The photomask industry: Minimizing a crisis in escalating costs
08/01/2002
By Robert N. Castellano, President, The Information Network, New Tripoli, Pennsylvania
The photomask industry is between the proverbial rock and a hard place. To recoup the costs of expensive e-beam and laser beam technology needed to generate photomasks for state-of-the-art ICs (130nm and below), photomask makers have had to increase prices geometrically.
A photomask set at 130nm can cost up to three times more than an 180nm photomask set. Over the past five years, pellicle and blank raw material costs have increased at a compound annual growth rate (CAGR) of 57%, while tool costs have increased at a CAGR of 37%. In 1995, a Core 2564 tool would retail for $2.5 to $3 million.
Today, tools needed for 0.13 and 0.10μm are $15 million each. Additionally, increased complexity means it takes longer to make the photomasks, shrinking margins further. For a photomask set with features >5μm, a tool could process 15 photomasks/day. Throughput drops significantly to three photomasks/day at 0.18μm.
An 0.18μm binary photomask for critical layers costs between $8000-22,000 compared to $5000-10,000 for 0.25μm design rules, and photomasks for noncritical layers cost between $18,000-20,000. OPC and phase shift add substantially to these prices. We estimate that a 0.10μm photomask set for critical layers will cost $100,000 and a complete photomask set for a 25-layer IC will be $2.5 million.
The combination of 90nm-feature ICs on a 300mm wafer yields a large number of devices, so an expensive photomask set can be amortized over long wafer runs for high-wafer-volume devices such as DRAMs and microprocessors. But what about medium-wafer-volume devices such as SOCs, FPGAs, DSPs, or ASSPs, or low-wafer-volume devices like ASICs?
ASIC manufacturers will be challenged to make profits unless the run rate is high enough to overcome photomask costs. As volumes grow, however, yields improve and advanced tools and processes are introduced. Improved productivity and yields lead to price reductions. While initial 0.13μm photomask sets may have approached $1 million, this price has been declining. The reason photomasks are costing more is simple: Tools to make them have risen dramatically in price.
Also at risk are fabless companies. Although they have depended on their foundry partners for photomask services, many of these companies are small design teams without the financial resources to warrant iteration after iteration of designs that require the respinning of expensive photomask sets.
Fear, uncertainty, and doubt (the FUD factor) have led some in the industry to express concern that the rising cost of photomasks will limit growth in technology, particularly when it negatively impacts cutting-edge developments fostered by clever design teams at fabless companies and universities.
Strategy and product solutions
Merchant photomask vendors, foundries, IDMs, metrology companies, and lithography tool suppliers all keenly aware of the ramifications of escalating photomask costs have initiated a series of strategies and products intended to minimize these problems.
Mask vendors and IC manufacturers are working together to ensure their continued edge in the race to create semiconductor generations with increased functionality in ever-smaller geometries. Captive photomask houses are abandoning complete photomask development in favor of a cooperative model for high-end masks designed to speed up development cycles and reduce costs.
As an example, Dai Nippon Printing (DNP) and STMicroelectronics have formed an alliance for the development and supply of leading-edge photomasks. DNP will create a new company called DNP Photomask Europe, and will become ST's primary leading-edge and high-end photomask supplier.
Similarly, Infineon Technologies AG, Advanced Micro Devices, and DuPont Photomasks will establish and operate a new advanced photomask facility in Dresden, Germany. An equally owned joint venture, Advanced Mask Technology Center GmbH & Co. KG (AMTC) will be co-located with a new commercial photomask production facility that DuPont Photomasks will establish in Dresden as a separate entity.
MPW services put many designs on the same photomask and are mainly used for small- and medium-sized runs, particularly experiments with new designs. The services have also been used for some ASICs. Companies that offer MPW include Atmel, Austria Microsystems, IBM Microelectronics, and TSMC.
Lithography vendors also tout new tool approaches to helping cut the cost-of-ownership of advanced reticles. Double exposure techniques such as Canon's Ideal method or Numerical Technologies Inc.'s Grateful approach can greatly offset the high cost of new phase-shifting photomasks.
NEC believes that the growing cost of custom IC photomasks and increasing complexities in chip development cycles are driving the need for a new mid-volume ASIC solution. Its new ISSP architecture will offer up to a 75% reduction in development and production turnaround times compared to regular standard
cell-based ASICs. The nonrecurring engineering costs are also expected to be reduced up to 10x compared to cell-based IC designs.
DuPont Photomasks has acquired software company BindKey Technologies, which is focused on developing innovative tools that reduce the number of iterations required from design verification to tape-out, or the release of the design for photomask production. Traditionally, manufacturing-verification software has detected layout errors that require time-consuming and costly manual repair prior to the tape-out step.
KLA-Tencor's PASS software enables photomask manufacturers to characterize defects without having to move the photomask from the inspector to an off-line aerial-defect-imaging tool. In addition, PASS automatically simulates how the defects will print onto the wafer before the photomask is shipped to a customer or used in production. Advanced reticle inspection and defect analysis are critical to assessing the effects of defects before they are printed onto wafers.
Merchant vs. captive markets
The rising cost of making photomasks has driven many semiconductor manufacturers to shutter their internal photomask-making facilities in favor of purchasing photomasks from merchant suppliers. STMicroelectronics and Infineon, cited above, are two recent examples. There have been numerous others, illustrating a trend that will continue at a rapid pace:
- DuPont Photomask acquired the captive photomask shops of Hyundai and Hewlett-Packard in 1998, IBM Europe in 2000, and formed a joint venture with UMC in 1999.
- Photronics acquired the captive photomask shop of Conexant in 2001.
- DNP acquired the captive photomask shops of Hitachi in 1999, and Toshiba and Fujitsu in 2000.
- Toppan acquired the captive photomask shops of Sony in 1999, Sharp and Mitsubishi in 2000, and NEC in 2001.
Figure 1. Comparison of worldwide merchant vs. captive photomask revenues. |
The worldwide captive photomask market decreased from $900 million in 1999 to $660 million in 2000 to $560 in 2001, as shown in Fig. 1. It is interesting to note that the total merchant photomask market increased from $1.71 billion in 2000 to $1.77 billion in 2001 at a time when semiconductor revenues dropped 30% and equipment revenues dropped 80%.
There have also been consolidations among merchant photomask suppliers. Photronics, for example, acquired PSMC, Align-Rite, and PKL in 2001.
Figure 2. Worldwide photomask market share, 2001. |
Consolidation within the photomask industry has had an impact on the relative shares of the merchant suppliers on a worldwide and a regional basis. Figure 2 shows worldwide market share for the top four photomask vendors.
DNP led the worldwide market in 2001 with a 26.8% share, up from a 25.7% share in 2000. In North America, Photronics was the market leader with a 48.7% share. DuPont Photomask was the market leader in Europe with a 42.2% share and in Asia (excluding Japan) with a 30.5% share.
Photomask demand is principally driven by new design activity, which is gaining momentum in all regions. Additionally, growth is being fueled by customers' ramp of new advanced steppers and scanners that utilize the most advanced photomask products, resulting in higher average selling prices and margins. Small feature sizes will register the greatest gains in photomask consumption.
In 2000, photomasks to replicate features of <0.20μm exhibited a 18.0% share of all photomasks, up from only 0.5% in 1998. In 2003, we predict that this sector of the photomask market will hold a 42.1% share.
Figure 3. Worldwide photomask market forecast. |
Figure 3 is a forecast of the combined worldwide merchant and captive photomask markets. Worldwide revenues decreased 1.5% in 2001 to $2.3 billion due to the large decrease in the captive market, as illustrated in Fig. 1.
Low-end reticle prices have dropped 20% in the last six to seven years, offsetting a rise in advanced mask prices. According to The Information Network, revenues will increase 10.5% in 2002 to $2.6 billion and will increase a further 18.0% in 2003 to $3.0 billion.
Robert N. Castellano is president of The Information Network, 8740 Lyon Valley Rd., New Tripoli, PA 18066; ph: 610/285-4548, fax 610/285-4547, e-mail [email protected], www.theinformationnet.com.
This article is based on The Information Network's report "Maskmaking, Inspection, and Repair: Market Analysis and Strategic Issues." Contact the author for more information.