Everyone is cutting
1st quarter forecasts

The current downturn in the chip business clearly wasn't expected by the chip industry. This week, 13 more semiconductor companies either downgraded earlier forecasts or reported a drop in sequential sales. In every case, the culprits were the same: inventory correction in the supply chain and a sharp decline in U.S. customer end-demand. Equipment makers got hit hard by order delays and cancellations from their chip maker customers.

So how long will it last? Chip makers sure don't know. They're so spooked by their bad forecasting that no one is willing now to predict a turnaround in business. Even Intel executives say they have no idea if or when a recovery will take place this year. Yes, it definitely is an industrywide downturn, but like I said last week, chances seem to be improving that the slowdown will be relatively short, albeit brutal.

Even big momma underestimated how quickly the chip business was falling out of bed. Intel, which previously had expected its first quarter revenue would fall 15% from the previous quarter, this week downgraded its forecast to a 25% sequential drop. That would even be a 19% decline from the year-ago quarter. Not a pretty picture.

Intel's business is down across the board. It is experiencing weak demand in both its PC and non-PC markets. Even its communications chip business, including flash and networking silicon, is weak. As a result, the chip giant is stepping up its major cost-cutting program, including the elimination of 5,000 jobs over the next nine months by attrition. This amounts to 5.7% of its nearly 90,000 employees.

Even R&D spending will be cut back by Intel, reducing its budget this year by $100 million to $4.2 billion. This is down 2.3% from last year. But Intel claims it will not cut its capital expenditures this year, which will amount to a record $7.5 billion.

Big foundries fall off cliff,
dropping as much as 28%

(See March 8 story.)

If you can't figure out short-term
outlook, just look out five years

Many market researchers seem always to take the long view--how a market will grow over the next five years, for example. That way they don't have to get involved in trying to make the tougher calls on short-term market moves--even though everyone is complaining now about the lack of short-term visibility.

Right now, a lot of people--me included--would love to see a forecast for the silicon foundry business in the next year or so. So what do we get? Right, a five-year forecast starting back in 1999. A new report from Cahners In-Stat declares that foundry revenues will increase by nearly five times between 1999 and 2004, despite the current slowdown in demand for processed wafers. "The foundry segment is expected to outperform the industry as a whole with foundry revenues expected to rise from $7.6 billion in 1999 to $35.4 billion in 2004," predicts In-Stat's Steve Cullen. Pure-play foundries will continue to push advanced process technologies ahead of integrated device manufacturers (IDMs), which produce their own chips, he says. Foundries also will lead the industry's transition to 300-mm wafers from today's mainstream 200-mm substrates, Cullen says. According to the glowing report, the need for foundry wafers will continue to grow faster than the rest of the industry because there are 500-to-600 fabless chip startups out there that need more production capacity. In fact, some IDMs have decided to convert older wafer fabs into foundry plants to tap this growing potential, In-State notes. It seems to be pretty late in the game to be doing that kind of thing.

(See March 6 story.)

January IC sales fall 5.7%,
but SIA is still optimistic

No surprise in the SIA's January data. Worldwide chip sales dropped 5.7% to $16.9 billion from $17.9 billion in December, reflecting the industry's excess inventory problems and a slowing of demand.

But the SIA continues to maintain its traditional optimistic outlook. "Current forecasts suggest the inventory adjustment will be completed by the end of the third quarter, and end-market product demand will improve later in the year," declares president George Scalise.

All regions slipped in January from previous month's sales. Sales in the Americas dropped 5.8% to $5.23 billion, while they fell 5.9% in Japan to $4.04 billion. The largest regional drop was recorded in the Asia Pacific market, which declined 7.9% to $3.97 billion. European chip sales fell only 2.7% to $3.62 billion.

Year-to-year, the best numbers were out of Japan, soaring 23.2% from year ago sales. Americas grew 15.4% over January 2000 and Europe climbed 14.6%. Poorest year-to-year gain came from Asia Pacific, which showed a revenue increase of only 2.9% over last year.

(See March 7 story.)

248-nm litho tools may be
in use for another four years

There may be more life in the old boy yet! Would you believe that 248-nanometer lithography tools may be cranking out products for another four years? Researchers at MIT's Lincoln Labs now think so. They used 248-nm lithography to produce a 0.009-micron (9-nm) device to show just how far today's exposure tools could go by using subwavelength photomask technologies.

This means that chip makers might be able to extend the life of their existing 248-nm exposure tools to perhaps 2005, pushing out the need for advanced, more expensive 193- and 157-nm scanners now in development. That may even include a delay in putting to work next-generation lithography (NGL) technologies such as extreme ultraviolet (EUV) and electron projection lithography (EPL).

MIT researchers claim they also set another world's record. Using Canon's EX-6 line of 248-nm lithography tools and a chromeless phase-shift mask from Photronics, they demonstrated a device with 210-nm pitch grating (105-nm lines and spaces) design rules.

"There is plenty of life left in 248-nm lithography tools, with strong resolution enhancement technologies," claims Lincoln Labs researcher Michael Fritze. "The 100-nm node is possible using high-NA (numerical aperture) 248-nm sources with resolution enhancements such as strong phase-shifting photomasks," he maintains.

This may be good news for chip makers, but it isn't welcomed by lithography suppliers who have invested a lot of money in their 193-nm scanners and are eager to start shipping them. Leading scanner vendors maintain 248-nm systems are quickly running out of steam.

(See March 7 story.)

DMD is 24 years old, cost TI
a mint, but now it's smoking

It seems like I've followed Texas Instruments' DMD (for Digital Micromirror Device) forever. It started 24 years ago when the Dallas company got a government contract to build an analog light modulator called Deformable Mirror Device.

TI has spent one heckuva lot of money on this technology, not willing to give up despite a highly elusive commercial market. It is now consists of an array of aluminum mirrors atop a layer of memory cells. The spatial light modulator receives images as digital input, stores the image, and then projects the image when light is reflected onto the mirrors.

I was surprised this week to learn that TI claims DMD technology is now a profitable part of TI's Digital Light Processing display business. The DMD chips are taking off in large-area projection display applications, and they are moving quickly into consumer televisions as well as movie theaters. This light-processing business amounted to just under $400 million in 2000 revenues, according to TI executives.

"This is not a laboratory curiosity now," declares CEO Tom Engibous. He figures TI has made a trillion mirrors since the early 1990s.

Now TI is beginning to push the micromirror technology toward a potentially huge market--optical switching in next-generation networks. It is looking at integrating multiplexers and equalizers on high-volume DMD chips, which would replace today's hybrid opto components and other parts used in fiber optic networks. These optical switching chips would be turned out by a 200-mm wafer fab just like any other semiconductor, according to TI.

Sources indicate that TI has already demonstrated DMD optical switching prototypes and concepts to such major transmission equipment makers as Siemens. The German giant wanted to buy DMD devices, TI says, but the Dallas company says it isn't ready to commit to volume production. It wants to make sure the switching devices are based on the same technology as TI's existing DMD display chips.

(See March 5 story.)

Toppan achieves
photomask milestone

Japan's Toppan Printing has achieved a critical photomask milestone. It developed a prototype photomask for 0.05-micron process technology and is now working to implement the technology by 2005.

The photomask, developed for Advantest, is still in the R&D phase, but Toppan is convinced that the prototype will be ready for volume production within four years. It is expected be suitable for use with Advantest's direct electron-beam, block cell technology, Hitachi's cell projection mask approach, and for Nikon's electron-beam method at the 50-nm process node.

Toppan, which had a 22% share of the $1.5 billion non-captive photomask market in 2000, claims its production masks are in a leading position to be verified for 0.13-micron production and is already entering volume production at that process node, said Kumamoto.

Toppan's biggest rival, Dai Nippon Printing, also is pushing hard to meet the same goals and says it's investing $42 million to build a factory in Kyoto to produce masks for the 0.13-micron design rule. Dai Nippon expects to increase the 20% market share it held last year to 25% this year.

(See March 5 story.)

Dow Chemical extends SiLK
process to 0.13-micron process

Going to copper wiring can drive up chip performance over conventional designs using aluminum interconnect and a silicon dioxide insulator by as much as 20-to-30%. Substitute SiLK insulation for silicon dioxide in that same copper-based chip will give you an additional 20-30% gain in performance, researchers estimate. So chip makers want to switch to a new low-k insulator when they move to 130-nm (0.13-micron) copper process.

Dow Chemical established SiLK as a leading low-k candidate when IBM said last year it would use the film in its upcoming 0.13-micron copper process. But in recent weeks, TSMC, Motorola, and AMD said they would use Applied Materials' chemical vapor deposition (CVD) technology instead for low-k dielectrics. Advocates maintain CVD films are easier to integrate into interconnect process steps than Dow's spin-on organic films.

But Dow Chemical isn't about to give up in the low-k dielectric battle. The Midland, Mich., company says it has successfully migrated the SiLK resin to the next technology node by introducing porosity in the material. It has extended its spin-on SiLK dielectric resin to an ultra-low-k constant value of 2.0 for ICs fabricated with 100-nanometer feature sizes. It claims it's the first supplier to develop a viable spin-on organic film with low-k values down to 2.0 and the 100-nm technology node.

"Chip makers are already looking to low-k materials for 100-nm device production because extendibility is a decisive factor in their material choice at 130 nm," said Mark McClear, business director of Dow's Semiconductor Fab Materials group. "Process technology must be extendible to multiple device generations to maximize chip makers' return on investment."

Sample films of the new SiLK material will be available in the second quarter of 2001 for testing and process development. Commercial production is now slated to begin later this year.

(See March 5 story.)

ASML stumbles on Congressional
concern over its acquisition of SVG

Did ASM Lithography Holding take a big chance last fall when it agreed to acquire Silicon Valley Group? The deal for $1.6 billion in stock would make the Dutch company the world's largest supplier of lithography tools for semiconductor production.

Early on, industry observers speculated the purchase would be very sensitive politically in the U.S. because of SVG's advanced step-and-scan exposure tools, the U.S. government funding it received to develop scanner technologies, as well as its position as the only major U.S. supplier of advanced lithography tools for volume production.

They were right. ASML acknowledged this week that its proposed acquisition has been slowed by concerns in the U.S. over national security and defense-related photolithography technologies owned by SVG. The Congressional committee on foreign investment in the U.S. notified the two companies that it will conduct a 45-day review under the Exon-Florio review process.

"We are obviously disappointed by this brief delay but pledge to work closely and cooperatively with the U.S. government in its review," comments CEO Doug Dunn. "We are optimistic that upon completion of this review, we will be in a position to close this merger promptly."

(See March 8 story.)

Intel soon will break out
its communications sales

It looks like Intel is about to break out its communications business as a separate profit and loss center as early as April. That could make it a lot easier for market researchers, investors, and competitors to determine how the chip giant is doing in this business. It may also help Intel's sagging stock price. Right now, this business is included in the chip maker's "other" category, making it difficult to figure out how this business is performing.

"I wouldn't be surprised if we didn't report something at Intel's annual analyst briefing in April," says Ron Smith, who runs the wireless communications and computing group.

This may be coming because the FCC requires that any business generating 10% or more of its revenue must be broken out as a separate reporting unit. But Smith maintains that "we're not just looking at the accounting rules, but are considering what makes best sense for our business." Intel could probably avoid the FCC mandate by dividing its communications business into small units, each representing less than 10% of corporate sales.

Dataquest figures Intel had communications-IC sales of $2.4 billion in 1999. Some 45% of that is in flash memory, which is being aimed almost exclusively now at communications applications. No has estimated that total for 2000, but based on flash sales of $2.7 billion last year according to IC Insights, Intel could have had communications-IC sales of around $5.5 billion last year.

Intel needs to shine some light on this emerging business. Some analysts question whether Intel has the management culture to enable the company to succeed in communications over the long haul. "There is a different mind-set in the computer world than there is in communications," comments analyst Will Strauss at Forward Concepts. "In the PC market, Intel is the de facto standard and can do anything they damn well please." But, he says, "in communications you have to have standards in place before anything happens. That's not to say there aren't people within Intel that don't understand that, but overall, Intel doesn't."

(See March 6 story.)

KLA-Tencor launches first in-situ
control system for copper polishing

Running a copper interconnect process on a volume production line is one tricky job that requires new kinds of measurement and control systems. KLA-Tencor has just rolled out one of them that it claims is the industry's first production-worthy, in-situ film thickness and end-point control system for copper metal polishing in chemical mechanical planarization (CMP) tools.

The new Precice unit will permit fabs to improve their copper interconnect processing steps by automatically compensating for variations in wafers during CMP polishing steps, which are used to level off metal layers in dual damascene processes, the company says. The ability to measure and control each wafer in CMP processing will enable chip fabs to improve their yields and ramp volume production of copper interconnect more quickly, it adds.

"Precice incorporates several unique features to provide real-time, accurate film thickness measurements during CMP," says Pete Nunan, vice president of technology development. The new system will enable the CMP tool to "dynamically adjust for film variations to ensure proper control of the CMP process at all times. This minimizes the yield-limiting defects that occur from over- or under-polishing wafers such as dishing, erosion and metal residuals," he adds.

The new measurement system combines optical and eddy-current technologies in order to overcome the limits of existing "optical only" in-situ metrology systems in CMP tools, the company claims. In-situ metrology embeds sensors inside the CMP tool to take measurements of film thickness and uniformity.

Today's in-situ "optical-only" systems provide only limited information, the company says. The major disadvantages of optical-only methods, it says, are unreliable end-pointing due to previous level pattern noise when polishing upper metal levels, and the inability to provide real-time CMP removal rate, and uniformity information.

(See March 6 story.)

Japanese lab close to making
single-electron switch circuit

This research may not result in a commercial product for another decade, but work now underway in Japan could lead to a new class of semiconductor devices that could have a major impact on the market.

After more than a decade of work, Japan's Institute of Physical and Chemical Research in now just months away from developing Japan's first single-electron, tunneling transistor that's capable of operating at room temperature. Masakazu Aono, head of the institute's surface and interface lab, is confident he can build his own single-atom switch circuit this spring or summer.

The tunneling transistor will make possible a 10 cm² microprocessor of terabit density, Aono claims. "This transistor will be three magnitudes of order smaller than the gigabit limit for MOS as predicted by Moore's Law," the researcher notes. "Fifty years ago Bell Labs created the first transistor and at that time no one imagined the present scale of integration," points out Aono.

Aono is developing a switch circuit consisting of three, 3-nanometer-wide polydiacetylene "wires" that act as a source and a drain, each separated from a well containing a 10-atom-diameter cluster of 500 silver atoms that acts as a capacitor. The gate lies on the other side of the separation and rests on a graphite substrate. When carrying a single electron, current can flow through source and drain, Aono says, but 1 volt placed at the gate adds a second electron to the capacitor, thus raising its potential and closing the circuit.

The lab, which now is using up the last of a five-year, $3.5 million grant, has just been given three more years of financial support. And more help could be coming from outside Japan. UCLA and Cambridge University, as well as two major U.S. chip makers, are said to be interested in co-development.

(See March 7 story.)

On Semiconductors will
expand China operations

The sirens from China continue to lure more investment dollars from U.S. chip makers to what has become the industry's final frontier. On Semiconductor plans to invest $100 million to expand its Chinese manufacturing joint venture in Sichuan province and to establish a design center in Shanghai. The new facilities will design, manufacture and test power management ICs as well as the firm's existing discrete semiconductors.

On Semiconductor was slated to sign an agreement this week with Leshan Radio to build a new 6-inch, 0.5-micron facility for analog ICs in the central Sichuan city of Leshan. The joint fab would become the first 6-inch production line in Sichuan and western China. An industry source said plans are pending for another 6-inch fab in Sichuan's capital city of Chengdu.

Back in 1995, when On Semiconductor was still part of Motorola, Leshan Phoenix Semiconductor was set up as a joint venture with Leshan Radio. So far, it has invested $280 million in the company. Its 1,600 employees now turn out primarily surface-mount discrete semiconductors for cellular phones and pagers. Arizona-based ON Semiconductor is now the largest shareholder, holding 51% of the company. Motorola still holds 10% and Leshan Radio owns the remaining 39%.

(See March 7 story.)

Growth of wireless Internet users
using CDMA will double this year

Why am I not surprised? The backers of code-division multiple access (CDMA) technology are super bullish about the growth of wireless Internet users using CDMA. These subscribers will more than double this year, according to the CDMA Development Group, a trade group promoting CDMA.

You have to wonder why, though, since later this year, carriers are expected to deploy wireless Internet services based on a pair of incompatible, third-generation (3G) standards: cdma2000 backed by Qualcomm, and widebandCDMA (W-CDMA), backed by Nokia, NTT, Motorola. It reminds me of the battle of the two different VCR standards, or the non-compatible stereophonic sound systems in the 1970s.

More than 22 million subscribers are now using wireless Internet services on existing CDMA networks. This is nearly 30% of the worldwide CDMA subscribers, according to the trade group.

Despite their incompatility, the two new CDMA services being introduced are expected to the growth of CDMA-enabled wireless data in 2001 and beyond. Unit sales of Internet-ready wireless phones will shoot over the 1 billion mark by 2004, CDG says. And by the end of 2002, nearly all wireless phones will be preloaded with minibrowsers and will be Internet-enabled, they predict.

By year's end, there will be more than 50 million cdmaOne and cdma2000-based wireless Internet users, the trade group predicts. This growth will come from the existing cdmaOne markets of North America and Asia, coupled with the continued expansion in Latin America and China.