If We Build It, A 450 mm Wafer Fab, Will Customers Come?

Implications: 1. New Markets will be created, enabled by the 450mm revolution. 2. Opportunity is there for new equipment strategic alliances with major chip players. 3. Smarter, simpler electronics projects need very intelligent, very large 450mm wafer-based chips.

Analysis: If we believe a survey done in 2007, the expected arrival date of 450-mm wafers fabs may be 2013. Then nearly 40 percent said 450-mm wafers would never happen. Now we read that Intel, Samsung, TSMC want 450-mm wafer pilot lines in 2012. These companies intend this target date as the start for 450-mm pilot line, not high-volume production lines.   Just as in the pre-300 mm fab era, many now doubt whether 450 mm fabs will appear. They feel that the risks and costs of 450 mm overshadow any benefits. Semiconductor manufacturing equipment vendors have called 450mm fabs flawed and misguided. The risks are that 450mm fab tool development costs could run as high as $100 billion for semiconductor manufacturing equipment vendors. Leading companies that make semiconductor manufacturing equipment, have said time and time again that investments in 300-mm manufacturing equipment have yet to be recouped. They believe that moving to 450 mm wafer size would yield little or no cost benefit.  However, Intel, Samsung, TSMC and Toshiba agree that a 30% percent reduction in cost per wafer over time will be achieved by 450-mm fabs. This cost reduction may expand many electronics markets, notably consumer, and many emerging geographical regions. Some experts challenge assumptions that 450-mm equipment will have the same throughput as an existing 300-mm machines. Their models for such fab equipment as etch, ion implantation, inspection and lithography challenge throughput assumptions. Assuming there are 400 dice per wafer, a 300-mm tool with a throughput of 150 wafers per hour could process 60,000 die per hour. Larger 450-mm wafers would more than double the die count to 920. Thus, a 450-mm tool with the same wafer throughput could process 138,000 die per hour. That is where models fall apart, claim equipment manufacturers. They fear that larger 450 mm wafers, having much finer than 300 mm geometry, processed on 450-mm tools will process but 74,400 dice per hour. That is a far cry from 138,000 die per hour. Thus chip makers would need to buy twice as many costly 450-mm tools. Further clouding sales prospects is that equipment vendors may have a customer base of just 15 companies who will build leading-edge 450 mm fabs compared to the 27 who now own 300 mm fabs. Who will drop out at 450 mm?