Debra Vogler, Senior Technical Editor
Among the topics explored at the International Sematech Manufacturing Initiative (ISMI) Global Economic Symposium, held in San Jose, CA, late last year, was timing of an industry move to 450mm wafers and a rational business model for silicon suppliers.
Admitting that his views are controversial, Applied Materials’ Iddo Hadar, managing director, corporate strategy, plunged right into his economic analysis. Data tracked by the Federal Reserve Board indicates that fab productivity added $500B to the US economy in the late 1990s. By his estimation, these gains came about because of nearly $40B spent on process R&D across the industry in the early 90s.
Hadar also noted that it was equipment suppliers – able to match R&D funding on the part of the semiconductor manufacturing industry – that enabled the two-year roadmap cycle. “But things have changed significantly,” said Hadar. “The ability of suppliers to match R&D funds has deteriorated.” He specifically pointed to the broad array of alternatives being pursued by the industry for the 65nm and 45nm nodes – all per the guidance of the ITRS. For example, there are four types of substrates being worked, two different device structures, five different kinds of gate dielectric materials, two new gate electrode architectures that, in turn require more than six new materials, and two types of source/drains. All of these factors contribute to most of the industry going back to a three-year cycle.
Perhaps more ominous was Hadar’s contention that any delay – similar to what happened to the 200/300mm transition – in going to 450mm wafer manufacturing will increase the development bill, just as happened during the 200/300mm transition. According to Hadar, the delay caused increased R&D spending on 300mm an order of 2-3x and also resulted in redundancy of efforts. In one scenario, Hadar’s calculations indicate that a $20B investment (on the part of the total food chain) would take 30 years to recoup, assuming a 30-year life-cycle and a peak at 50% of the silicon area processed. “That’s a long time and many companies don’t survive that long,” said Hadar. “We are living with the consequences of spending on 300mm which was delayed.” He believes that if the industry moves to 450mm manufacturing the same way it did 300mm manufacturing, it will be worse off and the industry would not come close to recouping its investment.
Hadar suggested that equipment suppliers and IC manufacturers must now share the risk in going to 450mm wafers and duplication of the start/stop/start experience with the 300mm transition must be avoided. He holds up the X-Initiative as one example of a successful collaboration between these groups. More fundamentally however, he points out that the industry has to decide what its investment priorities should be: installed base improvement, future technology, or the next wafer size. WN asked Hadar if he thought the current efforts by some suppliers to more heavily support the secondary equipment market and fab services was a decision to spend more on the installed base and he tended to agree with that notion.
While equipment suppliers have had problems coping with the dynamics of late, materials suppliers have had their own share of challenges. Karen Twillmann, director of market research, at MEMC, as well as a representative from the Silicon Manufacturing Group, told attendees that it was semiconductor units – not revenue – that drives the silicon market. She estimates that about $6B of the projected $15B of materials sales in 2004 will be in silicon.
According to Twillmann, since 1997, the 200mm silicon market has faced accelerated pricing erosion due to over-investment in capacity. She pointed to the decision last year by SUMCO and Siltronic to take out some of their 200mm capacity as a response to this supply/demand imbalance. She also reiterated a lesson learned in other industries, “No industry or company has survived over the long term with an inverted asset turn model. Today, everyone is cautious in making 300mm investments – the intent is to have a natural balance between 200mmm and 300mm demand.”
Reviewing some basic economics, Twillmann noted that if any company in the industry spent 25% of its revenue on capital expenditures – with its resulting 25% depreciation costs – then it would take a 45-55% gross margin just to break even (assuming 10-15% cash costs and 10-15% operating costs). She called that a broken ROA model. For MEMC, the obvious solution is that silicon suppliers should spend no more than 15% of revenues on capital expenditures.
There is some light at the end of the tunnel because of the diversification of semiconductor applications where long-term demand is expected to keep growth steady. Twillmann noted that applications such as SOI and strained silicon, along with high-quality wafer products such as low COP and COP-free wafers, and epi and annealed wafers, all help to de-commoditize silicon. Between 2003 and 2009, she noted that an 8-10% CAGR is projected while the 300mm segment is gaining share with 2008/2009 being the expected cross-over point. Currently, Twillmann noted that the 300mm market share is now ~12% but by the end of 2005 it is expected to be ~20%.
Along with the rosy outlook for the silicon market, however, Twillmann predicted that there will be consolidation in the silicon industry within the next few years as profitability and survival become key to winning in the materials sector. Currently, there are seven producers of 300mm wafers; in 1993, there were 11 producers of 200mm wafers and pre-1990, there were more than 20 producers of 150mm or smaller wafers. Answering a question from the audience about silicon suppliers in China, Twillmann noted that there are ~20, but very little is known about them – it is very difficult to track them, although industry groups, such as SEMI, are trying to do so.