"2002 will largely be defined by our success in the mobile and server space, and holding ground in the desktop space," CEO Jerry Sanders said at the company's annual analyst meeting Thursday, where AMD also revealed its product plans for the coming year.
"Instead of a (Microsoft-Intel) duopoly, we are going to have a holy trinity," he said.
The Sunnyvale, Calif.-based chipmaker's strategy will focus on lowering manufacturing costs and touting its chips through the mass of second-tier manufacturers and dealers, the so-called white-box segment of the market that is showing growing resilience, especially overseas.
"The white box market seems to be doing relatively better than some of the OEMs (original equipment manufacturers)," said Sanders, adding that in servers, the company is targeting "all white-box guys."
Intel also touted the importance of white-box sales, especially internationally, in its analysts' conference call in October.
As for costs, AMD said it would reduce overhead by trimming back operations. Expenses associated with moving to 300-millimeter wafers, for instance, will be cut because the company won't move to the standard until 2005, and then will do so by sharing the cost of a new fabrication facility with a partner.
In addition, the company said it would aim to undercut Intel by producing chips that are significantly smaller.
AMD will kick off 2002 with the release of "Thoroughbred," a version of the Athlon XP chip made on the more advanced 130-nanometer manufacturing process. The chip will cover 80 square millimeters in area, or 65 percent of the space of the "Northwood" Pentium 4 coming from Intel in early January. That chip measures 116 square millimeters, according to AMD estimates.
Both companies will shrink their chips again in 2003 by adopting the 90-nanometer manufacturing process, but AMD will still maintain an advantage, company executives claim. (The nanometer measurements refer to the average size of features on the chip.)
"We believe we have a 10 percent to 20 percent advantage on cost," said Hector Ruiz, president of AMD.
Along with keeping its chips small, AMD has other ambitious plans. In the fourth quarter, it will come out with a 1.3GHz Athlon and a 1.2GHz Duron for notebooks.
The first quarter will mark the appearance of the first Athlon and Duron notebook chips adopting the so-called model-number brand scheme, which downplays megahertz. In the first quarter, an 1800 Athlon for notebooks will come out, which will likely run at 1.5GHz, as will a 1500 Duron, which will probably run at 1.3GHz. In the second quarter, a 1900 Athlon for notebooks will emerge.
All of these chips will be based on the processor cores of Thoroughbred, or Appaloosa, a discount version of Thoroughbred.
These new chips will also consume less heat than current AMD notebooks chips. Combined with new packaging, the new chips will allow AMD to break out of the "desktop replacement" segment of the notebook market and get into the increasingly popular "thin and light" notebooks, said Dirk Meyer, group vice president of the computing products group.
In desktops, a 2200 Athlon that will likely run at close to 1.8GHz will hit the market, to be followed by a 2400 and a 2600 in the second and third quarters, respectively. Toward the end of the year, a new version of Athlon, code-named Barton, will appear. Barton is essentially the same as the Thoroughbred Athlons but contains IBM's SOI (silicon-on-insulator) technology, which better insulates transistors and is used to increase performance or lower the power consumption of a processor.
The plans for server chips will mostly follow the desktop plans in terms of speed, but will add capabilities for multiprocessing. A multiprocessing Athlon for servers will come out in the first half; a Duron version will appear in the second half.
The first Hammer chips will then appear commercially in the third quarter of 2002. The Hammer family contains a different architecture than the Athlon family and can read both 32-bit and 64-bit applications.
The desktop/notebook version of Hammer, called Clawhammer, is in the final stages of design, Meyer said, and will appear in engineering samples in the first half of 2002. It will cover 64 square millimeters on the 90-nanometer process. The first version will carry the model number 3400.
"Our eighth-generation Hammer product will, through architectural enhancements, materially outperform our seventh-generation Athlon," Meyer said.
Figuring out the chip speed by subtracting 300 or 400 from the model number, however, won't hold with the Hammer family. "The gap between the frequency and the model number will increase a little bit," Meyer said.
Sledgehammer for servers will come in the first half of 2003, representing another delay. Originally due in the first half of 2002, the chip was pushed to the second half of 2002. Sledgehammer will differ from Clawhammer in that it will contain more cache and be aimed at servers.
No AMD conference would be complete without some Intel trash talk, and Sanders happily obliged. The cross-freeway rival is in trouble, he claimed, because of the problems inherent in the designs of the Pentium 4 and Itanium, the 64-bit server chip. The Pentium 4 is expensive to make and underperforms Athlon, he said.
"I think Intel is trapped. When they designed the Pentium 4, they didn't expect to have a real competitor in the marketplace with a meaningful alternative," Sanders said. "They always underestimate us, and being underestimated is a good thing."
Itanium, meanwhile, will suffer because of the difficulty of producing software applications for the chip. Itanium uses a different architecture than other Intel chips. Hammer, meanwhile, is compatible with existing 32-bit applications.
"My biggest fear is that Intel will come out with a 32-bit processor with 64-bit extensions because it is the right thing to do," Sanders said. "The Itanium it turns out is a niche product...We are going to have a role in the industry because we better fulfill Microsoft's needs."
Life won't be easy next year, Sanders and other executives noted. AMD likely won't return to profitability until the second quarter. Flash-memory prices will also continue to decline because of overcapacity in the industry. Flash revenue may not increase at AMD until the third quarter.
The effort to get into servers will also eventually require endorsement from a major manufacturer, Sanders added.
"At the end of the day, we need to get a Compaq, Dell or HP," he said. "IBM is going to be tough."