The scare of overcapacity
Christian Gregor Dieseldorff, SEMI Industry Research & Statistics, San Jose, CA
How changes in capex affect fab capacity
Reports have surfaced recently that indicate companies including TSMC, Samsung, and Intel may change their capital expenditure plans for 2013. A number of market research firms have revised their economic scenarios and forecasts, including updates for both semiconductor industry revenues and fab spending.
For example, TSMC may increase its 2013 capex by about US$ 2 billion ??? to a robust $10 billion. Another media report in September indicated that Samsung may reduce 2013 spending, perhaps even significantly. Historically, Samsung cut capex in 2002 (by -30%), 2008 (-22%), and 2009 (-44%) ??? all years of economic depression. But can Samsung maintain its market share with a -30% or even -50% cut in capex?
Some media reports speculate that Intel may cut 2013 capex. Intel's equipment spending for 2012 was actually a bit lower than the year before. If Intel reduces total capex in 2013, SEMI's fab analysis suggests that Intel's Front-End equipment spending will increase as construction spending is expected to be a lower percentage of overall capex. Editor's note: This paragraph was edited after the magazine went to print, to emphasize Intel's specific capex expectations and deemphasize the impact on industrywide equipment spending.
The question remains: How will changes in capex affect capacity? An increase in installed capacity may lead to oversupply, which will then drive down average selling prices and subsequently revenue.
Capex affects capacity
If significant changes in capex occur, installed capacity is likely to be affected. Since the last economic downturn in 2009, the industry has seen a change in paradigm. The average yearly growth rate from 2003 to 2007 (before the downturn) was 16%. From 2010 to 2015 (after the worst of the downturn), it is expected to range from 6% to 7%. Before the downturn, most fab spending went into new installed capacity; however, more money is now being spent to upgrade existing facilities (Fig. 1).
Figure 1. Fab equipment spending vs. change of installed capacity.
Except for foundries and system LSI (SLSI), almost all segments of the industry have seen a decrease in the rate of new capacity additions in 2012 (Fig. 2). The foundry industry segment has greater installed capacity ??? both in absolute terms and in the rate of growth in comparison to other segments. Given industry consolidation and as more IDMs change to a fab-lite or fabless business model, foundries become even more important to the supply chain that provides manufacturing technology. The dedicated foundry sector has added capacity at greater than a 10% rate year-over-year (YoY) since 2010, and this trend will continue going into 2013.
Figure 2. Except for foundries and system LSI (SLSI), almost all segments of the industry have seen a decrease in the rate of new capacity additions in 2012.
Most current spending for the MPU sector goes into existing fabs. While no new capacity is planned for 2012, a couple of new fabs will begin ramping in 2013, pushing installed capacity up by 4%.
Cutbacks were made in NAND flash memory production in 2012 ??? by companies including Sandisk, Toshiba, Samsung, and Micron ??? in order to improve average selling prices. Anticipating higher demand, companies will likely need to increase flash capacity in 2013. While flash growth will increase only 4% YoY in 2012, it's expected to double in 2013 to 8%.
More consolidation is predicted in the DRAM market. Nanya is expected to exit the PC DRAM market, and Micron may acquire Elpida by 1H13. Ultimately, only three major DRAM makers will remain; no noticeable increase of installed capacity is expected in 2012 and into 2013.
Samsung's heavy investments in system LSI will singlehandedly spike overall SLSI capacity. One example is the recently announced $4 billion conversion of its Austin, TX, fab from flash to 28nm SoC logic devices. Installed capacity for system LSI shows healthy growth since 2010 ??? about 20% every year, with an expected repeat performance in 2013.
Samsung has begun an aggressive conversion of up to four existing memory lines to system LSI. A transition from flash to system LSI is not easy to manage, so a drop in memory capacity is expected. The company is expected to compensate for this memory capacity by continuing to ramp its Line 16 fab and constructing a new fab in Xian, China, with a massive investment of $7 billion; construction started in mid-September 2012. SEMI's World Fab Forecast gives more detail about the ramping phases of this leading-edge flash fab with huge potential capacity.
Needed: More 300mm fabs?
Merely shrinking die may not compensate for increasing demand for chips, such as for mobile and automotive applications. Volume production for the first 450mm fabs may not occur until 2018; it may take even longer until the cost-benefit ratio for 450mm volume fabs makes it an attractive option for all potential chipmakers. While 450mm production catches up to demand, the industry will still need additional 300mm fabs, especially for system LSI and flash. Of the 300mm NAND flash fabs currently in construction, ramping, or still planned, all are expected to reach full capacity by 2016 (Fig. 3).
Figure 3. Installed capacity of 300mm volume flash fabs. Actual data available in World Fab Forecast report.
The industry has a changed paradigm in that more fab equipment spending is allocated to upgrade existing capacity rather than adding new capacity. The industry reacts quickly to changes in average selling prices by reducing capacity outputs and slowing down expansion. In the five years before the economic downturn, capacity growth rates were in the double-digits. However, since the downturn, growth YoY has been more moderate, in the low single-digits.
The assumptions in any forecast are related to macroeconomic and political factors, like the Eurozone debt crisis, China's economy, the U.S. fiscal situation, turmoil in the Middle East, and the U.S. election. At this point, worldwide GDP for 2013 is expected to be higher than in 2012, while semiconductor revenue is expected to be heading for the upper single-digits in 2013. While some data suggests that 2013 front-end fab equipment growth could be as high as 17%, a range of 12% ($40.3 billion) to 14% ($41.1 billion) is more likely, in light of recent changes in the marketplace.
Solid State Technology | Volume 55 | Issue 9 | November 2012