by James Montgomery, News Editor
Six months after drastically rewriting their outlooks for a bleak 2007, most analysts surveyed by WaferNEWS appear to be moderating their expectations a bit. Most outlooks for 2007 equipment sales and capex are now a little better than they were in July, although across the board top-end growth predictions for 2008 capex are a little less ambitious than previously thought — the top estimate is now 23%, instead of 33%, while most others are in the low-teens. Outlooks for 2008 equipment sales generally haven’t changed much, still in a range of low to high teens, with only Gartner anticipating 20% growth.
Frontend equipment segments seen with best growth in 2007 include lithography (e.g., 193nm steppers), particularly for memory usage; process control, particularly for 45nm; macrodefect detection; and implant (medium current), also driven by memory, according to Klaus Rinnen from Gartner Dataquest. Most of the final manufacturing (testing/assembly/packaging) equipment segment will see a decline this year, but areas eking out some modest growth include wafer-level package inspection, bonders, and testers for memory devices, with growing adoption of systems-on-chip, and systems-in-package.
Most analysts predict overall capacity utilization rates will fall into the mid-80% range during 1Q07 before a slow climb back to the low-90% by early 2008. Memory firms are seen growing more cautions as the year progresses, with logic/foundry investments recovering in 2H07, although weakness in the device end-market could trigger rapid cutbacks in investments. Logic spending should be lower than in 2006, mainly because of pricing issues involving Intel, which already has said it wants to achieve “capex avoidance” of a couple billion dollars through the next two years. George Burns, president of Strategic Marketing Associates, pointed out that memory firms are bringing a lot of fabs online at the same time (DRAM last year, and flash this year), with the bulk of that in Taiwan, but capacity additions are coming in stages, which should keep equipment investments flowing.
Capex factors stabilizing, economics still unknown
A greater mix of electronics end-devices means a healthy diversity of demand, and less reliance on PCs (or laptops) provides some stability for the semiconductor industry, and thus stabilizes its investment plans too. Major market influences to consider for chipmakers’ 2007 capital investments include the adoption rate of Microsoft’s Vista OS by both consumers and businesses, and a rebound in ASPs. Rinnen noted that Vista’s arrival in 1Q07 won’t be a “miracle cure,” but it keeps the PC segment “fresh” alongside other end-use applications competing for dollars. A potential risk for downside in 2007 is that if memory makers don’t see the profitability they enjoyed in 2006, particularly in DRAM, it will affect their spending, noted VLSI Research’s Risto Puhakka. Semico’s Jim Feldhan warns of softness from cell phone sales, and that there could be another surprise disappointment in MPU and flash memory ASPs.
Analysts across the board agreed that one of the most influential and hardest-to-predict factors will be the macroeconomic environment. “Nobody is really able to forecast macroeconomics successfully,” noted Burns, pointing to potential unknowable occurrences such as an Asian financial crisis, or a shift in currency valuations vs. the US dollar, which could impact chipmakers’ sales and spending plans. Rinnen agreed that a slower macroeconomic environment is a possibility, which could elevate inventory levels.
Looking ahead, come 2009, “the party’s over,” quipped Rinnen, as the drivers of 2008 (housing rebound, US presidential election, Olympic games) fade, and an oversupply situation in memory takes its toll. Robert Castellano of the Information Network reiterated the sentiment that dominated the July 2006 forecast downgrades, that the industry still appears to be “robbing Peter to pay Paul.”
Will memory remember to throttle back?
A perennial concern in semiconductor capital spending is the discipline, if any, practiced by both NAND flash and DRAM memory firms. The consensus is that although memory firms have become a bit more responsible than they used to be, ultimately it’s a commodity-driven business where you “make everything you can and sell for whatever price you can get,” noted Future Horizons’ Malcolm Penn.
And so, look for memory spending to get even more out of whack, accounting for ~49% of total chip industry spending in 2007, up from 46% last year. And memory firms’ capital intensity (capex vs. revenues) is approaching 50%, vs. ~20-22% for the total chip industry — that’s above the previous peak years of 1995, 2000, and 2004, noted Gartner’s Rinnen. With continued pressure on NAND prices, he wonders how much longer the industry can absorb these imbalanced fundamentals before reevaluating the bottom line and adjusting spending habits. The real question is, he asks, “Are we going to slow down before we hit the wall, or are we going to run right into it?”
NAND flash is suffering through slowing growth and punishing price declines, amid mountainous supplies based on even bigger promises of future end-application use. Rinnen said that 2000-2009 bit demand for NAND flash could be as much as 15.4 trillion, a mind-numbing 11,000-fold bit growth for the decade. Puhakka noted that NAND flash growth is clearly slowing due to pricing, which indicates oversupplies, and if flash makers still press ahead with ambitious spending plans, “we will definitely see problems.” The emergence of new applications could rescue NAND flash oversupply, but he thinks there isn’t anything out there near-term to support 200% annual bit growth, though solid-state hard drives appear to be the best bet for the future. “As good as Steve Jobs is, there are only so many iPods, or cell phones, that can be sold.” — J.M.