Ten years ago, the good news was that Japanese makers of mainframe computers caught up with IBM after decades of effort. The bad news was that they had caught up with IBM. That cynical joke came to mind as I read about efforts to consolidate Taiwanese DRAM producers. In both cases, producers found it harder to sustain sales growth or profits as their industries matured. In the early stages of a market, small competitors enter and find opportunities as leaders struggle to keep pace with market expansion. At some point, stable leadership evolves and revenue growth slows. Smaller producers lack economic scale to survive price competition and lose ability to finance expansion. That weakens their competitive position further.

Ten years ago, analysts compared the AMLCD (large TFT LCD) industry to the DRAM industry. I thought such analogies were over done because AMLCD producers were on the path toward 80% variable costs while DRAM producers remained on the path toward 80% fixed costs. Now I wonder if other analysts were so far wrong. Looking at sales shares of the top-9 producers of DRAM and AMLCD in 2H’08, I see similar cumulative distributions. The rivalry index for DRAM was 5.4 producers while the index for AMLCD was 5.3 producers. In the past ten years, the AMLCD industry caught up with the DRAM industry in terms of industry structure. According to Lanchester Theory, Samsung Electronics has reached a stable position in both markets: 30% of DRAM sales and 27% of AMLCD sales. A market share of 42% or more would be needed for an unassailable position (theoretically) but a share above 26% tends to stabilize a market. It is also interesting to note that two Korean producers had a 49% share of the DRAM market and a 49% share of the AMLCD market in 2H’09.

Figure 1  Cumulative 2H’08 Sales Share of Top Nine Producers

Taiwanese producers of AMLCD had a 35% share, significantly more than their 10% share of DRAM. I assume this is why Taiwan’s Ministry of Economic Affairs (MOEA) decided to create Taiwan Memory with an investment of about $850 million from the National Development Fund and promote additional investment of private capital. Both AMLCD and DRAM manufacturers on the island have been suffering from over-capacity but the condition of domestic DRAM makers is more severe. For AMLCD makers, MOEA officials have suggested consolidation but limited their actions to studying potential changes to financial accounting standards. They may allow AMLCD producers to remove depreciation charges from earnings reports entirely or to reschedule depreciation over more time periods.

Recent news reports and rumors about which DRAM makers will join Taiwan Memory and which technology they will use (from Elpida Memory in Japan or from Micron Technology in the US) reminds me of prior rumors about consolidation among AMLCD makers. In both cases, the family interests of plastics companies appear to impede plans others find compelling. In the case of DRAM, Formosa Plastic Group controls several small producers and in the case of AMLCD, Chi Mei Group owns 28% or more of Chi Mei Optoelectronics, which is suffering from capacity expansion. Some observers wonder if consolidation can happen in either industry.

I wonder if consolidation is a desirable goal in any case. After all, combining several smaller fabs does not create a bigger fab. The average substrate size does not increase; it might decrease. Purchasing scale might increase but that would require uniformity, which is unlikely if a motley collection of substrate sizes and technology bases are merged. Intelligent merger requires protracted effort by talented people and generates significant opportunity cost, as demonstrated by AU Optronics as it digested Quanta Display. In addition, if a government fosters such a merger, it might feel obligated to sustain the organization even if it proves to be unprofitable.