Gathered below are the top nine stories for the week beginning Sunday February 22, 2004, as ranked by Silicon Strategies' readers, up to and including Saturday February 28. The ranking is based on the number of reader "views" or "hits" on a particular article.
1) IC industry sees shortages in '04, followed by downturn in '05 .... Worldwide wafer demand is expected to grow by 9.2 percent from 2003 to 2008 -- a level that raises concerns about capacity shortages in the short term and a downturn in the long term, according to a report from Semico Research Corp. issued Thursday.
2) Intel seeks to regain No. 1 position in flash, plans to triple capacity .... After stumbling in the flash-memory business, Intel Corp. is aggressively moving to regain its footing in the arena, with plans to take back lost market share from the likes of Samsung, Toshiba, and the Fujitsu/Advanced Micro Devices joint venture.
Intel may get unhappy if it is not number one, but its rivals will not give up market share easily, however much manufacturing capacity Intel throws at the problem.
Intel found itself making NOR, the less popular of two flash formats, letting Samsung and Toshiba leap over it in the 2003 rankings with their NAND flash sales. And according to observers, Intel has cost disadvantages compared with the compact non-volatile memory cells coming in at Spansion and Infineon courtesy of Saifun Semiconductor of Israel.
Of course, Intel has its multi-level cell technology Strataflash, but it seems to be unable to scale the number of levels per cell.
Secondly it can try accelerate moves to leading-edge manufacturing technology. But that buys a die size advantage only until the other manufacturers catch up.
The third thing Intel can do is to start making application-specific packages of processors and flash memories such as for the wireless handset business. That works best if a company effectively controls the architecture -- as Intel does in the PC space -- but even then the tactic is risky as it could be perceived as tied selling and declared anti-competitive. And in wireless Intel does not control either handset or PDA architectures.
The fourth thing Intel can do is ramp up the factories, bomb the price and have deeper pockets than the competition. If that is Intel's strategy it's good news for electronic equipment builders and bad news for Samsung, Toshiba Spansion, Infineon and the rest.PDC
3) Analysis: Looking into Intel's crystal ball .... In which Larry Greenemeier, of Information Week ponders where Intel is going over the next ten years.
For years, Intel has attempted to expand beyond its core microprocessor business. It is currently pushing hard in wireless, wireline and other markets, but the company still relies on its bread-on-butter processor business for growth.
In the next 10 years, the microprocessor may--or may not--be the engine of growth for the chip giant. So it's critical that the company find new markets to expand. After all, the PC business has been slowing down for a long time. Meanwhile, in the future, the "convergence" debate may come around again, where the PC vs. the TV will vie for the central computational center in the home. The PC won the battle in the first round, but the TV may win the eventual war over time.
Where will Intel be in 10 years? Intel CEO Craig Barrett makes a bold prediction. "Ten years out, it could be health sciences," he says. "The transistors that Intel makes today are already the size of DNA molecules or viruses. Doctors and researchers can start to think about coupling sensors, computers, and communication capabilities to diagnose and treat diseases on a molecular level. I would guess in 10 years--a wild guess--that this could be as big to our industry as the PC is today." ML
4) World's DRAM makers face criminal charges over price-fixing, say reports .... Prosecutors are investigating an alleged price-fixing conspiracy among the world's major DRAM makers that may have occurred before a sharp increase in prices two years ago, according to online reports that referenced the Thursday edition of the Wall Street Journal.
Not long ago, a sales rep from Micron was allegedly charged for destroying evidence in the DRAM price fixing scandal. However, it was really hard for me to believe that one only person from one company was involved in the DOJ's investigation. I wouldn't put it past the DRAMers to collectively fudge prices, especially during the worst downturn in history.
Alas. It looks like the big players were allegedly involved in something. The world's top four DRAM makers -- South Korea's Samsung Electronics Co. Ltd. and Hynix Semiconductor Inc., Micron Technology Inc. of the United States, and Europe's Infineon Technologies AG -- are said to be under investigation, "amongst others", according to reports.
5) Agere foundry supplier still dropping prices, says report .... As communications chipmaker Agere Systems Inc. sheds its own manufacturing capacity, it is not seeing any problems with tightness of foundry supplies, nor with negotiating prices down at one of its foundries, according to a Reuters report that quoted John Dickson, Agere's chief executive officer.
6) Optical litho to last another decade .... Optical scanners based on the 193-nm wavelength are expected to last for 10 years or more in production fabs -- a development that would eradicate the need for 157-nm and even extreme ultraviolet (EUV) lithography, according to an executive from KLA-Tencor Inc.
KLA-Tencor litho guru Chris Mack believes that 193-nm will be the last wavelength for lithography and said there is no need for EUV, 157-nm, and other NGL technologies. Thanks to immersion, 193-nm litho could extend to what some people call the 32-nm node on the ITRS roadmap.
Beyond that, the "optical forever" crowd is not promising anything. Intel sees the 32-nm node in 2009. But still others see Moore's Law slowing down and the 32-nm node will not show up for a long, long time and much later than 2009.
In a decade, however, some believe the semiconductor world could turn to carbon nanotubes or other exotic technologies, which will require a whole new set of chip-production technologies. ML
7) Gartner tags 2004 chip market growth at 23....or more than 30 percent .... The worldwide market for semiconductor chips is projected to reach $217 billion in 2004, a 22.6 percent increase from 2003, according to preliminary quarterly estimates from Gartner Inc. That is unless lengthening lead times become "pervasive", then it could be more than 30 percent growth, the forecaster said.
Fuzzy math has entered into the semiconductor market. Or is chip forecasting an inexact science? Either way, I'm confused by Dataquest's latest chip forecast. Makes no sense to me. The research firm says the market will grow 22.6 percent in 2004. Then, it makes another forecast. (FYI. Silicon Strategies predicts 22.7 percent growth for ICs in 2004.)
8) Nvidia chips to move to TSMC 0.11-micron process .... PC graphics chip vendor Nvidia Corp. said Tuesday (February 24, 2004) that it would be manufacture some of its forthcoming graphics processing units (GPUs) on a 0.11-micron manufacturing process at foundry Taiwan Semiconductor Manufacturing Company Ltd.
TSMC made a sensible decision to offer a 110-nm process. The company is reportedly the first foundry to offer the process, which is a shrink of its 130-nm technology using fluorinated silicate glass based dielectrics.
The technology is more or less a bridge solution to company's 90-nm technology, which is currently in so-called "risk production." Some chip makers may feel more comfortable moving to 110-nm first, then diving into 90-nm. With mask costs going out of sight, the 110-nm technology seems like a safe bet for some. ML
9) Nikon tips a $25 million, high NA stepper .... Nikon Corp. said Monday that it has developed a 193-nm wavelength lithography machine that will ship with the world's highest numerical aperture lens, making it suitable for 65-nm manufacturing processes.
When I first followed lithography back in the mid-1980s, a high-end reduction stepper sold for a grand total of $1 million--each. That was steep back in those days. In sharp comparison, the high NA, 193-nm scanner, which was rolled out by Nikon this week, will sell for a cool $25 million"each!
Here's even more shocking news for chip makers: If or when EUV hits the streets, the price for one tool is expected to range from $35-to-$70 million per unit. Besides Intel, who can afford these EUV monsters?
Karen Brown, a lithography expert from Idea Industries, put another shocking perspective on litho costs. Lithography costs represent 30 percent of the capital spending for today's fab, but that number is expected to increase over time.ML
Monday, February 23
Wacker Siltronic signs Soitec SOI license deal .... Siltronic AG, the silicon wafer division of Wacker Chemie GmbH, has signed a license for the Smart Cut method of preparing silicon-on-insulator (SOI) wafers from Soitec SA.
The plot thickens in SOI. Siltronic has licensed SOI technology from Soitec. The German silicon wafer maker reportedly is also a licensee of Ibis' Simox-based SOI technology.
Meanwhile, SEH has reportedly bought Ibis' SOI-enabled implanters. And MEMC has licensed SOI technology from SiGen.
Despite all the announcements, Soitec's bonded-wafer approach is reportedly winning the SOI race. AMD is using the approach with its processors, while IBM is reportedly looking at the technology. Intel has been on again and off again with SOI. Word on the streets is that it's on again at Intel.
Tuesday, February 24
European, Japanese transistor models vie to ease out BSIM .... CMOS field effect transistor models from Europe, and Japan are being developed to try and cope with problems scaling the established U.S. model format to the 90-nanometer manufacturing process.
Friday, February 27
Renesas files patent suit against Taiwan's Nanya .... Japan's Renesas Technology Corp. has filed a patent infringement suit against Nanya Technology Corp. Japan, a subsidiary of Taiwanese DRAM maker Nanya Technology Corp.