Wafer Fab Profit Opportunities and Costs

		At a Glance
	A comprehensive facilities lifecycle cost-of-ownership (LCOO) approach is described that includes considerations of variations in sites, facilities, companies and products.

A recent survey found that the average state-of-the-art (200 mm, 25,000 wafer starts per month) wafer fab completed or planned in 1997 cost $1.1 billion to build and equip. ¹ Projections show that the cost of establishing a new fab will rise at 18% per year for the foreseeable future. Semiconductor manufacturing equipment obsolescence is rapid,² requiring frequent updates and replacement. Operating costs can run $1 million per day. Such cost structures make semiconductor production the most capital-intensive and time-sensitive form of manufacturing on the planet. In order to survive and grow, a semiconductor company must find answers to obvious, but difficult, questions: 

• What is the best -- that is, the most realistic -- method of measuring overall costs of establishing and operating these facilities over their productive lifespans?
• What are key drivers of costs and productivity (output per unit of input) over time? How are these best estimated and accurately compared "up front"?
• What are key risk factors, and how do they relate to costs and productivity?

Our studies of the impacts of external factors on the cost and productivity structures of 200 mm wafer fabs are providing some answers. Over the past three years we have found that national and community infrastructure definitely affects wafer fab productivity and costs over the lifespan of the facility. These factors can vary widely among different sites and significantly impact operational risks and returns of these fabs.

Background of fab COO research

Our research³ shows that realistic measurement of fab cost must capture every dollar spent on analysis, planning, designing, financing, permitting, building, equipping, initiating, staffing, managing, certifying, operating, supporting, correcting, maintaining, repairing, upgrading and, finally, decommissioning a facility over its entire lifespan. Not unexpectedly, we have also found that such costs and overall facility productivity are closely related. However, in practice, in spite of industry cost-of-ownership (COO) programs, many of these cost/productivity elements are overlooked or inadequately defined. Over the long run, a comprehensive facilities lifecycle cost-of-ownership (LCOO) approach is needed to ensure profitability. This is especially true for traditionally lower-margin industry segments; but as margins begin to shrink in presently high margin products, it will become an industrywide concern.

Leveraging facilities LCOO re-quires looking at two sides of the ledger. The initial cost side is at least well quantified. The other side encompasses productivity, but is not always well defined. For example, many semiconductor firms treat facilities and real estate as largely undifferentiated "cost centers" and assume that by applying sufficient resources, high productivity and rapid capacity expansion can be achieved almost anywhere. That is bolstered by worldwide availability of seemingly attractive fab sites featuring good infra structure elements like water and power, combined with incentives and inducements from eager communities and governments. If productivity were the same at all sites, then site analysis and selection would be simple: Pick the location with the best combination of (apparent) costs, features and incentives.

However, we find that productivity, lifecycle (as opposed to initial) costs and risk profiles vary widely with location. An LCCO model analysis of a candidate greenfield site, anywhere in the world, requires input for expected productivity, infrastructure quality, company (and its products) and risks. Results often show that sites offering large financial incentives produce suboptimal or negative rates of return on investment or moderate returns with high risks. We have found a genuine worldwide shortage of financially attractive and low-risk sites that offer infrastructures that can support the combination of high productivity and low LCOO needed to permit the new generation of fabs to earn high returns on investment with low risk. 

Our findings are borne out by recent industry experience with apparently low-cost sites. During the past three years, the industry experienced "show stoppers" in places like Thailand and Malaysia and, to some extent, in developed countries like the United States. Each of these cases resulted in sizable financial losses even without including full lifecycle costs. One lesson is that productivity and competence (at any level of costs) cannot be assumed, a priori, for an untested infrastructure that may not be able to support state-of-the-art fabs. Indeed, many apparently low-cost sites were found to include high risks, hidden costs, poor productivity and other surprises. Thus, site and facility decisions are ever more critical; they are not just "cost" problems, but productivity, cost and risk opportunities. With the scale of such operations, such decisions can impact a company's overall productivity -- and the bottom line.

The business of infrastructure assessment and development

The idea that the productivity impacts of infrastructure can be systematically assessed and quantified is not new. The more astute semiconductor firms' players implicitly or explicitly presume that costs, productivity and risks can vary widely at different fab sites, and that has been generally beneficial. However, by using a "profit center" approach with the LCOO model, it is possible to maximize facility returns and minimize site and facility risks. For each candidate site and facility, a semiconductor firm will benefit by investing substantial research and analysis effort needed to do the following: Take a global, system-level view of costs and productivity; estimate total LCOO for specific programs and products "up front"; quantify risk exposure; and adopt facility selection, development and operations strategies to lower long-term costs and risks and to enhance productivity.

Systematically taking these factors into account allows a semiconductor company to optimize profitability and lower risks, rather than just controlling costs. FHI's approach to these problems in LCOO assessment began with the development of a set of tools, techniques, metrics and models founded on a proprietary database of established state-of-the-art 200 mm fab sites. The model differs from traditional site analysis models in estimating differences in productivity as well as differences in costs and risks at each site for different companies and products. Candidate 200 mm sites or facilities were then benchmarked directly against that worldwide LCOO database for direct comparisons. The model is now being extended to 300 mm sites as specifications for 300 mm facilities become available. Preliminary results indicate that 300 mm facilities are inherently more sensitive to these productivity and risk factors than 200 mm facilities running similar geometries.

Present fab COO status

COO is not new. Decades ago, the U.S. government recognized that some military systems cost several times the original price over their useful lifetimes. Today, all kinds of businesses are applying COO methods, notably in information system total cost-of-ownership (TCO). Obviously, that is driven by growing reliance on costlier systems, so it is hardly surprising to find the most capital-intensive industry, semiconductor manufacturing, using well-developed COO methods. However, there is a need to go much further.

Most of the work to date has focused "inside the box" -- that is, within a given wafer fab. Sophisticated COO techniques for individual pieces of equipment (tools) underlie most high-value procurement or upgrade decisions. Semiconductor Equipment and Materials International (SEMI) has published COO standards, and there is at least one commercial product (from Wright Williams & Kelly) based on those standards. ⁴ However, much of this has been done internally by individual semiconductor companies, creating differing proprietary approaches that make company-to-company comparisons troublesome. Also, incorporation of certain "soft" factors (like operator skill or management strategies) varies widely.

There is a recognized need to develop more "systems-level" COO capabilities. Some commercial products now operate at sub-unit, department or plant levels. While a good start has been made, there are clear needs in fab-level simulation and COO analyses -- and in harmonizing results obtained from differing approaches.

Extending the fab COO approach

Fig. 1. LCOO encompasses site, facility and company factors.

Our view of LCOO incorporates existing "inside the box" technologies but reaches "outside" to include key factors related to site, location and region. These outside factors are more and more significant in determining the real costs of fab ownership over the productive lifespan of the facility. They encompass all cost and productivity impacts of local infrastructure in the broadest sense of the term, including site analysis, site preparation, construction, operations, maintenance and support. They can be divided roughly into convenient categories, as illustrated below, but it is vital to remember that COO factors tend to be highly interactive with each other across all categories and types.

Figure 1 illustrates some basic LCOO categories. The facility factors (all elements of the facility itself, equipment, operations) are covered by existing COO techniques (shaded portion), some of which are quite sophisticated.

The company factors (on the right) encompass strategies, governance, ⁵ finance, core competencies⁶ and the specific products and programs planned for the new facility. These are key to overall firm success -- and highly individual. For example, we found two fab expansion programs where it was clear that government cash subsidies were critical, given those companies' financial situations. They were far more valuable than other forms of subsidies to these companies at that time, but would not have been important to a more cash-rich firm.

"Hard" and "soft" site factors

Certain site factors are quite familiar. "Hard" site factors such as costs, quality and availability of power, water, waste disposal, transportation and land are part of traditional fab site evaluation and are reasonably quantifiable. For fab sites, they remain important -- indeed tighter technical specifications (water quality, power quality, vibration levels, etc.) make some more critical than ever.

More attention has been focused recently on "soft" site factors such as political and social issues (anti-growth movements, environmental restrictions, labor laws) and economic factors (site recruitment incentives, national, state and local tax regimens). With an at least perceived shortage of high-tech professionals and workers and a greater awareness of the impacts of these factors on facility productivity, there is greater, but uneven, recognition of "softer" issues such as quality of life and educational systems.

This illuminates the dilemma of fab site selection. There is a widespread perception that the world is full of good candidates -- and many do include apparently attractive incentives and hard infrastructure. However, there are few sites that offer the industry a good business opportunity -- decent profits combined with low risks -- because so few have the kind of soft infrastructure that increasingly underpins high productivity in such plants.

The COO framework

Fig. 2. The basic COO framework incorporates all site and facility factors that significantly impact wafer fab COO.

A primary working tool is the basic COO framework (Fig. 2), which incorporates all site and facility factors that significantly impact wafer fab COO. A number of company COO factors are also incorporated, primarily strategic (product selection/development/marketing) and financial elements. This site/facility/company factor context provides a more comprehensive picture of actual COO.

This COO framework is used at different stages of facility lifespan, from initial site selection through planning, construction, startup, production ramp-up, maturity, upgrades and decommissioning, to provide "snapshot" views. While the approach is consistent, there will be a differing set of emphases for each "snapshot." The framework then becomes the basis for the more dynamic COO modeling.

An example of how the basic framework can be employed to develop the LCOO model for a specific situation is shown in Figure 3. This shows how the basic framework analysis at any given lifecycle stage of a fab must include COO factors at other stages of the fab lifecycle. For example, facility planning and design must include key company factors anticipated in production (1). Soft site factors that impact recruiting of technical personnel will have a very direct impact on company factors (and on facility factors) in all later production phases (2). Hard site factors would obviously have direct bearing on the productivity and COO of the facility itself (3), and that would impact key company factors at the time of decommissioning (5). Hard site factor analysis during site selection strongly interacts with the set of facility factors anticipated at the point of decommissioning (4).

Framework, LCOO and supporting techniques

Thus, the basic framework provides a series of linked snapshots of a particular site and facility over time, and that becomes the foundation of an inclusive and dynamic LCOO model. However, there are some obvious pitfalls.

First, like any valid COO modeling, it should include all cost impacts from initial selection and acquisition through the productive lifespan to final disposal, sale or decommissioning. While present COO techniques generally follow this scheme, the LCOO model is much more comprehensive in folding in the outside factors. Operating costs should reflect probable levels of technician and engineering availability, quality and experience. Salvage, sale or decommissioning should include a weighted analysis of potential liabilities (e.g., the future likelihood of more severe environmental remediation or employee out-placement requirements).

Second, systematic and consistent use of technology, market and product disaggregation techniques must be integrated into COO. It is relatively straightforward to link up-to-date technology roadmaps, marketing projections and product strategies to specific COO entities and factors, permitting quantification (or at least bracketing) of cost impacts from anticipated changes in technologies, market demands and products.

Experience with the model

The LCOO model for 200 mm fabs is new, but major elements have been tested over the past few years and benchmarked against actual cases. This has proven especially effective in highlighting unanticipated fab COO impacts in the following:

• Site issues
  • Soft factors
    • Community attitudes, understanding, preparation and trends
    • Overlapping and competing political and infrastructure institutions
    • Inappropriate/obstructionist permitting practices
    • Construction labor pool constraints and trends
    • Technician labor pool constraints and trends
    • Professional recruiting and retention
    • Overall and specific delay potentials
  • Hard factors
    • Power quality and availability status and trends
    • Water-related issues and trends
    • Hazardous materials constraints and trends
    • Telecom/communications constraints, risks trends
    • Environmental constraints and trends
    • Fiscal (including incentives) constraints, risks and trends
    • Business interruption risk/cost profiles
• Operational Issues
  • Facility vulnerabilities and liabilities
  • Operational support potentials and trends
  • Full tool-level, departmental-level and facility-level COO (using commercially available technologies)
• Decommissioning issues and trends
  • Potential liabilities
  • Potential cost growth

Recent industry examples

Delays and interruptions are a business fact of life, part of normal financial planning and seldom exceptional (barring major floods, fires, etc.). However, with $1 billion-plus investments and $1 million/day operating costs, the new generation of fabs is an exception. ⁷ One symptom is the typical 7 x 24 production schedule (seven day-a-week, 24-hour-a-day operation with perhaps a shutdown Christmas Day) to maximize investment and production efficiencies. This raises some interesting problems.

Table 1. Facility Implementation Times
Facility	Firm A	Firm B	Firm C
Announcement	May 1996	August 1995	May 1995
Groundbreaking	October 1996	November 1995	December 1995
Completion	October 1997	December 1996	June 1997
Construction	11 months	13 months	18 months*
First tool install	October 1997	January 1997	**
*Schedule not met **Confirmed data not available

For example, costs associated with any unanticipated delays in establishing and ramping up a new fab obviously become significant. Simply covering the debt service for such a facility is daunting if its revenue generation is delayed for weeks or months. However, delays can also create significant opportunity costs (from missed market opportunities and from obsoleting product and toolsets) and exacerbate staffing and training problems. It has become extremely important to avoid them.⁸ Well-planned construction programs can reduce the time between groundbreaking and facility completion, while excellent planning and coordination can expedite installation and qualification of tooling and equipment. It is interesting to examine some recent new U.S. fab sites.

We studied three new U.S. fabs of comparable size and program type but very different startup histories (Table 1). Why the disparity among these three sites? Our studies show that establishment and ramp-up of the sites for Firm A and Firm B were generally well planned and executed and represent good industry benchmarks for the current generation of fabs. We found that the degree of variation in this critical set of LCCO factors within the United States is substantial, and preliminary indications are that the variability of LCCO and risk exposure for international sites is substantially greater than for those in the United States. 

Fig. 4. Of the three firms studied, two had a high awareness of LCOO issues, while the third failed to recognize problems that could significantly compromise returns on investment.

The site for Firm C has attracted attention because a number of problems were discussed in the trade press as well as general-interest newspapers. Our own analysis indicates that the LCOO penalty they impose will run well into the millions on this facility when estimated opportunity costs, costs of delays, construction cost overruns and other factors are included. By itself, the seven-month additional startup delay (compared to the "best" case) accounts for a profitability penalty in excess of 1% over the lifetime of that fab.

Our observations of the three firms' site selection processes (Fig. 4) revealed a high level of awareness of LCOO-related issues with Firms A and B, and those operations still seem to reflect that. Firm C emphasized short-term incentives in its site selection. Without comprehensive LCOO analysis, that site probably appeared to be of moderate risk at worst. Events proved otherwise. LCOO will probably be high, productivity problems are likely and returns on the investment will be significantly compromised.

Conclusion

Rising costs and increasing risk exposure are major issues for the global semiconductor industry as it moves toward fewer and more expensive fabs at fewer locations. The impact of national and community infrastructures on LCOO and risk profiles is already significant -- and increasing. The LCOO model shows that in a developed country site, an increase of at least 1% in lifecycle fab output can be realized by optimizing facilities infrastructure and risk profiles. That is far more than the benefits of most government incentive programs. FHI's LCCO model was developed to help fab owners and the financial community understand, measure and optimize lifecycle return on investment vs. risk and to obtain a realistic net estimate of the true benefits of incentives.

Experience with greenfield sites demonstrates that different risk profiles and/or LCOO situations can exist at sites separated by less than a kilometer. The productivity and profitability of similar sites and facilities can vary widely with different companies (and programs and products). Thus, static site-ranking exercises or "best sites lists," while popular, are often deceptive since realistic evaluation must be site-, facility- and company-specific.

Integrating all three types of elements into the model permits semiconductor companies to treat facilities and sites as profit centers by optimizing facility location and product development choices, providing critical lifecycle financial, marketing and other benefits.

Acknowledgments

The authors are indebted to many people for their knowledge, insight and encouragement in the development of the LCOO model. Thanks go to Jim Burnett (Burnett Technologies), Alan Levine and Daren Dance (Wright Williams & Kelly), William McClean and Richard Skinner (IC Insights) and William Sherman (Total Business Continuity Planning).

References and endnotes

1. IC Insights, Inc., Scottsdale, Ariz.

2. Such as a new generation of very high-cost photolithography equipment about every 18 months.

3. Some of this work was reported in a series of Fisher-Holstein articles published in the February, March and April 1998 issues of Infrastructure (www. infras.com).

4. Described in the article "Understanding Equipment Cost of Ownership: Use COO for more than just equipment purchases," by Daren L. Dance, David W. Jimenez and Alan L. Levine (in this issue).

5. Governance is a term that has been used to describe the mode of management and internal direction of a company. It is a complex interaction of such elements as formal and informal policies and practices, other parts of the "command structure," organizational culture, and company history and learning.

6. Complex bundles of technical competencies (design, process engineering, etc.), management competencies (marketing, leadership, etc.) and other resources that form the basis of a firm's competitive advantages. See Prahalad, C.K. "The Role of Core Competencies in the Corporation," Research Technology Management Journal, Nov.-Dec., 1993, pp. 40-47.

7. Fisher-Holstein, "Fab Cost of Ownership: The Impact of Delays and Business Interruptions," Infrastructure (www.infras.com), March 1998.

8. Business interruptions (from fires, floods, weather, etc.) also become more important and carry other serious implications in terms of product yield and quality.

Talkback

We would love your feedback!

Post a comment

» VIEW ALL TALKBACK THREADS

Related Content

• Topics
• Author

Related Content

 

By This Author

There are no other articles written by this author.

SPONSORED LINKS