In Boy Scouts, we learned "leaves of three, let it be" to avoid the irritation and embarrassment of poison ivy. In the semiconductor industry, we learned "every chip manufacturing supply chain will eventually stabilize at three suppliers." Since its implementation in volume production in 1989, CMP has been slow to conform. While pads and slurries each began under the rule "suppliers: one; not much fun," CMP equipment hovered near the Rule of Three from the beginning. There were some wild excursions in the number of equipment suppliers in the 1990s, but once Applied Materials came on the scene, albeit slowly at first, the Rule of Three was never seriously challenged. The number of CMP equipment tombstones is about what one would expect for a field touted for more than a decade as "the fastest growing process sector in the semiconductor manufacturing business."
The roller coaster ride for CMP slurries has been a bit more thrilling, if not frightening, at least for some. From a high of over 35 self-proclaimed semiconductor slurry suppliers, we still have 15 manufacturers that can each claim a measurable market share of qualified volume production, according to the Techcet Group 2010 CMP Consumables Critical Materials Report. This supplier base is supported by the variety of process types currently in production, new processes and material sets being developed, and by regional market demands, particularly in Asia. The Rule of Three has not taken effect because we keep changing the rules – what materials do we want to polish, what materials do we want to stop on, what kinds of devices are we making, where are we making them, what are the cost pressures that govern.
Another factor that has foiled the Rule of Three is that CMP slurry is a performance material. For all the loose talk in which we may indulge about commodity pricing of ILD slurry, the fact is that no two slurries are interchangeable without some extensive process re-qualification and adjustment of quality control set points throughout the fab. This gives rise to a lore of legacy supply incumbency – once you’re in, you’re in for life – but this old behavior is no longer bankable, as fabs become more savvy about point of use dilution, reduced slurry flow, and other cost-reduction strategies.
The CMP pad business has quite a different story to tell. Long dominated by Rodel, the supplier itself was swallowed by Rohm and Haas, and then again by Dow Chemical, all the time holding on to its >80% market share. Competing pad startups have come and gone, unable to navigate the patent minefields successfully, or upon qualifying in a fab, being told that they need to be bigger before volume fab production will be committed to their pad. Only Thomas West found a sustainable niche early on, and holds steady to this day, mostly with its quiet corner of the tungsten pad market.
Along the way, there have been several disruptive threats to the CMP pad status quo. 3M’s fixed abrasive pads have captured about half of the selective STI market, but have struggled to expand beyond this single niche. ECMP promised a kinder, gentler approach to copper polishing with a Nafion? membrane enclosing a copper removal electrolyte, but was unable to concurrently meet the dishing specs across the requisite range of feature sizes. It may have found a niche in the read/write head process, but not in the chip fab. SemiQuest’s button pad still looking for the right opportunity to break into the pack with its promise of superior edge exclusion capability.
None of this lends itself to a legitimate conclusion that the Rule of Three has asserted itself in the CMP pad business, but there are certainly signs that the market dynamics are morphing in the direction of Rule of Three behavior. Fab engineers and purchasing agents are committing polishers and wafers to new pad qualifications. Pad conditioner suppliers are devoting more resources toward characterizing their products with non-Dow pads. Investors are showing renewed confidence in the ability to compete successfully with Dow, with new investment rounds secured for innoPad and NexPlanar.
Given the inexact nature of market forecasting, this is good enough for me to believe that CMP pads will eventually yield to the Rule of Three, plus or minus.