Greetings from Down-East Maine, The San Jose Mercury News ran a column this week on the 30th anniversary of the 4004--what Intel likes to call the world's first microprocessor. It makes me feel like ancient history because I ordered up the first story for BusinessWeek a couple of years later on the "computer- on-a-chip" and how it was going to change the world. We ran a picture of an Intel exec holding the 8080 microprocessor in his hand The 4004 wasn't practical to use for much more than a calculator and comparing it to the room-sized IBM 360 mainframe towering in the background.

Then in 1976, I wrote a cover story for BusinessWeek on the computer-on-a-chip revolution, which was picked up by Readers Digest. So I was there for the beginning of a technology that a smart-aleck columnist visiting the Intel Museum in Santa Clara "snaps a salute to one of the digital revolution's tiny soldiers." My gawd! But at least the Merc did tell the story of how a friend of mine, Frederico Faggin, converted Ted Hoff's brilliant idea into silicon. Those were exciting times!

Here's a shocker: spot
DRAM prices double . . .

Would you believe spot prices for DRAMs going into mainline PCs have doubled in the past week? Well they have, and no one seems to know why. This astonishing market turnaround has sent analysts and marketers scrambling to find out what's going on.

Here's the numbers. DRAMExchange.com reports the workhorse 16x8 PC133 128-megabit SDRAM was selling for the average price of $1.66 this past Monday, up from 85 cents just a week earlier. Converge listed the same part for $1.70.

The steep jump actually pushed prices ahead of the last negotiated contract prices, which analysts report were averaging $1.30 for the PC133 128-megabit part.

The flip-flop surprised most industry analysts, who had been watching DRAM prices die this year. "It is unclear what really spurred the price move," admits a puzzled Jonathan Joseph, chip analyst at Salomon Smith Barney.

(See Nov. 13 story.)

. . . But no one seems to think
the lower DRAM prices will last

Analysts and marketers were tossing out all sorts of guesses on why spot DRAM prices had suddenly doubled. Joseph Osha, chip analyst for Merrill Lynch Securities believes "Micron Technology and Hynix Semiconductor were testing the waters" for higher prices.

Farhad Tabrizi, Hynix marketing vice president, has a more complicated explanation. "When some of our DRAM competitors thought Hynix would not survive, they dropped the prices abnormally to kill us off." But after creditors approved the Hynix financial restructuring package and competitors "saw we weren't going to die, the competitors had to back off quickly on low pricing they couldn't maintain." Hmm.

And DRAM analysts in Asia, where the DRAM price hikes have been running the steepest, suspect the three biggest DRAM vendors--Samsung, Micron, and Hynix--had set a floor price of $1.50 or more that they wouldn't cut for chips sold into the spot market.

Samsung denies either cutting prices deliberately or setting a floor price to drive prices up. "The market is reacting to a confluence of factors moving prices up--seasonal high demand, a shift to Direct Rambus, and a lower supply of single data rate PC100 and PC133 SDRAMs," a spokesman says.

And Micron Technology was quick to deny setting any spot market floor price. "The market always determines the price that we can sell at," claims Michael Seibert, memory strategic marketing manager.

Jim Sogas, sales vice president for Elpida Memory, says move to higher prices "is certainly not demand-driven, so some factors must be at work in the supply side of the market." He adds, however: "If it isn't related to actual reduction in supply, then the spot market price increases aren't sustainable."

No one was predicting the higher prices would last. "Without a seasonal upturn in demand, we believe the recent DRAM price hikes will not stick," says one senior DRAM marketer. And the Samsung spokesman says the price jump "couldn't be maintained in face of a global supply glut that shows no signs of dissipating any time soon."

(See Nov. 16 story.)

Seems to me I've
heard this song before

I've heard the following complaint for so many years that I am beginning to wonder just how credible it really is. Listen up.

The EDA industry must move faster to close the gap between design tools and chip production capabilities so that chip makers can innovate. That demand comes from CAD tool user Vassilios Gerousis from Infineon Technologies, who presented his wish list of what EDA vendors could do to make users' lives easier.

Shrinking process geometries and growing IC gate counts are causing problems for chip designers in high-level design and physical verification, he notes. "We need to move to higher levels in abstraction to describe these large designs, but at the same time there are more physical effects such as crosstalk and inductance," Gerousis says.

Tool vendors need to come up with products that close the design gap and allow users to better exploit silicon, close the predictability gap to shorten product cycle times, and close the productivity gap to speed up product development, he notes.

To accomplish this, Gerousis suggests a radical licensing model where Infineon would pay EDA vendors for productivity improvements. The idea didn't appear too popular with EDA vendors on the same panel at a San Jose CAD conference this week.

Gerousis maintains Infineon "wants to work more closely with EDA companies so they can develop the tools we need." But this is the company that threatened earlier this year to build its own tools if EDA firms did not innovate.

(See Nov. 14 story.)

Motorola fattens up
chip unit for sale, or?

My morale would be lower than a snake's belly if I was employed at Motorola Semiconductor or was a Motorola shareholder. When I started flying to Phoenix in the early '60s to cover this division, it was a towering leader in the global chip business. And it was a vital part of the corporation. No more, it seems.

Now Motorola Semiconductor seems to be no more than an asset the company needs to fix up so that it can be peddled for the maximum amount of money. CEO Chris Galvin, whose grandfather founded the company, was in New York this week to speak at a Wall Street conference. And he compared his semiconductor unit to the company's government business division that was sold recently.

"Two years ago we were losing a lot of money in the government business, and we said there is no sense in selling it now or disposing of it now," Galvin says. "We fixed it. We got it generating significant positive cash flow and double-digit pretax profits. We sold that business last September."

Semiconductor has to be worth more than it is now before Motorola would consider selling it, Galvin indicates. "If you look at it from a marketplace standpoint and the business is valued at two, four, or six, or eight times sales, one can decide whether you want that in the portfolio or not," he says. "That's the essence of the plan we are working on right now for Semiconductor." Chicago never did understand Phoenix.

(See Nov. 15 story.)

So is ATE business
coming back now, or not?

The automatic test equipment (ATE) business is so bad that I find myself looking frantically for any evidence that this industry had hit bottom. So I jumped this week when LTX told the financial community that the magic day had come--at least for this ATE vendor.

Business in general is looking up for LTX, declares CEO Roger Blethen. "During the October quarter, we saw signs that our industry has hit the bottom," he says. "We also see significant increases in orders for the quarter that ends in January."

But analysts aren't so sure, and neither am I. Some of them wonder whether LTX is still hunting for the bottom, given its wide-ranging outlook. It projects that sales will decline sequentially anywhere from 6% to 30% in the January quarter.

The October quarter was a disaster, with revenues falling 55% from the year-ago quarter and off 30% from the previous quarter. Bookings were off 30% sequentially during the period, which was "weaker than expected," Morgan Stanley says.

(See Nov. 14 story.)

No news is bad news
at Applied Materials . . .

If you follow the semiconductor industry closely to try and figure out what's going on, then you have to watch what's happening at Applied Materials. Even when nothing much seems to be changing. That was pretty much the case this week when the chip equipment giant checked in with its results for the quarter ended Oct. 28th. What wasn't changing was the direction in which the equipment business was headed.

There was no signal that business had bottomed out--in fact, business seems to be continuing to get worse. Applied's sales declined sequentially by 20% to $1.26 billion in the quarter, and were off a whopping 57% from the year-ago quarter. The company almost met Wall Street's consensus estimate of net income, but so what?

Worse yet, the chip equipment leader expects another 20% sequential drop in revenues in the current quarter. So things keep getting worse here. New orders in the quarter fell 9% sequentially and were down 69% from the year-ago quarter rate. This was greater than anyone expected because of orders being cancelled for 300-mm tools.

For fiscal 2001 ended Oct. 28th, Applied had net sales of $7.34 billion, down 23% from fiscal 2000. Net income hit $508 million for fiscal 2001, a 75% drop from the $2.06 billion in the previous year.

(See Nov. 14 story.)

. . . as recovery 'postponed'
in semiconductor industry

Like it or not, Applied Materials is hunkering down now to wait for clear signs of an end to the downturn in semiconductor capital spending.

"The problem is we cannot get chip industry executives to commit to when they will sign the purchase orders," declares CEO James Morgan. "There are a lot of potential purchase orders out there," he notes, "but I don't think we'll know until the early part of next year what they are going to do."

"We believe this lack of visibility, added to an already slowing economy, has postponed a recovery in the semiconductor industry," comments CFO Joseph Bronson. As a result, Applied has downgraded its projections for chip industry revenues and capital spending. "This is the first downturn in history where both unit volumes and dollars have declined for both memory and logic chips," he points out. "We expect capital spending to remain weak in the first half of 2002 due to continued wafer fab under utilization and poor customer profitability," he adds.

Applied now estimates that the chip industry will purchase about $6 billion worth of 300-mm systems in 2001, a sharp drop from its summer forecast of $10.2 billion. It expects this spending will climb to $11 billion next year. Copper-deposition tool shipments are expected to grow 30% in 2002 to $1 billion, Bronson says.

A major problem facing the industry next year is the lack of leading-edge fab capacity and continued delays in investments as a result of uncertainty in the marketplace, Morgan says. The CEO says that many chip makers are now at the crossroads and must make decisions on adding new leading-edge capacity for 0.13-micron and below process generations.

Only 4% of the industry's current capacity is capable of producing 0.13-micron and below devices, Morgan estimates. So when demand finally takes off, the need for new fab equipment will "go vertical," he figures, which will be a problem of a different color. That's the kind of problem we would all like.

(See Nov. 15 story.)

Infineon scrambles
to get bigger in DRAMs

Why do you suppose Infineon Technologies wants to keep growing in the red-ink-plagued DRAM business? Is its strategy to become one of three or so suppliers of the commodity chips--few enough vendors to make a profit possible?

That strategy was thought to be behind the German company's move to enter into talks with Toshiba, which reportedly was looking for a way out of the shrinking DRAM business. It now appears that the likelihood of a merger occurring between the two companies' DRAM businesses is fading fast, given that they were unable to agree on a preliminary agreement by the end of October as had been hoped.

Now the Germans are turning to Taiwan to investigate what the chances are of picking up a DRAM supplier, according to sources. They say it is unclear whether the talks with Toshiba hit a snag or if Infineon is putting pressure on Toshiba by courting Taiwan's troubled DRAM suppliers. Another possibility: Infineon became disenchanted after Toshiba refused to include its flash memory business in the proposed joint venture.

In Taiwan, where the top six DRAM makers lost a total of $1.23 billion so far this year, several vendors say they've been contacted by Infineon on the subject of a merger. The German company already owns a 47% stake in ProMOS Technologies and may be interested in taking over complete control of the DRAM maker, or even merge with Mosel Vitelic, sources suggest.

"Infineon has talked to everyone in Taiwan about possible cooperation," including Nanya Technology, Powerchip Semiconductor, Vanguard International Semiconductor, and Winbond Electronics, claims Hander Chang, an assistant vice president at Winbond.

Taiwan's DRAM producers all license their process technologies from overseas partners, making them vulnerable in an industry down cycle. A merger of Infineon with any of the Taiwanese DRAM makers would not much to relieve the current supply glut, but it would help Infineon expand its market share.

Infineon now accounts for 8% of the world DRAM market, trailing Samsung Electronics' 22% share, Micron Technology's 20%, and Hynix Semiconductor's 17%, according to Nomura Securities. Teaming with Toshiba or a Taiwanese chip maker in DRAMs would give Infineon a larger critical mass to compete against the big three producers. With 300 mm wafer fabs coming on line, analysts figure that only five or six DRAM suppliers would be needed to meet global demand."

(See Nov. 13 story.)

Infineon sees six more
months of uncertainty

Outlook for global chip markets is still uncertain for another six months, as far as Infineon Technologies is concerned. The German chip giant expects to continue seeing price erosion in DRAMs due to excess manufacturing capacity and notes that cellular phone shipments still haven't recovered from their sharp drop off in early 2001.

Revenue and profits continued to worsen in the quarter ended Sept. 30th. Sales recorded a 15% drop sequentially, causing a net loss of $465 million. This means more belt tightening. Infineon cut its capital spending plans for the year beginning Oct.1st to 900 million euros, only one-third its original estimate of 2.8 million euros.

Memory products had a sequential drop in sales of 27% for the quarter just ended, causing a loss before interest and taxes of 522 million euros. Wireless communications revenues actually rose 7% sequentially from the previous quarter, but lost 178 million euros before interest and taxes. The wireline group showed a sequential decline of 24%, resulting in a loss before taxes and interest of 128 million euros. Only Infineon's automotive and industrial group showed earnings before interests and taxes. Sales were 6% lower sequentially.

(See Nov. 13 story.)

Can Mattson pull it out?

We were surprised back in October to learn that Brad Mattson was resigning as CEO of Mattson Technology and planned to retire. Now we are beginning to see why.

This week Mattson reported a net loss of $187 million on net sales of just $37 million in the third quarter. The loss includes $128 million in charges related to two major acquisitions at the beginning of the year and $21 million for excess inventories.

It comes as no surprise that the company has a lot more restructuring ahead to fit its operations to next year's anticipated business. This will include "another substantial reduction" in workforce in the fourth quarter. Mattson had already laid off 20% of its work force in August.

Mattson was hit hard by the current downturn in semiconductor capital spending because it came just as the California company was attempting to digest a complex three-way merger. Mattson acquired Germany's Steag Electronics Systems and CFM Technologies of Exton, Penn., to expand its tool business into rapid-thermal processing and wet-wafer cleaning.

(See Nov. 13 story.)

AMD rushes ahead on
silicon-on-insulator SOI front

It looks like Advanced Micro Devices is going through with its aggressive plans to use silicon-on-insulator wafers to make its next-generation microprocessors.

SOI promises higher operating speeds at lower power levels compared with circuits fabricated on standard bulk silicon substrates. AMD's position on SOI differs markedly with its chief MPU competitor, Intel, which doesn't see the need to switch to SOI substrates for its PC processors.

This week, a Bernin, France-based supplier, Silicon-On-Insulator Technologies received a multi-million dollar order for 200-mm SOI wafers from AMD, which will use them to produce its new Hammer series of 64-bit MPUs. Soitec, which claims to be the world's leading supplier of SOI substrates, says the AMD order was biggest in its history.

SOI substrates have become a prime candidate for central processing units in both high-end computer workstations and portable PCs, but higher costs remain a concern for most commercial applications. But AMD has already started pilot production of SOI devices with Hammer MPUs scheduled to go into pilot production by the second half of 2002. AMD claims it will ultimately convert all of its PC processor production to SOI wafers.

Soitec, which now claims 80% of the SOI wafer market , is building a second SOI wafer factory to push up its capacity to 1.2 million eight-inch substrates annually.

(See Nov. 12 story.)

An 'old friend' comes
back to haunt industry

Heck, I thought that computers were now keeping the industry's inventories in complete control and such industry problems as "double ordering" were ancient history. Boy was I wrong.

An out-of-control supply chain was the primary cause of what is being called the worst downturn in the history of the electronics industry. That's the growing feeling among some senior industry executives. One of the major culprits is the growing role of contract manufacturers in the industry, they say.

"There's a general illusion that inventory management is much better today than it was," says Wilf Corrigan, CEO of LSI Logic, who ought to know (He's been around the industry as long as I have.) But it's worse, he says. "It's worse because you have several layers now in the pipeline, and it's not really under control."

Because of the contract manufacturers, he says, "inventory is kind of out of sight, out of mind. Supply chain management has clearly deteriorated." Ten or 15 years ago when more companies were vertically integrated, inventory control was often in the hands of a single manager," Corrigan says. But inventory today winds up in the hands of contract manufacturers and distributors channeling products to them. "So instead of one single inventory, there are two maybe three inventories in series," Corrigan points out.

"There was a much higher percentage of outsourcing among our customers" in this cycle, says Rich Templeton, COO at Texas Instruments. "The result was that things got very overheated. Everybody was worried only about capacity and how much they could buy in 2000. There wasn't the attention to detail, attention to forecasting, and knowing where the inventory was at different points in the cycle," he maintains.

"So that's clearly one of the things that amplified the demand and inventory growth that we're paying for now as an industry--and that's got to tighten up," Templeton says.

"In boom times, orders are placed, factories are running at full capacity, and manufacturers quote longer lead times," points out Hau Lee, director of Stanford University's Global Supply Chain Management Forum. "Customers see lead times getting longer so they ratchet up their orders and build up their safety stock. And manufacturers see that as stronger demand."

"That's exactly what happened last year," he maintains. "The industry has a myopic view of the supply chain and a myopic view of the future." I think I wrote a similar story 20-odd years ago, quoting similar words about similar inventory problems.

(See Nov. 15 story.)

Agilent goes to 'Defcon 3,'
cuts another 4,000 jobs

Some industry observers say that Hewlett-Packard should have kept the original roots of the company instead of spinning them off as Agilent Technologies. But this year, these operations would have hurt rather than helped HP in its struggles to make a profit with its PC operations.

In August, Agilent announced its first workforce reduction of 4,000 jobs (Remember, HP used to be the no layoff company.) This week it decided to reduce its headcount by another 4,000 jobs, meaning that the Palo Alto-based test, measurement, and chip supplier will have sliced its overall workforce by 18% by the middle of next year.

Agilent's senior management also will take a 10% pay cut starting next year, while the rest of the employees will take a 5% pay reduction starting in the second quarter. All this is expected to save the company about $700 million on an annualized basis, with one-time restructuring expenses amounting to about $175 million.

So why all this upheaval? Well, the fourth quarter ended Oct.31st basically was terrible. Sales were down nearly 50% from the year-ago quarter to $1.6 billion. And "earnings-before-goodwill" were $275 million in the red, compared to a $328 million profit last year. Net orders declined 8% sequentially and 56% from the same period a year ago.

"Continued weakness in customer demand clearly indicates that the recovery will be delayed and more gradual than we expected," says CEO Ned Barnholt. "As a result, we are taking additional steps to further reduce our cost structure and return the company to profitability."

(See Nov. 15 story.)

Bulk silicon looking better
for IC use to end of decade

Here's this week's top technology story. Intel has made more progress on the development front that assures its continued use of bulk silicon to the end of the decade.

The chip giant broke its own record for the world's smallest transistor, claiming development of a 15-nanometer device that will be used to make microprocessors and other chips by the end of this decade. The transistor is a CMOS-based, 0.8-Volt device that will be able to switch at speeds of 0.38-picosecond. Last June claimed the record by developing a 20-nm transistor capable of 0.75-ps switching speeds.

The 15-nm device is expected to become a key element in the development of chips based on the company's P1268 process technology. Intel is expected to develop chips based on this process by 2009. By then, the company could have high-speed processors running at speeds of 20-gigahertz or faster, according to analysts.

The latest transistor proves that bulk silicon will continue to be a viable technology in semiconductor manufacturing for the foreseeable future, says Gerald Marcyk, Intel's components research director. "If you look at Moore's Law, we are trying to shrink the transistor 30% every two years," he points out. "This technology will take us out at least until the end of the decade."

But before the end of this decade--or even sooner--Intel and other chip makers will face major challenges in developing ICs with bulk silicon and still keep power consumption down. The challenge is to make devices with low "standby currents and standby power," Marcyk says. "Developing smaller and faster devices is not a problem. The challenge is to make devices smaller, faster, and with lower power."

Beyond the next decade, it is unclear whether silicon can remain a viable technology. "That is as far as I can rationalize it," Marcyk says.

(See Nov. 15 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!.)