The technology gap between SMIC and the rest of the pack in pure play foundry is even wider. Of the 14 pure-play foundries ranked between fifth and 18th in sales, only four (TowerJazz, Grace/HHNEC, Dongbu, and Xinxin) are expected to be able to produce ICs using 90-nm or below feature sizes in 2012, IC Insights said. Even this production is likely to be relatively limited, the firm said.
Collectively, these 14 foundries are forecast to account for $4.6 billion in sales, or about 15 percent of the total pure-play IC foundry market in 2012, IC Insights said.
For foundries, keeping up with leading edge process technology can be accomplished in two ways, IC Insights said. One is through joint ventures and licensing agreements, such as the partnership between IBM and Globalfoundries. The other is through significantly increasing R&D spending to develop advanced technology, as TSMC has done, Globalfoundries said.
Revenue from 30-nm and below technologies is expected to account for about 30 percent of total pure-play foundry revenue in 2012, up from 22 percent in 2011, IC Insights said. Older technologies—80-nm and above—are expected to make up only 13 percent of sales in 2012, down from 14 percent in 2011 and 15 percent in 2010, IC Insights said.
Relying purely on licensing from partner to move up to the technology value chain is inherently risky. R&D, pilot production and mass production could be very different animals to be tamed. Even it is successful, a good portion of the profit will go to the licensing fee that could significantly undercut its capability to compete during tough time. That is what happened to Taiwanese DRAM industry. On the other hand, develop technology internally requires intensive investment, visionary management, time and miss opportunities. However, if done right, the payoff is significant and such companies are normally the market leaders in their industry or market segment.
Globalfoundries certainly has deep pocket but like to see short term results. Hopefully, they can be successful with their unique combination, and gain long-term competitiveness in pure play foundry. It would be good for the industry. Time will tell.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.