November 6, 2003 4:16 PM PST
AMD plans for more chips, fab
The chipmaker on Thursday sketched out its strategy for the PC and cell phone markets for the next two years at its annual analysts' meeting and disclosed its plans for another multibillion-dollar chip fabrication facility, or fab.
AMD could once again return to profitability in the fourth quarter, depending on what happens with flash memory, said company CEO Hector Ruiz. The company hasn't reported a quarterly profit since the second quarter of 2001.
"We expect the microprocessor business to be profitable, and we are looking at flash to get there, but it's up in the air," Ruiz said during a question-and-answer session at the company's headquarters here. AMD's two main products are microprocessors, used in PCs and servers, and flash memory used primarily in cell phones.
Ruiz further said that the first software that can take advantage of the 64-bit capabilities of AMD's most recent desktop chips is on the way. "The consumer-type apps will begin to pop up toward the middle of the year," he said.
The mood was in stark contrast to last year, when the company was reeling from massive losses, product delays and layoffs.
In 2003, however, the company launched its Opteron and Athlon64 processors, landed deals with companies such as IBM and Sun Microsystems to promote AMD chips in servers, and absorbed the lion's share of a flash memory joint venture with Fujitsu. From the third quarter of 2002 to the most recent quarter, revenue has climbed 88 percent, while expenses grew only 18 percent, according to AMD.
Computer world tour
Over the next two years, AMD will place more emphasis on getting its chips into servers and notebooks rather than desktops. Desktops likely will continue to account for the bulk of AMD's revenue and margin--the amount of money left over after expenses--but the opportunity to expand lies in the other two markets, said Dirk Meyer, senior vice president of the company's computational products group.
In the second half of the year, the company will introduce Odessa--a high-end notebook chip made on the 90-nanometer process--and Dublin, which was designed for inexpensive notebooks and made on the older 130-nanometer process.
Oakville and Trinidad will then come to the notebook market in 2005. Both are 90-nanometer chips. A nanometer is a billionth of a meter, and the designation means that the average feature on a chip measures 90 or 130 nanometers. Chips with smaller features are generally faster and provide greater performance.
In desktops, the second half of 2004 will see the introduction of San Diego, a 90-nanometer chip for game PCs; Winchester, a 90-nanometer chip for midrange boxes; and Paris, a 130-nanometer chip for budget buyers. Toledo, a high-end chip, and Palermo, an inexpensive one, will then follow in 2005.
Chips made on the K-9 design will appear around the same time, Meyer said.
As for servers, Athens, Troy and Venus will be released in the second half of 2004, while Egypt, Italy and Denmark will appear in the second half of 2005.
While the company said it's seeing good results with 90-nanometer manufacturing, AMD chief scientist Bill Siegle conceded that it's late in mass manufacturing chips with the new process. The company earlier had expected 90-nanometer chips to be ready by the end of this year.
AMD also will experiment with producing processors for "a new class of PC-type devices" that will cost even less than today's desktops, Ruiz said. These devices likely will contain AMD Alchemy chips, which are based on a different architecture. A Chinese manufacturer is already using Alchemy chips in this regard on an education project.
Siegle said AMD will disclose the location of a 65-nanometer fab in the next few weeks. The 65-nanometer facility, which will process larger 300-millimeter wafers, needs to be ready in about two years, relatively rapidly, he said. AMD will not build the facility alone, but will share the cost with another manufacturer.
"We are not in a position to share the details, but we have a site selected and groundbreaking scheduled," Siegle said.
Although company executives would not disclose the location, one of the more likely places appears to be Singapore, where it still may own land. The company was going to build a 65-nanometer fabrication facility with UMC there, but the deal fell apart early this year.
Another possible location is New York. IBM and AMD researchers already work together on research projects in IBM's East Fishkill facility. New York is aggressively courting semiconductor makers. On Wednesday, at a dinner in Silicon Valley, New York Gov. George Pataki said that the state has created "shovel ready" manufacturing sites complete with water, power and transportation facilities.
Other possibilities include China, which is offering major tax incentives--usually the overriding factor in determining where a factory gets built--to manufacturers. The difficulty with China is that cutting-edge manufacturing tools have to be snuck in through U.S. subsidiaries to avoid running afoul of U.S. trade laws.
Two other strong possibilities are Dresden, Germany, and Austin, Texas, where AMD already runs existing facilities.
In flash memory, the company will try to cut costs and expand its market share by shrinking the size of its chips, especially its MirrorBit flash parts. Shrinking the chips will essentially allow the company to make twice as many or more chips at once for around the same price. Intel and AMD are neck-and-neck in flash memory, according to statistics from AMD.
"We want to double our capacity in the next 12 months," said Bertrand Cambou, CEO of the AMD-Fujitsu joint venture.