The thrust for this equipment spending is at the foundry level followed by NAND flash memory manufacturing, according to SEMI's presentation at SEMICON on 6/9/12, and as summarized by ElectroIQ. The reasons are pretty much what you've already guessed, the continued high demand for mobile devices from smartphones to tablets (for more detailed data, see the article by ElectroIQ).
Furthermore, ElectroIQ, quotes Ken Dulaney, Gartner's VP & Distinguished Analyst, who spoke during the SEMI/Gartner Market Day at SEMICON West, as expecting enterprise clients to "largely skip Windows 8 because the conversion to Win7 was so recent, many custom applications are still being ported to Win7 and Windows 9 is already anticipated for 2014." Additionally, the wider adoption of tablets to replace industry required manuals, such as is the adoption trend in airline cockpits, according to Dulaney, and the potential for enterprises to drop RIMM's Blackberry line and switch to Apple's iPhone devices will further push the market for handheld mobiles.
Based on post-SEMICON financial and industry analyst meetings, with particular attention to Citi's analyst reports, ElectroIQ summarized the foundry tool situation as follows:
Speculation has been "markedly negative" on H2 2012 semiconductor demand, Citi says, including talk of NAND capitulation, skepticism that foundry might weaken, and questions on whether Intel might reduce its blockbuster capex. Meetings at SEMICON West are largely confirming tool suppliers' suspicions that there will be weak orders in Q3, picking up in Q4. Part of 3Q’s weakness may be amplified by heightened seasonality that arises given higher customer concentration, Whalen says. However, Citi accepts the potential for improving semiconductor demand into Q3, which might strengthen capex in Q4.
Foundry orders will rebound sooner in H2 2012 than NAND, which see increases in H1 2013, Citi predicts. Based on its interviews with Lam Research and others, Citi expects flat or slightly better foundry capex in 2013, driven by steady 28nm deployments.
One of the associated questions to these discussions is the state of 450mm switch-overs. With CAPEX spending running the gamut presently, the costs of fabrication rising dramatically, and then the shrinking margins, what do analysts expect with the move to 450mm? WSJ today reported on the latest transition question, the highlights follow:
“There was a false start,” agreed Mark Fissel, a vice president at Lam Research, another company considering development of 450-millimeter production tools.
Another factor for equipment makers this time is a dwindling number of customers. Hasserjian put the number of companies who built 200-millimeter factories at 75, while 25 built 300-millimeter factories.
At 450 millimeters, the number could shrink much further, amid estimates that the cost of fully equipped factories may rise from $4 billion or so to $10 billion. "I can't get beyond five, maybe six companies total" that will build 450-millimeter factories, [Applied Materials' corporate VP, Kirk] Hasserjian said.
As ElectronicsWeekly noted,
Whereas 75 companies built 200mm fabs, only 25 built 300mm fabs […] At yesterday's IFS 2012, Future Horizons CTO Mike Bryant said that 450mm fabs will cost $10bn and that the companies which build them are, in order of building: Intel, TSMC, Samsung (all by 2021) followed by Globalfoundries, Toshiba, Hynix, Micron, UMC, ST/Infineon, SMIC, Renesas (assuming a new management team), Texas Instruments.
As we transition to 450mm over the next five years, the playing field will certainly shrink because of the costs involved in retooling. Looking at the semiconductor supply chain, the continued reduction of companies with fabs will have a significant impact on the industry in terms of pricing, availability, capacity, and the impact of disruptive events such as natural disasters.