Roughly every few months, we end up talking about CAPEX. While it’s very “upstream” it’s also very relevant to the final component pricing and supply-demand that many are concerned with. This year we saw jaw-dropping announcements of unparalleled CAPEX spends by the major players at the fab level for new and upgraded facility spending (cf. here for one example, or click on 'CAPEX' in the tag cloud for a full listing). Not all of these announcements have resulted in actual financial comitments though.
Samsung's CAPEX and related business strategies have certainly paid off; with their major investments still moving ahead, they've played a smart card and are quickly moving to rival their competitors (and outpacing chip sales at Intel, as the numbers show here). By 2014-2015, it is likely that Samsung, not Intel, will take the IC throne, according to the same recent review of IC rankings by EETimes.
With a roughly 24 month window from financial commitment to production, the business and financial strategies surrounding fabs are mission critical. The high price tag (minimally in the US$ hundreds of millions, usually in the US$ billions) means that betting on the correct architecture, lines, and market demand stability is essential. Perhaps as a result, many of the CAPEX commitments we saw voiced earlier this year have either been temporarily suspended or shifted, especially in the case of memory shifts from DRAM (due to more ASP bottom-dropping events) in favor of NAND (due to increased demand from smart wireless devices (SWDs)). Granted, the diversification of those shifting the weighting of investment from DRAM to NAND is, simply put, smart. As it is, there are likely to be NAND shortages as we move into 2H11 and the further adoption of SWDs based on NAND flash solutions takes hold (see the upcoming MarketWatch Quarterly on the SSD-HDD debate, and this Quarterly article on SWDs and this one on memory shifts).
With SEMI's recent review and forecast of fab capacity and fab spending published, what are we to take away from the present CAPEX and fab story? Well, SEMI's numbers show a relatively modest fab spending increase for 2011 of roughly 8% year-over-year (same as for 2010) followed by roughly 9% for 2012. The take away message from these numbers is one of conservative corporate strategies for large expenditures. These numbers are particularly down in light of the new fab construction starts forecasted to see double digit drops for 2011 and 2012.
Meanwhile, on the fab capacity side, we're seeing good numbers and reasonable forecasts for the same next couple of years. As SEMI points out, "the current focus is on upgrading and ramping existing facilities; […] [T]he industry may not have enough capacity in the next two years, as new fabs slowly come on line. […] It appears that many companies are hesitating, waiting for more proof that the market has stabilized." (cf the same SEMI report, last section)
Demand for next generation architecture is there, as seen by the tremendous rise in SWD purchases and therewith, end-products (i.e., from tablets to smart phones, and now SSD-only notebooks will be tested). Those few companies who have stayed committed to their well devised CAPEX strategies to meet the upcoming demand (especially in NAND) will find the competition sparse and pricing likely to be strong in 2011-2012.