When Intel barks, vendors grimace, then listen very carefully. Two of the big three stepper manufacturers—ASML and Canon—claim to be staying the course with their 157-nm programs, while Nikon says it is reevaluating its position. Texas Instruments, Infineon, and other chipmakers have reiterated their qualified support for 157, while acknowledging major challenges in the mask, pellicle, and optics areas. Such support may be welcomed by suppliers and researchers that have already poured millions into 157-nm R&D, but Intel's roadmap alteration may put market forces in motion that could lead to the nascent technology's demise.

Intel's decision, although significant for the leading-edge contingent, is frankly not that critical to most industry folks I met in Minnesota, Missouri, and Ohio. Pure-play foundry upstart PolarFab, working the analog and mixed-signal niche, is running its PBC4 0.50 µm process and ramping its Polar35 0.35 µm line in its 200-mm module: It is years away from needing 193-nm litho, let alone 157. In fact, the Bloomington, MN, company has bought mostly used and refurbished tools to meet its needs, even finding gear that company veteran Jim Wittek described as "still in the original wrapper.... We're careful to get tools that fit with our existing toolset. We don't play in state of the art." Polar's doing quite well on the trailing edge: the privately held company expects to see 2003 revenues increase 50% over last year, according to vp of sales and marketing, Dennis Gaetano.

Of course, the bleeding edge is in the eye of the beholder. Senior metrology manager Darrell Louder showed me around Seagate's giant magnetoresistive (GMR) fab, also in Bloomington. While many laboring in the 90-nm semiconductor salt mines think they're ahead of everyone else's curves, the strapping PhD told me some surprising particulars about the thin-film-head manufacturing world.

"We use metrology a lot more than the chip guys," he said. "We have 88 different steps to measure physically." He noted that Seagate gets 31,000 isolated-feature die per wafer on its new densified designs. Keep in mind that's per 150-mm wafer. Each wafer gets FIBed (focus ion beamed) at least once, if not twice, during production. Atomic force microscopy must be used in-line to help control roughness during CMP. The most mind-blowing factoid Louder laid on me was the thickness of certain parts of the GMR stack walls: Individual layers are down in the single-digit angstrom range in a multilayer stack measuring in the tens of nanometers. "The specs are getting so tight that it's almost impossible to build and measure. We can see the secondary effects, but not the primary ones."

For most tool, materials, and components suppliers, Intel's 157-nm announcement is a mere blip on the corporate radar. The companies I saw during my trip are more concerned with sharpening and streamlining their manufacturing practices, focusing their R&D and intellectual property activities, consolidating acquisitions, shutting down business units, and getting better at understanding and serving their customers.

BOC Edwards's Chanhassen, MN–based chemical management division has adopted a "cradle-to-grave" project team approach and branched out into the materials management segment, scoring a major contract with TSMC. FSI's shedding of its track system unit has been painful, admits chairman/CEO Don Mitchell, but it has allowed the Chaska, MN, company laserlike focus on the cleaning and conditioning segment, where it garners significant 300-mm market share. FSI neighbor Entegris has quickly absorbed the recently acquired wafer and reticle carrier unit of Asyst, with the product line already integrated into the parent's manufacturing infrastructure.

Brewer Science has opened a spacious new R&D facility in Rolla, MO, and is actively developing its MEMS materials business, while fellow Missourians at MEMC have been pleasantly surprised by a discrete-driven spike in their 150-mm wafer business to go along with healthy 300-mm orders. Swagelok has dedicated part of its Cleveland-area manufacturing base to support its newly created semiconductor services company in high-volume as well as customized product applications.

Maybe Intel's decision to scrap 157 isn't such a big story after all.