As we discussed last week, investment professionals are always looking for an "edge."  As Mark Cuban reportedly said, “When you sit at the business table you always look for the sucker or fool.  If you don’t see one, it’s you."

Beyond last week's report showing wafer area shipments surging faster than semiconductor sales, not to mention to an all-time wafer area record, something else we think nobody else seems to be considering is directly related to that tidbit.

DRAMeXchange reports on all aspects of memory.  They obviously focus on DRAM IC and module activity, but also follow NAND, NOR and even SRAM market conditions.  Of course, in commodity parts like these, the biggest market "condition" is pricing.

Back in April, the SOXX was 382, the street was pounding the table that PC growth on top of DRAM growth per PC would cause semiconductor, especially DRAM manufacturer's, earnings to ratchet upward.  Sure enough, the SOXX catapulted from 382 to 430+ in five short weeks  ("Are we just perpetually bearish", May 24, 2005).

At the time of our mid-May'05 article, the 256MB DDR-400MHz market price was only $2.30 (vs. an industry average cost estimated $2.85 to $3.00) while 512Mb DDR2-533MHz was $5.89.   DDR's fall from $3.50 -$3.60 pricing in February was considered only temporary. (Denali memory report, April 6, 2005.) 

"Just wait," was the mantra.  Prices will take off in the second half of 2005.

By August 30 (one-third into 2H05's much-anticipated recovery), "I see cheerleaders" reported,  256MB DDR-400MHz prices had fallen back from a momentary pop to $2.65 down to $2.47.  This still-below-cost pricing was despite the alleged back-to-school sales push the street tells us to expect, and buy stocks for, at that time of year.

Last week (past the two-thirds point of 2H05's alleged recovery), DRAMeXchange chimed in again with, "Inventory clearout for both DRAM and NAND Flash chips."  In other words, the "smart money" memory brokers in Asia decided to get rid of their inventory.

DRAMeXhange reports 256MB DDR-400MHz prices are even lower than May's $2.30, now at only $2.17!  (Utilizing the increasingly acceptable "un-tested" material get pricing down to $1.92!)  Then 512Mb DDR2-533MHz chips were all the way down to $4.34, a -26% drop since May's forecast of a "recovery" in pricing and profits.

If this is a DRAM "recovery," what will actual weakness look like?

Some folks think DRAM sales "don't matter."  They claim NAND sales now drive the memory market.

However, current NAND prices show 8Gb prices are down -20% while 4Gb are down -32% since May.  Since volume are probably up more than these prices are down, total NAND sales are growing.  But what are the actual profits?

We will also note, as reported by Denali in June, 2005, Denali DRAMs were over 2.5x total NAND sales (http://www.viewfromsiliconvalley.com/id174.html).  With DRAMs losing money, how much profit does a DRAM manufacturer have to generate in NAND, assuming NAND actually turns a profit, just to break even?  What happens when that manufacturer needs to build a new state-of-the-art 300mm fab with a price tag running to $2B?  Where does the money come from?  (Besides Apple and Intel...)

Conclusion: Even if I buy into the story NAND sales will continue to increase units faster than prices fall, I have a hard time believing losses in DRAMs "don't matter."  I also have to stop and wonder why the brokers are doing an "Inventory clearout."  The obvious explanation is the "smart money" sees further weakness dead ahead.   Could it be related to the all-time record in "wafer area shipped" being broken?

* * * *
The above is strictly for entertainment purposes and should not be construed as advise to buy, hold or sell any security, property or financial product.