TAIWAN: According to DRAMeXchange, a research department of Trendforce Corp., 2H’May contract prices will remain flat from the previous period. DDR3 2GB’s average contract price is $18.75 (1Gb $1.02) and 4GB is $36.5 (2Gb $2.12). The momentum in price increase, derived from the influence of the Japan quake, is weakening.
Shin-Etsu and SUMCO have improved the poor yield in the 40nm production. Also, the yield rates are going back to the previous full capacity. Both Shin-Etsu and SUMCO have made promising progresses in getting back to the previous capacity. Shortage in raw wafer is no longer a concern. Slow PC demands and the increasing DRAM inventories of PC OEMs (to around four weeks), contract prices in June are susceptible to changes.
DRAM migration has reached bottleneck, entry barrier is established with declining supply bit growth
After the fierce competition in 2008, DRAM makers have accelerated their post-70nm technology migration, and time to market has become the most critical factor to their profitability. We are now seeing two new technologies in a year, comparing to one technology that lasted for years. DRAM makers are moving into 40nm, and even 30nm domain.
In general, each migration step brings around 20 percent to 30 percent increase in bit growth. The improvement and the stability of the yield rates have surfaced as serious road blocks. Some makers have demonstrated technology migration is evidenced very challenging.
Samsung outperforms any other DRAM makers. Its 35nm technology will produce 30 percent of the total output by the end of 2Q11 and over 50 percent output by the end of the year. Hynix is taking on the challenges from both 38nm technology and 6F2 layout. Its output from 38nm will be very limited this year. Elpida has fully migrated to 45nm in 4Q10 and will commence mass production on 38nm in June with the target to raise the portion to 50 percent by end of 2011.
Nanya and Inotera are concentrating on their moves to 42nm and the corresponding yield rate improvements. A further move into 30nm is a possibility at the very end of the year. Other than technologies, the makers are also allocating more wafers to higher margin products like server and mobile DRAMs to improvement the profitability of their own.
The race for finer technologies is on. DRAM makers will continue on their refinements of 30nm technologies and yield rates. As the DRAM makers further move into the 20nm frontier, there lies a bigger challenge to the validity of Moore’s Law. Beside the design and stability requirements, the expensive EUV equipment will also pose as an entry barrier for the competition into the next generation. Only a handful of highly profitable makers can plunge enough capex to stay in the game.
The only good news for the rest of the group is that we expect 30nm will remain as the mainstream technology until 2012. Hopefully, it will leave enough space for the Taiwanese DRAM vendors to catch up. DRAM bit growth until then will slow down and the price pressure will be eased on the over-supplied commodity DRAM. Hopefully, the DRAM prices will come back up to rational and profitable level.