Industry Downturn Perspectives..Forward and Backward

Recent Results Signal Better Times Ahead; How Much Better?...Little Consensus, But Some Interesting Scenarios are Possible.

After being in something of a free-fall for the better part of six months, the past 30 days' results have offered some hope for the semiconductor and memory industries in terms of improved capacity utilization, price stability (if not price increases), and better demand filling up the order books. Intel's 2Q guidance, even off a horrible 1Q result (horrible, but only by Intel standards) that the worst is passed, is welcome news. Not unexpectedly, the worst suffering has been among memory makers, whose markets were the most out of balance in every way; tried and true industry 'ballast' product lines in linear ICs and discrete components, dropped hard but faced no huge restructuring issues. PCs seem to have turned out not as bad as expected. Wireless has fared likewise: down but not disappeared. While still grim, the grimmest views of last fall, wrapped in large uncertainties, have been moderated and sometimes reversed, as more perspective was gained over time, and the dimensions of the demand fall-off became more clear.

We would have to say, in many cases the industry 'overreacted' to the early barrage of bad news: companies with preemptive layoffs, dire forecasts for 2009 and beyond, and apocalyptic visions of the future. Not everyone acted such, but many did. Seeing the 4Q08 results, industry forecasts for 2009 plummeted to the 'worst chip recession ever', anticipating a drop off of 25-30%, and unspecific calls for 'overdue industry consolidation'.

With better business in March and April (still way below 'before' levels, though), most are now calling the bottom, saying it is not getting worse but that it is starting to turn for the better.

This time vs. 2000-01: Compared to the 2000-01 downturn, this one is quite different in some regards. Both were the consequence of bursting bubbles (though different bubbles, maybe the same loose cash), but more important was the run-up into the abrupt demand fall off. In the earlier downturn, the industry was hugely profitable right until it went off a cliff. This time, the industry had been grinding for more than a year before 4Q08's fall off: memory makers had lost $2B in 2007, mostly in 4Q07, DRAM prices had been under extreme and steady pressure for more than a year. NAND had its ups and downs in 2007, but 2008 was down, down, down.

Inventories: Over all products and in all markets, the world's ability to assess actual inventory levels throughout the whole ecosystem has improved markedly over time, due mostly to the pervasive use of computer management methodologies. The sale of an apple at a Piggly-Wiggly market in New Hampshire is instantly registered at the Corporate headquarters in South Carolina, and a replacement apple is scheduled to ship in a week.

It is the same with semiconductors. But when the demand schedule changed so abruptly as it did last fall, and visions of The Great Depression fill the public dialog, it became immensely more difficult to determine "what the right inventory level should be"...inventory being a function of how much you needed in stock to avoid line-down situations and in anticipation of new orders. The guiding stars of demand and production management disappeared for a time. But the cloud cover is moving away over the past month, orders that were filled by inventory in October through February, now need to be filled with new production, and we can now see much more clearly our forward order book.

That DRAM prices only budged upward in the past month or so is some evidence of the amount of excess inventories and production capacity sloshing around in the system as we rolled into late 2008. In the past 4-5 months, tens of thousands of 200mm wafer starts/month have been ramped down to nothing; forward motion to the 50nm node has slowed due to high transition costs; Taiwan has idled huge amounts of near-prime 70nm capacity; Qimonda's 10% worldwide DRAM marketshare production machine has been shuttered in Virginia and in Dresden, plus at its foundries in Winbond and Inotera. Even with all this take-down in production capacity, prices moved up almost not at all through the end of March.

NAND inventories also suffered from a sharp downward shift in demand expectations: Actual year-over-year production of 130% annual GB growth for 2008 through the summer months, were instantly written down to just 90% growth for 2009 on 1 October 2008; only the absence of Taiwanese in NAND kept that market from getting too far out of control, though the "NAND Quartet" did their best to outrun the market, with annual CapEx levels that were close to annual NAND sales, and with (again in retrospect) the requisite production glut. NAND makers have invested 'with reckless abandon' for years, and now they overshot the market in the face of weakening demand. Apple coming in with an upside demand at the end of 1Q09 may have saved their collective skins for a while, but like OPEC, the NAND industry's idle wafer starts are a visible temptation to make more product, and capture more of today's higher prices.

So, today, though it is all quiet and prices are uncharacteristically stable for DRAMs and NAND, and both up from their recent lows, respected iSuppli Analyst, Nam Hyung Kim cautions that excess production capacity remains a major threat in both markets if prices go much higher, and, like some low-grade infection, will prevent a true market recovery unless more actions are taken or more time passes, for demand to creep back up. Of course, with NAND growing so much faster in GB/year than DRAMs, it is likely to come into balance earlier.

But whatever the future holds (and with some tempering of 'industry disaster scenarios' mixed with 'we are far from out of the woods'), there are still some analysts willing to make the case that the worse is passed and 'something good for vendors' in near at hand, if the overall economy cooperates in sufficiently restoring demand growth. One can read about this in the works of Future Horizons' and IC Insights' recent market newsletters.

IC Insights' outlook: This possibility is intriguing and worth understanding. It argues that the industry might be in for an upside surprise in terms of tight capacity, AUPs and associated revenue growth in 2H09. It is based on a few rather indisputable facts about the production and investment side of the industry, and some not-so-indisputable understandings about industry behavior.

(1) In 1Q and 2Q08, according to the quarterly SICAS report, industry capacity was tight all over, especially in leading-edge lithographies, with 0.12um capacity utilization running at more than 90-95% of available capacity. Then ("way back then"), the reasonable concern was that if the industry continued even at its then-current demand growth levels, capacity utilization would top out, causing shortages and arresting price declines.

(2) Though memory makers had made huge and irresponsible (in retrospect) capacity investments in 2006 and 2007, CapEx overall for 2008 was retreating at a substantial rate, and eventually dropped by more than 25% from 2007's elevated levels; to wit, the overall industry in 2H08 was arguably NOT making enough investment in new capacity to support moderately optimistic growth forecasts; excessive investments in memory fabs, but not enough in logic fabs for MPUs, ASICs, and ASSPs. TSMC and UMC were going forward with tepid expansion plans as we rolled through 2008, even in the face of the industry shift to more fabless companies, and those fabless in fact or fabless wannabees, reluctant to row their own boat for 65nm and 45nm process. These foundries' fab utilization fell off a cliff in 4Q08 and 1Q09, but was showing good recovery in March 2009 and into 2Q09.

(3) Especially in memory, but not limited to memory, price declines are taken as a given, and chip markets grow by increasing sales volumes faster than prices decline. Both DRAMs and NAND experienced huge GB and unit shipment growth in 2007 and 2008, which were about equal to the best two years in the past decade. But revenue growth was curtailed by the industry's massive capacity expansion that caused prices to also drop at record rates: 50% to 60% year over year declines is a good estimate of what vendors faced.

But, as we have seen from time to time in the industry (notably 1992-95 in DRAMs), if prices cease to drop, 'volume growth' flows straight through to become 'revenue growth'; the DRAM market, during that era of unexpectedly stable DRAM prices, went from $8.5B to $13B to $21B to $41B in consecutive years to 1995, as Japanese makers retreated and Koreans advanced but were too small to fill the vacuum. Bit growth almost translated directly into revenue growth, as 16M DRAMs cost about $50/unit for more than three years.

IC Insights looks at the industry from an Average Unit Price (AUP) basis, which, with all the necessary caveats in dealing with huge aggregations of products and wide range of prices, can be useful in understanding 'our revenue future'. In most terms, users' purchases are insensitive to price stability and even price-ups...so long as they can see they are paying the same as everyone else, and not losing out competitively. Tough markets drive down industry AUPs, in aggregate, but shortages of capacity can arrest price declines and reverse them. Then unit shipments and GB growth become additive to the price uplifts, to make the world once again beautiful for memory makers.

CapEx for 2009 is expected to be down another 40%+ from 2008, and substantial 200mm capacity has been decommissioned, esp. among memory makers...in effect, cancelling out SOME of those excessive CapEx in 2004-5-6. While it is hard to calculate with any exactness, "memory industry GB production capacity", taking into account relevant CapEx levels, process shrinks and productivity improvements, as well as capacity taken off line permanently, perhaps memory production capacity in 2009 is only about equal to that which was in place or in progress, at the end of 2007. Furthermore, having been on a diet for two years, further significant increases may be very hard to jump start in response to any recovery of demand growth that might occur. Even if DRAM and NAND prices come up, it will take a year for vendors to get the right cash flow and the right confidence, and the better part of another year to buy install and ramp new tools into production.

As noted above, however, and especially in memories, there are good reasons to believe that the markets are still so over-inventoried and over-production-capacity, that demand will have to grow for a very long time before it again bumps up against limited supplies.

We do feel, however, that important milestones have recently been crossed in supply take-downs and CapEx constrictions that are already having payback in better pricing; tight cash, continued restraints and a modicum of demand recovery can do much to improve the prospects of memory makers for the remainder of 2009 and into 2010.