EDITOR'S PAGE

The American alternative

The October 2004 edition of the World Fab Watch database states that two-thirds of probable future fabs will be built in the Asia-Pacific region. These data should surprise no one who's familiar with the exodus of chip manufacturing to Taiwan, South Korea, Singapore, and China. For many Americans, the trend is seen as part of a troubling erosion of technological and economic strength. For most players in the semiconductor world, this outsourcing is just a part of doing business in a rapidly evolving market.

But not every U.S. chipmaker has jumped on the Asian fab migration bandwagon. To the surprise of many observers, Texas Instruments has decided to build its next 300-mm fab in Richardson, TX, not far from the company's Dallas headquarters and existing 300-mm factory, DMOS6.

Several presentations at October's ISMI Manufacturing Effectiveness symposium in Austin offered insights into the planning for the new Fab One. In her keynote address, Shauna Sowell, TI's vice president and manager of DFAB, laid out the rationale for the decision to build the new fab in the Lone Star State. Of course, tax breaks and other incentives from the state played a major role, but one of the most intriguing reasons Sowell noted is Fab One's sustainability model.

Sowell cited a benchmarking study of global 300-mm fabs revealing that the Asian projects had a 50% cost advantage, mostly because of lower initial facilities costs. In order to compete on a cost basis, the TI design team focused on lowering certain initial costs (Fab One will be a two-level factory, for example) and life-cycle costs, since utility expenses often exceed construction expenses over a fab's life. According to the executive, these efforts to bring down life-cycle expenditures will help meet the company's goal of reducing Fab One's costs per square foot by 30% compared with DMOS6.

Other papers at the symposium offered specifics on TI's plans to construct and operate Fab One in a sustainable fashion. Paul Westbrook described how his team's approach uses energy modeling, works with the tool and support equipment vendors on source load reduction, takes advantage of minienvironments in the HEPA coverage and gowning areas, and optimizes and integrates the facilities systems—all without raising capital expenditures. He estimates that these efforts will save TI about $750,000 in the first year of operations and about $3 million annually at full build-out.

In his presentation on water sustainability, Pablo Ruiz noted that changes in the management of the three main types of water at the new fab—DI, industrial waste, and city—could result in significant savings. One example showed a scheme for reusing industrial wastewater for point-of-use abatement systems and scrubbers that could save more than $1.3 million a year.

Later in my Texas sojourn I toured DMOS6 and had a chance to chat with two key members of the management team: fab manager Sima Salamati-Saradh and company vice president Larry Tolson. We
discussed how the lessons learned during the ramp of the 300-mm facility have improved TI's overall manufacturing abilities and bode well for a successful start-up of the new facility.

When I saw DMOS6 in March 2001, the large, open ballroom had one functioning litho cell and other tools scattered around the central area and adjacent rooms. Three-and-a-half years later the scene was quite different. Most of the waffle-floor was occupied by production toolsets, the overhead automation tracks were active with wafer lots, and more equipment was coming in, as the 90-nm ramp continued. Salamati-Saradh said that getting the fab automation to work, especially the software, was "a painful learning curve" but that the system has now "proven out to be great." She estimates that the fab is running in the "mid-80s" range in terms of automated delivery, a number Tolson believes is higher than many 300-mm fabs he's seen.

Sowell, Salamati-Saradh, and Tolson all admitted that DMOS6's 130-nm ramp was not without its difficulties. One humbling moment came when benchmarking studies revealed that TI's Taiwanese foundry partners had gone, in less than a year, from well behind the IDM in yield and production metrics to way ahead. After taking their medicine at 130, the TI executives say the 90-nm ramp has gone much better, with Tolson emphasizing big improvements in the interactions between the process development and production teams.

When you combine the 300-mm process and manufacturing expertise developed at DMOS6 with the innovative design-for-environment approach the company is taking with Fab One, TI presents a promising, cost-conscious alternative to the further erosion of the U.S. chip-manufacturing base.

Tom Cheyney
Editor
tom.cheyney@cancom.com