Hold on; foundries aren't
taking over world after all

To investors and even to some customers, it must seem as if the silicon foundries are on their way to taking over the entire semiconductor business. But it ain't going to happen, according to a panel of leading industry executives at the Electronica trade show in Munich this week.

There are a growing number of reasons why foundries are coming up short now, they say. For one thing, traditional, large integrated device makers (IDMs) still hold a distinct advantage over pure-play foundries in the emerging production of complex systems-on-chip (SoC) designs, they claim.

"Systems-on-chip is not a main logic process," points out Ulrich Schumacher, CEO of Infineon Technologies. "It requires embedding all kinds of memory on chip. It's RF radio-frequency, and other technologies. You don't have foundries today providing all of these technologies," he says.

Fabless semiconductor companies were able to succeed in recent years by concentrating on high-value intellectual property (IP) for specific system functions, says Fred Shlapak, president of Motorola Semiconductor. But this strategy weakens, he says, when these suppliers move into new markets or attempt to sell system-on-chip products.

The senior execs also question whether their companies could ever make a decent profit by going to foundry wafers for high-volume, leading-edge products. "It is virtually impossible to make money with foundry wafers at the leading-edge technologies and at high volumes," claims Shlapak.

"The cost of a foundry wafer in the past 18 months has been two times higher than our own manufacturing costs," estimates Infineon's Schumacher. "But even when foundry supplies become plentiful again, foundries must still make at least 30-to-35% return on investment," he predicts. "That return on investment will make their wafers unprofitable for major chip houses to buy in a number of highly competitive markets."

Such factors seem to be changing some corporate strategies. Motorola initially was pursuing an aggressive strategy to outsource as much as half of its chip making, but now appears to be changing its mind. Shlapak says that Motorola's foundry target is now 12-to-15% of its total frontend capacity--far lower than its earlier goal.

Foundries face a growing problem as requirements increase for specialized processes beyond leading-edge CMOS, says Shlapak. "Where many of us do business wireless and communications markets, it is virtually impossible to compete at high-volumes in baseband technologies from foundries and to still make a profit." In fact, he says, "It's impossible."

In fact, it is becoming harder for pure-play foundries to be competitive in a number of new chip technologies that are needed for next-generation communications and wireless products. Foundries today cannot supply leading-edge R-F technology or even power-chip technology, says Ulrich Schumacher, CEO of Infineon Technologies.

"It is significantly easier to develop 0.18-micron CMOS than to develop indium phosphide, gallium arsenide, or silicon germanium technologies," he argues. "It will not be so easy because it is not just a question of money. It's also a question of expertise."

(See Nov. 21 story.)

How Intel is
betting the store

When 85% of your business is in just one line of products--and you're about to introduce your first new design there in five years--it is definitely a "bet your company" situation. That's true even if you're the biggest chip maker in the world.

That's exactly where Intel finds itself this week as the microprocessor leader rolled out its long-awaited Pentium 4. "The roll out of the Pentium 4 is critical to the company's whole future," declares Tony Massimini, who tracks this market for Semico Research.

The processor business is "still the cash cow for Intel," he points out. "Intel's other four operating divisions are operating in the red," he notes. For this reason, he says, it's important for the company "to get the ball rolling to the next-generation" of processors.

Intel certainly faces big challenges. Its biggest comes from Advanced Micro Devices, which has steadily gained MPU market share in recent months. So Intel was quick to point out that its new processor will keep it ahead in the high stakes race with its only real rival. "With this launch, we will be 300-megahertz ahead of them," declares Intel EVP Paul Otellini. The current speed leader, AMD's Athlon, is rated at speeds up to 1.2 gigahertz, while Intel is rolling out Pentium 4 chips that run at 1.4- and 1.5 gigahertz.

As the first major change in Intel's processor architecture since 1995, the Pentium 4 initially is geared for the high-end desktop/workstation market. It will eventually become the company's mainstream processor line, replacing the current Pentium III Celeron chips.

But there is a performance gap between the Pentium III and Pentium 4, says Semico's Massimini, and AMD has a product that fits between this gap. "The only thing that's holding AMD back here is the lack of manufacturing capacity," he says.

(See Nov. 20 story.)

Ignore last month's hiccup;
chip gear orders hit record

A month ago, the semiconductor equipment market hiccupped, causing some observers (like me) to worry a bit and question whether the global semiconductor markets were about to start heading south. But it's like the man said: don't look at the stock prices every day--stand back and watch the longer term trends and ignore the daily jiggles.

This month, North American suppliers of semiconductor equipment posted record orders for the month of October, according to the SEMI trade group. They posted record orders of $3.04 billion in October, which helped to push the book-to-bill ratio up slightly to 1.17, according to SEMI. That means orders were 17% higher than shipments for the month, indicating an expanding market.

Bookings were 5% above the previous month and 89% higher than the $1.61 billion in orders posted in the year-ago month. October shipments three-month average hit $2.59 billion, 5% higher than September and 75% higher than October last year.

Here's how SEMI CEO Stanley Myers sees it: "Following what appears to have been a seasonal pause in growth through the summer months, orders for new semiconductor equipment re-accelerated in October. Front order growth was the strongest we've seen in six months." So from now on, I guess, we should look at these numbers for trends and ignore monthly fluctuations.

(See Nov. 21 story.)

Substrate shortage could slow
IC makers moving to 300-mm

Cautious blank-wafer suppliers, who have been hurt badly in the past by adding new capacity too quickly, may have a surprise in store for chip makers finally shifting to the next-generation 300-mm wafer.

Wacker Siltronic, for example, is moving ahead with a feasibility study for constructing a new high-volume plant that would dramatically increase shipments to IC makers. The big Germany firm, a division of Wacker-Chemie and the world's second largest supplier of blank silicon wafers to chip makers, wants to hold on to its one-third market share of 12-inch wafers.

But Wacker officials caution that they haven't decided on a launch date. "It will be three to four months before we can be at the point of a reliable decision," says CEO Robert Rmer. In fact, a go-ahead is not likely to happen even next year, he notes, if uncertainty continues to exist over when chip makers go into volume production of the 12-inch platters.

That kind of caution, however, could result in big problems for chip makers. Suppliers are not likely to be ready to for high-volume shipments of production-grade substrates if they don't get guarantees from chip makers within the next few months on how many wafers they will need. In fact, there's already a good chance that 300-mm wafers will be in short supply next year, warns Rmer.

Even when Wacker decides to build its new 300-mm production plant, it will take another two years to get production started, he says. Such a "greenfield" plant will cost as much as $440 million. Based on current projects, worldwide demand for 300-mm wafers is expected to grow from the current 25,000-to-30,000 a month to as much as 200,000 next year, based on current projects. Wacker hopes to push its pilot line to the 50,000-per-month range, but managers say they fully expect to see a shortfall in 12-inch supplies as new 12-inch production fabs begin to ramp their volumes

A major problem for the 300-mm movement has been the depressed prices for 200-mm wafers. Shortages are now beginning to appear, so prices have moved up slightly by 5%-to-10%. But this is far short of what's needed for major new investments in wafer blank capacity worldwide. Blank wafer suppliers remember all too well the losses they racked up in 1998 and 1999--analysts figure the industry lost as much as $1 billion annually.

(See Nov. 22 story.)

When you think Infineon, think
communications ICs, not DRAMs!

Infineon Technologies doesn't like the beating it is taking on the U.S. stock market--down nearly 60% from its high this year. None of the other beaten-down chip makers do either, of course, but the former Siemens division says it isn't getting a fair shake from investors.

So company management have been busy in Silicon Valley telling their story to financial analysts and the press, hoping to change the perception of the big German company. "We feel we don't get a fair valuation in the market based on our presence in communications," declares Mark Tyndall, vice president for the Munich-based company.

It's the DRAM that seems to grab most of the media's and Wall Street's attention. "We are more than a DRAM company," Tyndall insists. "We're also strong in communications."

But analysts believe the big chip maker has fueled its own perception in the market, especially by promoting its aggressive expansion in the DRAM business. Nearly half of its worldwide sales, in fact, come from memory products, which amounted to $2.98 billion in fiscal 2000--doubling the previous year's memory chip business.

In contrast, its sales of communications chips amounted to $1.86 billion last year, up from $1.36 billion in the previous year. Wireline communications chips showed a loss, which widened in the fourth quarter. So Infineon has reorganized this group into six divisions to streamline these operations.

(See Nov. 20 story.)

Micron is now sampling
256-megabit DDR SDRAM

Early next year, Micron Technology will be cranking out production quantities of its new 256-megabit, 266-megahertz double-date-rate SDRAM.

First PCs to use the new chip will be those systems based on Advanced Micro Devices' Athlon processors. But third-party vendors are beginning to turn out DDR chip sets that will make it possible for PC makers to use Micron's 256-megabit DDR with Intel's Pentium III.

Sampling has already started on the new chip, which is built in a 16 x 6 configuration. Later this quarter, it will also be available in 64 x 4 and 32 x 8 configurations. Micron also is delivering samples of the device in PC2100 modules amounting to 512-megabytes. One-gigabyte DIMMs will be available in the first quarter of 2001.

Micron wasn't talking pricing yet for the new DDR chip in production quantities, but says that it will be comparable to production prices of 256-megabit single-data-rate PC133 memory chips.

(See Nov. 20 story.)

Infineon 'de-emphasizes' G.Lite
and will focus on fullrate ADSL

It doesn't look good for G.Lite, a stripped-down version of the digital subscriber line (DSL) standard that was developed as a low-cost technology that could easily be deployed to the mass consumer market.

Developed in 1998 by Compaq Computer, Intel, and Microsoft, G.Lite originally was expected to hit the mass markets in late 1999, but the technology has yet to be endorsed by the U.S. major carriers. The digital-modem standard sends data at speeds up to 1.5-megabits-per-second, while the full-rate ADSL standard moves information at 8-megabits-per-second. Unlike asymmetric DSL, G.Lite does not require the installation of a "splitter" at the home to separate the phone and data services that share the line.

This week Infineon Technologies threw in the towel and said it will "de-emphasize" its G. Lite product and focus on the full-rate asymmetric DSL (ADSL) technology.

G.Lite encountered major problems in field tests, due in part to compatibility issues among chip and equipment suppliers, says Mark Tyndall, Infineon vice president. Data-connection rates were also hindered, he says, due partly to the poor quality of copper wiring in most households. And signaling problems between the phone line and the modem forced the carriers to install an expensive passive component called a microfilter in every phone jack.

Demand for G.Lite was also poor, Tyndall complains "It's disappointing," he says. Because the G.Lite market has not taken off, Infineon's focus will be on full-rate ADSL, he says. This chip market is one of the company's most strategic product areas. "The VDSL market race is between us and Broadcom," says Tyndall.

The German company will push its full-rate ADSL chip line, which will begin shipping in volumes in the first half of 2001. It also is counting on a high-end ADSL technology called VDSL that supposedly will move data, voice, and video at speeds up to 26-megabits-per-second to homes and businesses.

(See Nov. 20 story.)

Foundries fight to offer
multi-project' wafers

You'll be hearing a lot more about multi-project wafers because they're becoming increasingly critical to the large silicon foundries for moving customers to their most advanced technologies. The battle heated up this week between the two largest foundries--Taiwan's United Microelectronics (UMC) and Taiwan Semiconductor Manufacturing (TSMC).

UMC is expanding its multi-project wafer program to include its new 0.13-micron copper logic process technology and dedicated "silicon shuttle" wafers for advanced mixed-signal and radio-frequency CMOS devices.

The foundry intends to speed up prototype shipments from multi-project wafers by reducing fabrication cycle times with a "hot-lot" schedule that puts the substrates on a fast-track at its fabs. This means that each photolithography layer will be completed in about one day. Fabrication of multi-project wafers with six-layer metal and 0.18-micron feature sizes will be finished in less than four weeks, UMC promises.

The UMC expansion comes weeks after rival TSMC said it was planning to increase its multi-project wafer shipments by 50% in the first half of 2001. TSMC also is extending its "cybershuttle" multi-project wafer effort to include its new 0.13-micron copper process technology.

Multi-project wafers have become increasingly important for the large foundries, since foundry customers can use them to cut costs by sharing photomask sets and processing steps because the multiple IC designs are packed on a single silicon substrate. This is becoming a big deal since the cost of some reticle sets are running as high as a half million dollars.

Multi-project wafers are becoming a useful tool for customers to try out the latest process technologies for new product designs as well as a viable way to serve low-volume products. The huge 300-mm substrates that are due shortly to go into volume production will make multi-project wafers even more important.

(See Nov. 20 story.)

An ambitious Applied Materials
expects to keep on its growth path

So what do you do for an encore after you hit your revenue run-rate goal of $10 billion that was set back in 1995? If you're Applied Materials, you set a new target--$20 billion in the next three-to-five years.

"We have set the bar another notch higher," pointed out CEO James C. Morgan after the world's largest supplier of chip production equipment finished the fourth quarter of its fiscal 2000 with record sales of $2.92 billion.

Applied aims to make its new goal not only by growing its market share in next-generation wafer-processing equipment, but also expects to do it by bundling process technologies, metrology capabilities, and "total" services with its tools. He expects to gain $3-to-$5 billion in annual business within a few years from this "total services solutions" business, which provides outsourced tool maintenance, guaranteed wafer-processing operating results, and spare-parts management in wafer fabs. The two-year-old business already is pulling in about $1 billion in annual revenue.

But Applied is running behind schedule with another new product that it expects will help it move to the $20 billion level. This is its process module program, which bundles advanced technologies and metrology with complete tool sets for wafer fab applications, such as copper interconnects and low-k dielectrics.

Applied's stand-alone metrology systems business continues to grow, but the overall strategy to integrate those systems with wafer-processing tools has not been launched. The equipment giant still hasn't released a product for integrated, or in-situ metrology, but does expects to ship its first integrated metrology systems and guaranteed process models for revenue sometime in the February-April timeframe.

(See Nov. 21 story.)

Germans chip market
duplicating U.S. growth

The outlook for the German semiconductor market just about matches the predicted growth of the U.S. market.

In its traditional Electronica forecast this week in Munich, the German Electrical and Electronics Mfrs. Assn. (ZVEI) said the German market will hit nearly $10 billion in 2000, up 42% over last year, making it the best year in the nation's history. Next year's growth will slow, according to ZVEI, with revenues climbing 20% to nearly $12 billion.

The German trade group also expected global chip sales to match the earlier forecast made by the U.S. SIA trade group. It says global chip sales will grow 37% to $205 billion this year. And in 2001, global chip sales will climb 20% to $246 billion.

German sales of all electronics components--semiconductors, passives, electromechanical, and printed circuit boards--are expected to grow 29% to $16 billion this year from $12.4 billion in 1999, according to ZVEI.

About 26% of the European components market is expected to be purchased by the computer industry this year, compared to 22% in 1999 and 21% in 1998. Automotive applications in Europe will account for 24% vs. 26% in 1999 and 25% in 1998. Telecommunications will consume 26% of electronics components in 2000 vs. 24% in 1999 and 24% in 1998, according to ZVEI.

(See Nov. 21 story.)

If you have any comments or questions, don't hesitate to E-mail us at bhenkel@aol.com. Have a great weekend!

(Click here for last week's Semiconductor Alert!.)