Picture is turning
gray for chip makers

We know who the grinch is who stole Christmas! It definitely was the PC makers, who sounded the alarm this week that the sky was falling. Disappointing holiday sales of PCs, as well as slowing sales in several other end markets is having a cascading impact on chip makers.

A flood of warnings flowed out from chip makers this week as they lowered fourth-quarter earnings estimates. Cypress Semiconductor joined the crowd on Thursday, a list that already included Altera and Mosel Vitelic, among others. Cypress says that due to slower sales in a range of end markets it has sliced its quarter-to-quarter revenue growth to 4% or 5%.

Now is not a good time
for record wafer starts

Wafer starts at fabs worldwide took a sudden and somewhat unexpected jump in the third quarter, based on new data this week from the Semiconductor Industry Association. Starts surged nearly 7% over the second quarter.

This big increase comes at a difficult time for chip makers, who are trying to deal with cutbacks in new orders by major customers and inventory adjustments in key end-user markets, points out analyst Bill McClean, president of IC Insights. If this rate of increase continues in the fourth quarter, "we're looking at wafer starts increasing 25% this year," he says. That would be a "record for wafer-start capacity increases," he says. In 1995, the last boom year for chip growth, wafer starts increased 22%.

Fabs are now running at their highest capacity utilization rate since the mid-1990s. In the third quarter, they were operating at 96.4% of their installed capacity, up from 95% in the second quarter and 90.9% in the year-ago quarter. Low point in fab capacity utilization came in the third quarter of 1998, when the rate hit 80.8%.

The jump in capacity, however, wasn't caused by the surge in capital spending. It appears that the capacity growth came from existing chip plants squeezing out more unit volume with extra work shifts and new tools that have higher throughput rates. Capital spending by chip makers is at an all-time high in 2000, climbing nearly 80% from last year, but most of that investment in fabs and new equipment won't be felt until next year at the earliest.

This is causing increased concern about 200l. McClean has cut back his forecast for capital spending in 2001 to an increase of 10% or less. "We will see capacity spending turned off pretty quickly at the first real sign of weakness," he predicts.

(See Nov. 30 story.)

More bad news
for DRAM biz

The provider of DRAMs, flash memories, LCD driver ICs, and foundry services originally had expected to make a profit of $351.1 million on sales of $1.08 billion for 2000. Now, the Taiwanese chip maker says it expects to report a profit of $157.9 million on sales of $847.1 million in 2000. That's quite a change.

(See Nov. 27 story.)

This DRAM maker
is still bullish--but he's Taiwanese

Nanya Technology, world's eighth largest DRAM supplier, still has its sights set on booming revenues next year. It expects to grow 2001 sales by 88%, up from a projected $585 million this year. But Taiwan chip makers are well known for their overly-optimistic outlooks.

Big talk, yes, but the Taiwan DRAM maker is putting its money where its mouth is. It expects to begin building a 300-mm wafer fab building in April, about the time it will get its second 200-mm plant up to full production of 25,000 wafers a month. It plans to be fabbing 12-inch wafers by early 2003.

These aggressive plans by five-year-old Nanya come at a tricky time. DRAM pricing and the chip market are getting clobbered by inventory dumping and slower-than-expected memory growth, but the DRAM maker is convinced that conditions will stabilize early next year, followed by long-awaited shortages in the summer of 2001.

"We actually think that next year will be pretty good," exclaims Ken Hurley, Nanya's North America president. "It will start off slow, but our projections have us at $1.1 billion in sales next year, and we have not revised our targets."

By the end of this year, Nanya expects to make its goal of breaking into the Top 10 global DRAM makers, reaching the No. 8. position. And next year, it is trying to move up another notch or two and pass local rival Mosel Vitelic as the largest memory supplier based in Taiwan. "We think we will have some significant growth in market share next year," Hurley says. We'll be watching, Ken.

(See Dec. 1 story.)

Hyundai Electronics' troubles
deepen, so spinoff set for 2001

It seems that Hyundai Electronics Industries' financial problems get worse and worse. This week the huge DRAM maker said it is speeding up its planned spinoff from the parent chaebol and will try to raise $3 billion via new loans and securities and asset sales next year.

The spinoff was originally targeted for 2002. Analysts believe HEI hopes the move to an independent company will distance the business from its troubled sister chaebol affiliates and increase investor confidence.

But HEI has its own massive debt problems. It has, by itself, some $9 billion in outstanding debt, a growing problem that will be exacerbated by a large chunk of three-year bonds maturing next year.

But HEI management says it already has arranged a $435 million loan from Citibank and four Korean banks. The company also hopes to raise another $435 million in domestic loans next year, and to raise $1.15 billion from bond sales during the second half of the year. Finally it plans to raise another $425 million by selling off some assets and account receivables.

(See Nov. 27 story.)

Applied aims to grow share
of ion implanter business

Ion implantation is getting more important than ever for making chips. So Applied Materials is "significantly increasing" its spending for ion implanter capacity and R&D in order to keep pace with the growing demand for the tools in next-generation processes and 300-mm wafer fabs.

Dataquest says implanter sales will more than double over the next five years, growing from $649 million last year to $1.7 billion in 2005.

"Advanced ion implantation has become a critical technology as chip makers scale their device designs to the sub-0.13 micron technology node and 300-mm wafer size," comments Craig Lowrie, general manager of Applied's Parametric and Conductive Implantation Group.

Applied's investments will go for increased hiring, applications development, and facility expansion. It is upgrading its implant facility in Horsham, England, to triple its production capacity. A new manufacturing line will also be added in North America and more Class 1 cleanrooms will be added to R&D labs in North America, Europe, and Japan.

In addition to increasing its production capacity, Applied also is expanding into two new and competitive segments of the booming ion-implantation market--medium-current and high-energy. This week the equipment giant unveiled a combination medium-current/high-energy system for use in 200- and 300-mm wafer fabs. Applied has, for years, sold a line of ion implanters that compete in the other two major market segments--low-energy and high-current.

(See Nov. 27 story.)

Ion implanter business gets
more important...and bigger

Ion implanters have been around for a long time in the semiconductor industry. I remember when I saw my first system back in '70 or '71 at Mostek--remember the DRAM boys? But implanters are getting a lot more attention these days as they become more critical to achieving the shrinking geometries required in the latest IC design.

VLSI Research is the latest market researcher to look at this growing equipment business. The San Jose firm expects the global market to grow 58% between 1999 and 2001, or from $671 million to about $1.35 billion by 2001.

"Implantation is the only tool in a fab used 15 to 20 times in the semiconductor manufacturing process," notes Babak Adibi, global product manager for Applied Materials' implant division. "As you shrink the device, the ion-implanter becomes even more critical," he says. "If you want to improve the performance of a transistor," he says, "you have to control the channel of a device." And that means using the ion implanter to shoot a beam of electrical ions called dopants to form the transistor. "Doping accuracy is the most important thing going forward to improve transistor performance," Adibi says.

The global ion-implanter equipment market is dominated by three major U.S. players--Applied Materials, Axcelis Technologies, and Varian. Axcelis and Varian tied for the market lead last year with sales of $250 million each, VLSI Research says. Applied was in third place with sales of $120 million.

(See Nov. 29 story.)

Transmeta recalls some
Crusoes in latest mishap

Transmeta, the secretive, yet high-profile startup that's building a zippy low-power microprocessor, was making news again this week--but not the good kind.

In its second setback in recent weeks, the Silicon Valley fabless chip maker has identified a faulty batch of its X86 microprocessors installed in NEC's notebook PC and is now replacing the bad chips.

In an attempt to minimize the problem, a Transmeta spokesman insists the problem involves only 300 of its Crusoe microprocessors. "It's a limited batch," he claims. The problem is not in running applications, he notes, but could happen if a user reinstalls the PC operating system.

But Transmeta may not have its arms entirely around the problem. NEC already has shipped a bunch of systems equipped with faulty X86 processors in the channel. "It's unclear how many systems are in the channel," the spokesman says. But he maintained the problem is limited to NEC.

(See Nov. 29 story.)

LSI Logic's loss
is Altera's gain

Well, the Valley gossips were right. John Daane, 15-year veteran and communications chip chief at LSI Logic who resigned from that post last week, is taking over as CEO at Altera. He succeeds Rodney Smith, who had disclosed his intention to resign last year.

Daane seems to be right for the job. At LSI Logic, he was responsible for the company's communications chip lines and for its design services, and intellectual property development and methodology.

Smith, who continues as chairman, says the San Jose programmable logic device supplier will not change its focus as a result of the change in command. "John's in-depth knowledge of the semiconductor industry, and the ASIC market in particular, will be invaluable in leading Altera's growth as we focus on system-on-a programmable-chip solutions for the communications market."

(See Nov. 27 story.)

TI talking big about
cable-modem chips

It has been less than 18 months since Texas Instruments entered the cable-modem business, but already the big chip maker is talking big, Texas-style.

Using Dataquest's market forecast, TI's Houston-based digital-signal-processor division is bragging that it will ship nearly 3 million of the 10-million chips that will be sold in 2000 for data-over-cable services. That would give the company a 30% global market share for DOCSIS integrated circuits.

By 2004, Dataquest expects worldwide cable modem shipments will reach 30 million units. Analyst Patti Reali expects North America to start shifting away from worldwide standalone cable modem units as the market shifts to integrated broadband access and voice devices.

(See Nov. 28 story.)

Fairchild isn't changing outlook;
sees 20% gain in sales next year

Despite slowing industry growth in some markets, Fairchild Semiconductor is sticking with its current forecast for the fourth-quarter. The Maine chip maker expects its quarterly sales to hit $490 million, which would be a 3% sequential gain in revenues over the third quarter's $476 million.

Bookings were soft in the first part of the current quarter as chip customers adjusted their backlogs to balance inventory levels for more modest growth in 2001. "These inventory-driven backlog adjustments are continuing," acknowledges CEO Kirk Pond. "We believe backlog adjustments will diminish toward the end of the quarter, and we believe we can book and ship enough additional orders to achieve our $490 million sales target," he says.

Adds CFO Joe Martin: "We continue to believe that this period of slower sequential growth is due to inventory adjustments in segments of the communications and computer markets." "For the past 18 months market forecasters have projected 2001 growth to slow to a still-healthy 20-plus-percent growth rate from the torrid 35% pace of 2000," Martin adds, "and our long experience has been that when growth rates slow even slightly, backlogs and inventories need to be adjusted."

Sales growth in 2001 will be about 20%, predicts Pond. "This is in line with the recent forecast issued by the SIA for the total semiconductor market," he notes. "This reflects our outlook that after the seasonally soft first quarter, sales will strengthen through the second quarter and second half of the year."

(See Nov. 29 story.)

Tokyo Electron jumps into market
for copper electroplating systems

Competition is going to heat up in the U.S. market for copper electroplating systems. Tokyo Electron, second largest chip production equipment maker in the world, is diving into the market. It estimates the global market for copper electroplating systems will reach ¥50 billion--or about $500 million annually--in 2003.

Currently, these electrochemical deposition systems are currently being sold in the U.S. by Applied Materials, Novellus, and Semitool. They are used to fill in copper wiring and vias, a new process that is replacing conventional aluminum metal interconnect technologies.

Tokyo Electron, or TEL, will be selling its copper electroplating systems through NuTool, a U.S. startup. The marketing, sales, and support pact covers chip makers in Japan, U.S., and Europe.

NuTool says this electroplating technology is capable of depositing planar copper films on the surface of wafers with low film thickness--50% of the feature depth or less. As a result, the technology significantly reduces problems of dishing, depressions in the copper interconnect, and erosion, the company claims. The thin planar surface also reduces requirements for chemical mechanical polishing (CMP), it says. The NuTool 2000 system already is being pilot tested at customer-investor sites, according to NuTool.

(See Nov. 30 story.)

How about a $100 set-top box?

Philips Semiconductors has a new set-top box chip that it claims will make possible a multi-function set-top box for as low as $100 in terms of bill-of-material costs.

Based on a RISC processor core from Mips Technologies, the new chip supports two-dimensional graphics, audio/video decoding, Internet browsing, MP3, personal video recording, and V.90-compatible software-modem, says Thierry Fautier, product marketing manager. "What we've developed is a chip for the mid-range segments of the set-top box market," he claims. It also supports major broadcast standards such as Digital Video Broadcasting (DVB), he adds.

The chip is the second part in Philips' new Nexperia Home Entertainment Engine line. In October, it rolled out its first Nexperia chip, a higher-end chip based on a dual-processor architecture and designed for high-end set-top boxes.

The chip also includes a digital signal processor that supports MPEG-1, MPEG-2, Dolby Digital, G729, and G729A. Priced at $20 each in 50,000-unit lots, the new chip will begin sampling in December, with production slated for the second quarter of 2001.

(See Nov. 27 story.)

Kawasaki and Sonicblue
invest in new wireless LAN

ComSilica--first spin-off company from Berkeley Concept Research Corp., a technology incubator company set up by Kawasaki Steel--is developing a next-generation broadband wireless technology for the IEEE 902.11a standard. Called WhizNet, it is being designed to serve home networking applications with data transmission speeds of up to 54 megabits-per-second in the 5 gigahertz band.

Kawasaki Steel and Sonicblue formerly called S3 both have invested in the Berkeley, Calif., startup and are hoping to use the technology to offer new products for next-generation wireless local area networks. The investment will give Sonicblue "preferential access to high-bandwidth wireless technology for use in our Access division's home networking products," notes CEO Ken Potashner, CEO of Sonicblue, which has refocused itself on home automation and Internet appliances.

Sonicblue is counting on the market for broadband and wireless local area networks to explode. Wireless LAN users are expected to total 23 million by 2003, nearly five times the 5 million in 2000, the company predicts. Revenues from wireless LAN are expected to grow to nearly $2 billion by 2003, up from just over $400 million this year, the company says.

Kawasaki Steel hopes that WhizNet technology will help it penetrate what is expected to be a huge Japanese market for wireless LANs.

(See Nov. 27 story.)

Moto says its new GaAs process
superior to current technology

Motorola Semiconductor figures it has developed a new process to build gallium arsenide circuits that's superior to the existing depletion-mode pseudomorphic high electron mobility transistor (pHEMT) devices. The chip maker says it has qualified a true enhancement-mode (E-mode) heterostructure field-effect transistor process for GaAs circuits.

The new process makes possible single-power-supply devices with "exceptional and cost effective performance in both linear transmitter and receiver circuits for low-power wireless applications," Motorola claims. The first products using the new E-mode GaAs technology are expected to be on the market in February.

"Because of its low off-state leakage current, this true enhancement-mode device eliminates the drain-supply switch required for depletion-mode pHEMT and metal semiconductor field effect transistor (MESFET) devices," says Karl Johnson, director of Motorola's compound semiconductor lab. The technology is "ideal for use in transmitter and receiver circuits, for both analog and digitally modulated portable products," the company says.

The technology is said to be ideal for use in transmitter and receiver circuits, and for both analog and digitally modulated portable products. The E-mode GaAs technology helps to reduce the cost and size of the end product by eliminating both the negative-voltage generator and the drain-supply switch within the handset power amplifier section, as well as eliminating additional passive components, the company claims.

Motorola already has built a three-stage power amplifier designed for a 1.900 gigahertz time-division multiple access wireless handset applications. And the E-mode technology has demonstrated excellent noise figure and linearity performance for small signal applications, the company says.

(See Nov. 27 story.)

Is Broadcom trying to be
the chip industry's Cisco?

Cisco Systems has certainly demonstrated that an excellent way to grow a network systems company is by swapping stock to acquire dozens of startups and other emerging companies. Now it looks like Broadcom is trying to do the same thing in semiconductors.

In its latest acquisition, the Irvine, Calif., chip company is paying $777 million in stock for VisionTech, an Israeli supplier of digital video/audio compression and depression chips. Its encoder chips are used in MPEG-2 applications for personal video recording, interactive videoconferencing, and Internet Protocol systems.

The four-year-old, fabless chip supplier's latest MPEG-2 video/audio encoder chip has been adopted by a raft of set-top box manufacturers, including Motorola Broadband Communications Sector, Scientific-Atlanta, Pace, Microsoft WebTV, and Replay TV.

Personal video recording (PVR) has really got Broadcom excited. "It's one of the most compelling consumer applications which will radically change the way people watch TV, allowing simultaneous recording and playback of live broadcast-without a VCR or videotape," points out CEO Henry T. Nicholas. "VisionTech's capability to compress live video in real time will be used across all of Broadcom's product lines for applications ranging from distributed video over home networking to IP video streaming over the Internet," he says.

"By 2005, PVR functionality will become an integral component of virtually every set-top box," claims Amir Morad, CEO of VisionTech. Cahners In-Stat also predicts that shipments of PVR systems will surge over the next few years. Between 2001 and 2003, the market researcher says, unit shipments of PVR-enabled set-top boxes are expected to increase 275% annually, reaching 8 million systems by 2003.

(See Nov. 28 story.)

IBM, Infineon, UMC
claim 0.13-micron lead

Are the floodgates about to open with an outpouring of the next-generation, 0.13-micron process chips? That's what the three partners said this week, claiming they had beaten their competition to the punch.

The process technology team--called the WorldLogic alliance--was an ambitious plan revealed 10 months ago to develop common process technologies for 0.13- to 0.10-micron logic ICs with integrated mixed-signal circuitry and embedded DRAM.

The partners--IBM, Infineon, and United Microelectronics--said this week that they already had begun fabricating advanced ICs with their process. They declared the technology "production ready" and now available to customers. UMC already has started offering multi-project wafers, which combine 0.13-micron IC designs on a single chip aimed at cutting costs in prototyping.

Dozens of customers are currently designing chips based on the compatible 0.13-micron logic processes, IBM claims. First customer shipments of chips for network communications and computing applications are expected early in 2001.

"This is a new development model for the industry," claims IBM VP Bijan Davari. "By cooperating on the development of the base technology, each of us can concentrate on our unique chip design and manufacturing capabilities to differentiate our offerings in the marketplace."

(See Nov. 28 story.)

NEC, Hitachi venture
will build 300-mm fab

The joint DRAM venture formed a year ago by NEC and Hitachi is moving right along. Early next year, the company--called Elpida Memory--will start up its own sales operation take on new product development activities. And this week, it disclosed plans to build a 300-mm wafer fab for volume production of 256-megabit memories using a 0.13-micron process.

Construction on the fab, which will be located in Hiroshima, Japan, will begin in January, with initial production slated to start in the first half of 2002.

The joint venture will invest ¥160 billion ($1.4 billion) in the new fab, and in addition, NEC and Hitachi will kick in another ¥20 billion ($180.2 million) in the joint venture to fund construction of the fab shell building. The new 300-mm plant will eventually have a capacity of 20,000 wafers per month equal to about 48,000 eight-inch wafers per month.

(See Nov. 28 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!.)