Greetings from Down-East Maine--where most folks this week are worried more about their rising school taxes than they are about the gyrating stock market. This new weekly column, successor to a weekly newsletter we wrote during the 1990s, analyzes and comments on the important developments of the past week in the chip world.
A glut of processed wafers by 2001? Don't rule it out
We always think of Manhattan apartment buildings when the global chip industry begins to explode with new projects to add fabs and capacity. When there's not an apartment to be had between 96th Street and the Village, the big builders start digging foundations for 50 new high-rise buildings. Of course, when all this added capacity goes online two-or-three years later, New York will be in the midst of a recession and every new building is giving three months free rent to sign people up.
I'm getting those feelings again after hearing news this week that the world's three largest dedicated silicon foundries plan to boost production of wafers by an amazing 116% over the next two years. Of course, managers at the three companies and even analysts are quick to say that this boost in new capacity will not result in a glut of processed wafers by the end of 2001.
Yeah, sure. The Big Three foundries are expected to boost production this year to 6.77 million 8-inch equivalent wafers and to 9.1 million wafers by 2001. This would account for 16% of the world's processed silicon that year. Last year, Chartered, TSMC, and UMC produced 9.6% of all wafers (See April 3 story).
Chip sales fell? Yeah, but only a little bit
Sales of semiconductors in February slipped surprisingly. Global sales fell 2% from January's total to $14.56 billion. But they were still far ahead of year-ago figures, running 33.4% higher than the $10.92 billion in February 1999, according to the Semiconductor Industry Association. Best growth came in the Asia-Pacific region-up 45.4%--and in Japan--up 42.5% from last February. Trailing were the Americas market, up 24.7%, and Europe, 25.1% higher than 1999 (See April 6 story
Big growth expected to continue for fabless companies
Market forecasts are like weather forecasts. More often than not, they're going to be wrong. But we still like to hear them. Partly because they're usually feel-good predictions.
Take the forecast from the Fabless Semiconductor Association--or FSA--which this week put on its usual dog-and-pony show for its annual predictions. Its members expect to consume 39% more foundry wafers this year than they did in 1999. And next year, wafer consumption will climb another 48%.
A year ago, these same guys were bullish, predicting their wafer consumption would grow 43%. The boom comes mostly from the communications markets. This year, 37% of the wafers sold by fabless companies will be analog and mixed-signal products, prime components of comm gear, which is 50% higher than 1999's 25% share. The fabless chip suppliers figure they'll need 2.6 million 8-inch equivalent wafers from foundries this year. But independent device makers also are going to foundries for more of their manufacturing. This is beginning to strain the production lines of foundries, according to the FSA (See April 5 story).
ASML rolls out 248-nm scanner for 0.13-micron volume output
ASM Lithography, along with many others, is convinced that 0.13-micron feature sizes will be required for IC production in the second half of this year. The problem is that the new 193-nanometer processes, reticles, and photoresists are not likely to be ready for volume production for at least another year.
To bridge the gap, the Dutch lithography equipment maker has come up with a new 248-nm step-and-scan tool designed to fabricate ICs with 0.13-micron design rules. It was unveiled this week at the Semicon Europa trade show in Munich. The new scanner is an enhancement to ASML's deep-ultraviolet 700 platform, introduced last year for 0.15-micron chip production. Phase-shifting masks and optical proximity correction technology will be used to get exposure resolutions down to 130-nm. Shipments of the new scanner should begin in the second quarter. The $8.6 million tool has a throughput of 120 eight-inch wafers per hour.
ASML doesn't expect its customers to begin moving 130-nm systems into volume fab lines until next year. By then, it expects to have introduced a new volume production scanner (See April 4 story).
Don't count on Bluetooth for five years-at the earliest
It figures. A great idea comes along for a neat product made possible by rapidly falling chip prices. But despite all the interest and hype, this promising application keeps getting delayed because chip makers can't get their prices down fast enough. So it goes with the long-promised rollout of short-range personal-area wireless networks based on the Bluetooth standard.
There was a big debate over pricing at this week's Bluetooth conference in Geneva, which attracted more than 1,000 developers. It looks like it will take years to implement this technology due to the high cost of embedding it in cellular phones and other mobile devices. Magic number for a Bluetooth connection was $5 as far as the initial members of the Bluetooth consortium-Ericsson, IBM, Nokia, and Toshiba-were concerned. But chip makers this week couldn't agree on whether they could hit that price, much less when it would happen.
It gets even more problematical when they look beyond the first point-to-point cable replacements to such applications as wireless Internet ports for laptops in hotels and airports. Says International Data Corp.'s Randy Giusto: "The magical $5 figure is not doable within the next three years."
While Bluetooth add-ons will begin to roll out this year, the cost of hardware alone will be in the $20-to-$25 range, meaning that these first applications will be limited to laptops for mobile professionals and other high-end devices. Total cost of a Bluetooth chip set won't drop below $5 until 2005 at the earliest (See April 7 story).
Europeans spending more $$$ to add capacity
The action this week in the chip production equipment world centered in Munich. And what a difference a year makes. Suppliers showing their wares at the Semicon Europa trade show are seeing a strong recovery in demand for new tools and technology from IC makers.
The European market for production equipment fell 10% in 1997 and another 6% in 1998 before bouncing back 14% last year to $3.24 billion. This year, the European market for chip-making gear is now expected to increase by 30-to-40% to more than $4 billion. Much of Europe's capital spending growth is being fueled by its own chip makers.
STMicroelectronics is raising its investment in new production capacity by 63% to $2.2 billion in 2000, up from $1.4 billion in 1999. Infineon Technologies, which was going to invest $1.2 billion in 2000--a growth of 10%, could go higher now that the company expects to invest $1 billion to move 300-mm wafers into volume production. Philips Semiconductor currently expects to raise its investments by 4% to $950 million in 2000 (See April 5 story).
SEMI can't decide in global chip gear forecast
You got your choice of two global forecasts for chip gear sales this year-a "slow ramp" outlook and a "fast ramp" version. The way business has been jumping in recent weeks, we'd probably go for the fast ramp. In the slow ramp, the Semiconductor Equipment and Materials International (SEMI) trade group looks for worldwide equipment sales this year to rise 20% over last year's $25 billion to $30 billion, then peaking at $48 billion in 2004. This week SEMI also released its fast-ramp forecast, which has chip gear sales growing a whopping 44% to $36 billion in 2000, rising another 19% to $43 billion in 2001, and peaking at $49 billion in 2004 scanner (See April 5 story
AMD finally gets it right-to report record quarter
Advanced Micro Devices just finished the kind of quarter it has been dreaming about for the past couple of years. Its two main product lines--flash memory and PC processors--hit records in both units and revenues.
That was the word from a happy Jerry Sanders in Tokyo this week for the 25th anniversary of his Japanese subsidiary. Jerry says that AMD surpassed its public goal of selling at least 1.2 million Athlon processors during the first quarter and gaining unit market share. As a result, AMD will report its first billion dollar quarter on April 12. That would be a gain of more than 5% over the preceding quarter when AMD had record sales of $967 million and earnings of 43 cents per share. First Call/Thomson Financial, which tracks analyst forecasts, reported the consensus estimate was first-quarter profits of 48 cents a share (See April 5 story).
Does Infineon have 6-month lead in 300-mm production?
Infineon Technologies and Motorola may have got their 300-mm wafer pilot line up and running faster than some people expected. The Munich company now believes the joint-venture is six months to one year ahead of its competitors in getting the new 12-inch wafers into production.
The Dresden line claims it is now achieving 90% production yields on 64-megabit DRAMs. By obtaining these yields on the new-generation wafer, the German chip maker (spun off from Siemens) believes it can slash the cost of producing 64-Mbit memories by 30% from those made on standard 200-mm wafers. Infineon says it has been shipping 64-Mbit parts to customers from the Dresden pilot line since last September. It plans to move its new 256-Mbit DRAMs to the 300-mm line and use an advanced 0.17-miccron process to make them (See April 3 story).
Low-k dielectrics are big story this week
This was the week when newspaper and TV reporters were having a hard time explaining why a new low-k dielectric material was so important to IBM and future chip products. Even though the announcement had been expected, it was decidedly big news when IBM went public this week with its low-k dielectric material for fabricating the next generation of faster chips.
Insulators with lower capacitance like IBM's new material allow signals to travel faster through a chip, increasing overall performance. While other chip makers have been trying to use glasslike substances similar to silicon dioxide that now dominates current chip making, IBM went to a completely different substance--an organic polymer called SiLK developed by Dow Chemical. Other companies had tried organic polymers, but no one could marry them to a copper interconnect process, IBM claims. It solved the problem by creating a polymer sandwich that is firm enough to withstand normal chip processing.
One big advantage of the IBM technology is that the new polymer will use traditional spin-on processing tools, which means production costs will not change significantly. Volume shipments of chips using the new material will begin early next year (See April 3 story).
CVD developers still in race, but some hedge their bets
Because IBM has led the way in producing copper chips in volume, you'd think Big Blue's disclosure that it would use a new spin-on low-k dielectric material would be a major setback for tool suppliers promoting other approaches to low-k dielectrics.
Not so. Three tool suppliers-Applied Materials, Novellus, and Trikon Technologies-have all been promoting chemical vapor deposition (CVD) technologies for the same job. Executives at these three companies declare the race is far from over. In Munich this week for the Semicon Europa meeting, the managers agree that CVD film still has big advantages over IBM's spin-on approach in mechanical strength, lower production costs, and process integration with dual-damascene copper processes.
Novellus says that CVD is still favored by 80% of the people in the industry. The problem now with spin-on films is the high price of chemicals. It costs up to $10 a wafer for spin-on materials, compared with less than $2 per wafer for both CVD materials and equipment depreciation. But after all the talk, it turns out that Applied Materials and Novellus are hedging their bets on CVD and are exploring options that would move them into IBM's spin-on camp. Applied reportedly is working on a new track system for spin-on low-k dielectric applications (See April 6 story).
Mature markets looking good for Fairchild
The big guys got out of the commodity chip business because they didn't see enough growth or profit. That is proving to be short-sighted, at least as far as Fairchild Semiconductor is concerned.
The Maine chip maker is flying high these days with its new power and interface products and it's doing a lot of talking about it. Sales are so good that the company has revised its first quarter estimates for a second time (It will report on April 25.) Two weeks ago, Fairchild said it was expecting a 6% quarter-to-quarter increase in sales, but this week it raised its estimate to 10%. Earnings per share also will be higher than the current analysts' consensus of 38 cents per share.
As a result, Fairchild is expanding a lot faster. It is spending 15% of sales this year on capital expenditures this year. That means it will invest a total of about $255 million: $70 million in South Portland, Maine, $44 million in West Jordan, Utah, $30 million in Penang, Malysia, $79 million in Puchon, South Korea, and $32 million in Cebu, Philippines (See April 3 story).
Infineon officially joins the 300-mm race
As expected, Infineon Technologies is going ahead with a 300-mm wafer module at its fab in Dresden, Germany. Adding the next-generation line will cost more than $1 billion over the next three years. It will incorporate technology developed with Motorola at their Semiconductor 300 joint venture fab in Dresden. When operational, the new line will employ 1,100 workers. Groundbreaking will take place at the end of May (See March 31story
Despite lousy time, Cabot IPO shines
It may not have been the best week to go public, but while the stock market was scaring everyone with its roller-coaster ups and downs, one semiconductor equipment maker did just that. Cabot Microelectronics Corp., the largest supplier of slurries for chemical mechanical planarization (CMP) equipment, raised $80 million in it successful offering. Opening at $20, the stock rose to $27.50 later the first day. The Aurora, Ill., company was formerly the Microelectronics Materials Division of Cabot Corp., Boston-based chemicals manufacturer, which still holds 82.6% of the stock. It plans to distribute those shares to its shareholders in the next year or two (See April 4 story
If you have any comments or questions, don't hesitate to E-mail us at firstname.lastname@example.org. Have a great week!
See last week's Semiconductor Alert!).