hard. Take MEMC Electronic Materials in St. Peters, Mo., for example. It is now saying that it could run out of cash by the end of September, as it struggles with mounting losses.
This week the third-largest supplier of blank production wafers posted an $88.1 million net loss in the second quarter, including $22.3 million in restructuring charges. This came as the supplier reported a 29% sequential drop in quarterly net sales to $157 million. The work force has been cut 20% so far this year.
"We are working closely with potential financing sources to address our liquidity and cash needs," says Klaus von Horde, CFO of MEMC. The company now believes that cash generated from operations--together with the liquidity provided by existing cash balances and credit facilities--will be sufficient to satisfy commitments for capital expenditures, repayment of current maturities of long-term debt, and other cash requirements only through the third quarter.
(See Aug. 10 story.)
Tseng elected deputy CEO
at TSMC--Chang's successor?
Morris Chang seems almost irreplaceable. I regard the CEO of Taiwan Semiconductor Manufacturing Co. as the father of the Taiwanese semiconductor industry and the innovator who pioneered the silicon foundry business. I first met him more than 25 years ago at Texas Instruments where he was a vice president.
But the time always comes when a strong corporate leader must stand aside for his successor. And the first steps along that path may be happening at TSMC. New top management changes this week included the possible heir apparent to Chang.
F. C. Tseng was named to the new post of deputy chief executive of the Hsinchu-based foundry giant, while Rick Tsai, marketing EVP, was elected president and COO succeeding Tseng. Both will continue to report to Chang.
The moves signal to analysts that Chang will hand more responsibilities over to senior executives--especially Tseng. Tseng, a 14-year TSMC veteran, has run the company's day-to-day operations for the past three years. Now he will start looking at longer-term strategies. Until now, Chang was solely responsible for the company's long-term strategies.
Chang is not giving up any control of TSMC yet. But observers speculate that Tseng may be the heir apparent. There's no doubt that Tseng is the rising star at TSMC. "Dr. Tseng is expected to take still greater responsibility in planning and managing the company's long-term strategies," Chang says.
(See Aug. 7 story.)
Dataquest revises forecast again;
says IC sales will fall even faster
Once again, Dataquest turned more bearish about this year's global chip market. This time it has lowered its forecast significantly and is now predicting a 25.8% decline to $168.2 billion from last year's record $226.5 billion. Until this week, the San Jose firm was forecasting a 17% fall.
Next year's global IC forecast was also cut, from a 13% increase down to an 11.8% rise. The market researcher changed its mind because "the seasonal demand cycle for PCs and cellular handsets has not kicked in as of yet--and there is still too much inventory on the shelves."
But the industry may have hit bottom during the second quarter, Dataquest speculates. I don't agree, but the second quarter was painful for nearly all key semiconductor markets. PC shipments were down from a year ago by 1.9%--the first decline ever. Chip shipments for wired communications systems were off by 45% from a year ago, and were 14.6% lower for workstations.
It pretty much depends on how memory pricing goes, but third quarter chip sales could still fall, Dataquest figures. The new forecast shows a revised "downside" scenario with revenues falling 5% from last year.
In what may be indicating a sea state change, Dataquest is now forecasting that chip industry sales between 2000 and 2005 will rise at the compound annual growth rate of just 5.8%--far below the historical average of 17% annually over several decades. That's scary.
(See Aug. 8 story.)
Vitesse to play new role as
Vitesse Semiconductor is striking out in a new direction in hopes of developing a new revenue stream. The Camarillo, Calif., chip maker, which saw its revenue in the June 30th quarter swoon 47% to $60 million, is entering the foundry business by offering chip-making services using its new indium-phosphide (InP) process technology.
The major shift in strategy moves Vitesse into competition with a small group of foundries, including Kopin and TRW, that offer InP-based wafer outsourcing. It also will compete against silicon-germanium. Atmel, IBM, and several others now provide SiGe-based foundry services for high-end networking and wireless applications, the same applications that Vitesse is going after.
Vitesse is going full-speed on other InP fronts as well. It expects to begin sampling heterojunction bipolar transistors (HBTs) for 40-gigabits-per-second (OC-768) applications by the first quarter of next year. Vitesse claims its VIP process is different than other InP processes in that it uses standard process equipment and manufacturing techniques.
(See Aug. 6 story.)
Mentor begins crash effort to
develop FPGA design synthesis
Mentor Graphics has started a crash program to develop design synthesis software for field programmable gate arrays (FPGAs) with the goal of executing a new technology roadmap within the next 18 months.
Putting system-on-chip (SoC) designs on programmable logic platforms is forcing EDA vendors to adapt their tools so they can deal with such "unprecedented technical challenges as massive frequency increases and exponential gate count growth," points out CEO Wally Rhines. "Technology and chip complexity are moving faster than anticipated in the programmable logic market," he says. To keep up, Mentor Graphics is investing "significant corporate resources," he adds.
Project Atlanta calls for new synthesis capabilities to be delivered in three phases over the next 18-months. In phase one, Mentor will focus on a new approach to FPGA synthesis called "heuristic synthesis," which will focus its design knowledge on larger building blocks such as CPUs, ROMs, RAM, and content addressable memories. It will be released as a beta program in the fourth quarter.
Mentor plans to deliver in phase two what it claims will be the industry's only true physical optimization solution for FPGA-based designs. The company plans to leverage technology from its ASIC timing closure tools to merge physical information and data with synthesis algorithms. FPGA designers will be able to improve design performance by replacing inaccurate wire load models with data that properly accounts for interconnect delays, the company says.
Finally, Mentor plans to make available synthesis software that enables engineers to design at a higher level of abstraction, allowing for rapid design exploration and implementation. Sounds challenging.
(See Aug. 6 story.)
Will vendors get premium price
for 333-MHz DDR SDRAM?
DRAM suppliers have to be an optimistic lot. As they start ramping up 266 megahertz double-data-rate SDRAMs, some of them already are getting ready to sample 333 megahertz versions later this year.
The chip makers are hoping to get a premium for the new higher-speed chips since they will be initially aimed at servers and high-end workstations--markets that have less price elasticity, says Bill Gervasi, technology director at Transmeta. But observers are quick to point out that the fiercely competitive DRAM producers already have stopped charging price premiums for the PC266 chips over single-data-rate PC133 SDRAMs, even though they are at the very beginning of the DDR lifecycle.
Elpida Memory, Micron Technology, and Samsung Electronics all say they will have PC333 DDR samples by the fourth quarter and will be in production in the first half of 2002. And early next year, Via Technologies expects to have chip sets supporting PC333 for Advanced Micro Devices and Intel microprocessors.
DDR chip makers will reach 333 megahertz as they move into 0.15-micron processing, Gervasi says. "They also get some speed improvements from a tighter delay-lock-loop design for greater accuracy during memory read cycles," he adds.
(See Aug. 6 story.)
Samsung wants 30% share
of SRAM market in 2002
Samsung Electronics is going all out to increase its market share in the SRAM market. It plans to sell $1.8 billon worth this year and grab a 30% share of the global market share next year.
The Seoul company is moving fast. It already is moving into volume production with its 0.13-micron process three months ahead of schedule. And it has started volume production of its new line of low-power, 8-megabit SRAMs using the new process and plans to launch a 16-megabit version of the low-power SRAM in the fourth quarter.
Full-scale production of the 8-megabit SRAM is expected to begin in the fourth quarter. The low-power chip is aimed at next-generation mobile phones, which need the low-power static memories for video, audio, and data capabilities.
Samsung also is moving ahead with a 0.10-micron process for low-power SRAM production in 2002 and a 0.08-micron process in 2003.
It is counting on the worldwide mobile handset market to grow at an average of 15-to-20% annually. Samsung estimates 380-to-410-million handsets will be shipped in 2001 and more than a 1 billion by 2004. Good luck.
(See Aug. 6 story.)
Big loss doesn't stop Nanya
from going after fifth place
Nanya Technologies may be in the toilet this year like every other DRAM supplier, but that doesn't mean it's backing off its goal to move into the Top 5 DRAM vendors next year.
The Taiwanese DRAM maker, which expects to post a net loss of nearly $167 million this year because of falling prices, revised its forecast for this year's sales downward to $456 million, half of its original target of $882 million.
But Nanya is not making any more cuts in investments or production plans this year. It still plans $347 million in capital spending to add capacity to its two existing 200-mm wafer fabs. It also is sticking with its plans to break ground for a 300-mm wafer fab next March.
The Taiwan chip maker also is getting ready to migrate its DRAM production to a next-generation process with 0.14-micron feature sizes. That will happen in the first quarter of next year. Device shrinks and increases in production capacity will play a major role in Nanya's drive to the Top 5 suppliers.
At the end of last year, the company was ranked No. 10 in DRAM market share with about 2.2% of worldwide sales. Estimates by Nanya and some analysts now show it growing to a 5.5% share this year, moving it up to No.7 and near the No. 6 position.
But Nanya has its work cut out. The DRAM problem now is both demand and falling average selling prices. "We think there will be some strengthening in the fourth quarter," says Ken Hurley, president of Nanya's North American subsidiary. "Whether we call that a recovery now is premature."
(See Aug. 8 story.)
Agere pulls major coup
against Intel in China
Agere Systems has pulled off a major deal with the "Cisco of China" that observers say could threaten Intel's communications chip business in China.
The Allentown, Penn., chip maker has just signed an alliance with Huawei Telecommunications, China's largest networking equipment supplier.
The problem is that Intel considers Huawei one of its key accounts in the network processor area. The Chinese company reportedly uses Intel's line of network processors in its equipment and the two companies have been developing a line of chip products at a laboratory in China, as part of a long-term alliance between the two.
So what's going on? Analysts believe that Huawei is still working with Intel--at least for now. Indeed, the Chinese company could even be developing two sets of equipment, one based on Intel chips and the other on Agere products. In any event, analysts believe the deal between Agere and Huawei could weaken Intel's position at the Chinese company.
Listen to what Huawei has to say about the new alliance. "Agere's strength as a leading supplier of communications components confirmed our decision to partner with them," says vice president Liu Ping. "With Agere's technology, we believe we will gain significant time-to-market advantages over our competitors in the global networking marketplace."
Under the terms of the multi-million dollar deal, Huawei initially will use Agere's PayloadPlus family of programmable network processors and PI line of switch fabrics in its line of networking-equipment products. At the same time, the two companies will work together to develop a new line of chip products.
Another reason why the Huawei account is significant for Agere is that China's communications market seems to be growing at a time when this market in the rest of the world is experiencing a severe downturn.
(See Aug. 6 story.)
Is Motorola raising
needed cash the easy way?
I can remember visiting Motorola's military electronics operation way back in the 1960s. Sitting out in the desert in Scottsdale, Ariz., the division was a smallish, moderate growing unit that had a sterling reputation with the Pentagon and the Feds for quality products. And it was profitable.
So when Motorola disclosed this week it was selling its defense and government systems business to General Dynamics, I couldn't help but wonder if the troubled company wasn't eating its seed corn--or however that old saying goes.
Granted that GD assumes the unit's debts and pays $825 million in hard cash to the cash-short company, but there always seemed to me to be a good symbiotic relationship between the system house and Moto's chip unit. The Motorola group has 3,000 employees developing and building products for secure and integrated communications. Revenues next year should hit $830 million.
Here's Moto's official reason for selling it. If you can figure it out, lemme know. According to CEO Chris Galvin: "As Motorola increases its focus on business areas central to its long-term strategy, we feel certain that General Dynamics, with its strong government focus, will be the right home for Integrated Information Systems' employees." I wonder if Chris has checked the profits and stock prices of aerospace firms recently?
(See Aug. 6 story.)
Online chip exchanges
head for major shakeout
Just a couple of years ago, a good many people were predicting traditional brick-and-mortar semiconductor distributors were going to end up being replaced by the new wave of Internet startups. And soon dozens of these online exchanges were being formed. But today the outlook is completely different. This group of online companies is headed into a major shakeout and it's difficult to say just how long some of them will survive.
Today's tough economic climate as well as the drastic turn in the supply/demand environment is forcing several of these dot-com companies, including eChips, FastParts, and Need2Buy, to overhaul their strategies.
Many of them failed to influence the electronic components supply chain the way people thought they would a year ago. "Adoption and use are dependent on a growth environment. There is no growth right now in the industry," notes management consultant Bob Moncrieff at Pittiglio Rabin Todd & McGrath. It is beginning to look like some of these companies may be better off shutting down or being bought than pouring more money and resources into their development.
FastParts, which got $30.5 million in venture funding last year, has been struggling. Earlier this year, the dot.com company cut its 50-person staff by 25%, reportedly because of pressure from its venture capital investors to reduce spending. Now the company is seeking strategic partners, and sources say it is in acquisition talks with DoveBid, an auction services company.
Some of the giant traditional distributors also have gotten caught up in this problem. Over the past two years, for example, Arrow invested $60-to-$70 million on several of these on-line start-ups. And a hybrid business model emerged last fall when Arrow's Advantage telesales business unit merged with QuestLink and ChipCenter to form eChips. These investments that were made by Arrow, as well as by its rival Avnet, were as much a defensive strategy as they were an offensive one, analysts say.
(See Aug. 7 story.)
If you have any comments, criticisms, or questions, don't hesitate to E-mail us at firstname.lastname@example.org. Have a great weekend!
(Click here for last week's Semiconductor Alert!.)