Head full of numbers

My head is full of numbers. That's not unusual this time of year, given the seasonal blossoming of economic forecasts, market analyses, and corporate earnings as well as the inexorable drill-down of various technology nodes. Sorting out the annual digit-o-rama has left me more flummoxed than usual, and I need some relief.

The photomask plays an increasingly crucial role in the semiconductor community's efforts to follow the numbers and obey Moore's law. International Sematech just shifted more than half of its lithography budget to the mask and reticle side, since the challenges of 157-nm and next-generation cannot be met without more-aggressive R&D on the mask side. A TSMC executive told a recent industry conference that the mask-set costs for the 100-nm process node could run to as high as $2 million, up from some $650,000 for the 130-nm designs.

But the number that truly boggles is the size of the mask design files that have to be transferred between the mask houses and chip fabs. Industry experts say that for each 130-nm mask set, 30 Gb of file space are needed. That number is expected to grow 10x by 2005 to about 300 Gb. Keep in mind that's 300 gigs for just one set. Imagine terabytes of files taking up pedabytes of storage space, and you can foresee a company's IT infrastructure choking like an errant pretzel in a chief executive's windpipe.

The market analysis mavens have also burdened me with a heavy load of statistical baggage. A couple of weeks ago at an IC Insights seminar in Sili Valley, Bill McClean shared some of his 2002 report. Aside from the usual industry charts and tables (he forecasts flat action for the chipmaking side in 2002 and negative capital expenditure growth, by the way), the veteran analyst foisted several chin-scratchers on the attendees. For the first time, the worldwide electronic systems market—that is, all the gadgets, appliances, cars, servers, and the like that chips go into—declined from 2000 to 2001. Not only did the market go down, it plummeted 10%. Even in other down years, this has never happened.

He also noted that the R&D spending of chip companies slipped year to year, dropping 9%. Again, even during downturns, research spending has always headed north. An optimistic glimmer did emerge from the R&D spiral though, since research outlays as a percentage of overall industry revenues rose 19%.

McClean reminded those in the chipmaking community with an exaggerated sense of self-importance that Wal-Mart's annual corporate revenues easily exceed those of the entire semiconductor industry. However, chipheads can take some comfort that without digital technology, the house that Sam Walton built wouldn't have grown into a Fortune 500 megamansion.

More numbers tell the tale of two U.S. companies headed in opposite directions. AMD has feathered its nest by partnering with UMC to build a high-volume 300-mm, 65-nm microprocessor fab in Singapore by 2005. The two companies also inked a separate foundry deal that gives AMD added 130-nm logic capacity with the Taiwanese foundry. The moves should help AMD strengthen its R&D and capacity positions and ultimately aid Sanders' Army in its market-share battle with you-know-who.

The preceding announcement also signals the end of AMD's technology arrangement with Motorola after the 100-nm node. Their cooperative efforts led to the successful development of copper interconnect processes now in high-volume production in Dresden's Fab 30 and elsewhere. Once an unassailable bulwark of the industry, Motorola's fortunes have been steadily deteriorating, leaving many former Moto lifers and fans to wonder whether the company's semiconductor products division will be sold off or spun off, either as a weakened whole or piecemeal.

While AMD's move to add capacity via the foundry and partnership route seems like a crackerjack strategic move, Motorola's "asset light" approach might better be labeled "asset flight." The company will spend a piddling $200 million in cap ex in 2002, down from $610 million in 2001 and $2.4 billion in 2000. The shuttering of multiple fabs, layoffs of thousands of employees, and shift of at least 50% of its manufacturing to foundries brings into question the electronics giant's long-term viability and intentions in the chip sector.

I hope I'm wrong and that such harsh reorganizational moves will turn Moto's fortunes around, but I wouldn't count on it.