If you look at the semiconductor industry from 30,000 feet, three new mountains stand out. The first is that the basic materials are changing, bringing headaches aplenty. The transition to high-k gate oxides promises to be difficult and may go hand-in-glove with a switch to metal gate electrodes. Silicon-on-insulator and strained silicon will be sought by circuit designers who don't want to leave performance on the table.
The second mountain is the rise of the pure-play foundries.
The third is that logic chips are becoming system-on-chip designs, the complexity of which requires a design-for-reuse intellectual-property strategy.
With those challenges defining success, keep a close eye on IBM's decision to declare itself a "high-end foundry."
"We want to be a major player as a high-end foundry, using our process leadership and . . . partnering with leading-edge fabless companies," said Michael Mayer, general manger of IBM's Microelectronics Division.
One difference between IBM and the successful foundries is that Big Blue sells IBM-branded chips (although the sale of IBM's cell phone chip set unit to Triquint is part of the new strategy). Will IBM compete with its foundry customers in key product areas?
"There was a point where we tried to assess whether we could penetrate the standard-product areas mainly in networking and communications," Mayer said. "As we have matured in our understanding of the market, in the standard products area, we want to concentrate on PowerPC and related product areas, like Infiniband and network processing."
Can IBM keep its lead in process technology? Probably, but with the inherent risks that leadership entails.
According to one reliable source, in the later stages of the transition to a low-k dielectric, IBM engineers discovered problems with mechanical stress failures in the vias. Although IBM is getting its SiLK dielectric back on track, the reliability issue forced some key IBM ASIC customers to redesign their libraries to IBM's less-aggressive fluorinated silicate glass (FSG) dielectric, setting back their schedules by several months. (An IBM spokesman said its customers were well-prepared for the switch to FSG.)
Another challenge: While IBM has great design tools and a formidable library of internally developed IP, it is playing catch-up in bringing commercial EDA and IP resources to designers.
IBM's decision has major implications. Success will depend on its building the 110 percent commitment to the foundry model that any winner must have.
Send feedback to dlammers@cmp.com.
RSS
RSS| 
Digital| 
Mobile
