October 28, 2004
View from Silicon Valley - Building inventory and building fabs
www.viewfromsiliconvalley.com

Long-time semiconductor salesmen are fond of saying there are only a few nanoseconds in every cycle when both plenty of parts and plenty of purchase orders are available. The rest of the time you have plenty of one but not the other.  It's either "allocation" or recession.  Feast or famine.

Per an on-line commentary a few weeks ago(&), Cypress reported "'widespread buildup of inventory at key customers.' Same song, 87th verse."  There have not been 87 reports of softness but they correctly observed many, many semiconductor companies, large and small, have reported "soft" conditions (fewer orders), in multiple markets over the last quarter.  Inventory is starting to pile up. What is the significance?

A correspondent of mine captured what seems to be the prevailing Wall Street sentiment on this topic perfectly: "I'm aware that the news flow could be come positive as Street sentiment is still awful... and... 4Q is always the strongest time for demand. So any positive data points on the margin will likely be viewed positively and many will 'hope' the cycle just took a quick breather."

In other words, the semi guys and Wall Street would have us believe this inventory accumulation is just a transitory blip in a still-strong growth cycle. In the month since I received her input, many analysts are indeed proclaiming "seasonal" strength going into the end of the year.  (Let's not get into if or how the election might also impact semi business or stocks in general.)

Nonetheless, if orders are slowing, inventory is accumulating. Semiconductor fabs are not like car factories. You can't shut a fab down for two weeks and just turn it back on when everyone returns from their two-week furlough.  The "story" here is inventory.  It's the elephant in the room nobody wants to admit is there.

Perhaps the world's most famous inventory today belongs to Intel, who reported working down a lifetime high $3.223B July inventory to "only" $3.118B (-3.3%) in October. This inventory reduction was accompanied by an unspecified inventory write-down.  Intel burned $2.5B cash last quarter on a stock buy back but refused repeated requests for details on the scope of this inventory write-down.  (Count on these parts becoming a piggybank to assist in future earnings management.  I would also bet the write-down was bigger than the $125M balance sheet inventory decrease.)  Bottom line, Intel knew they had a big fat target on their head to reduce inventory and they still had to resort to undisclosed financial engineering to barely "accomplish" the goal.

Coming at the issue from a different direction, TSMC, the world's first and largest contract fab house, doesn't hold inventory but their customers do. On October 8, (*) TSMC, "said sales fell for the first time in seven months and... warned they could continue to drop as demand falters." Then, on October 26, TSMC, "forecast a surprisingly large drop in its factory utilization rate," and, "There is a more gloomy feeling," per Chairman Morris Chang's statements.  In other words, bad news on October 8 was replaced by worse news on the 26th.  Then in the two trading days after this last announcement, TSM went up 5%+. Heck, what do I know?

Other contract fabs tell stories similar to TSMC's(**):

"We see a slowdown now, but it's not severe,"  (Tower Semi President)

"There is quite an inventory correction taking place right now," (X-Fab VP)

"The visibility is not good out there," (Jazz Semi CTO)

"In Q1 2005, they (customers) are saying: 'We need to order again.' " (X-Fab VP)

If the above are the most positive things insiders can say, alarm bells should really be going off.  Especially when accompanied by comments about poor "visibility."

As an aside, "visibility" comments are disingenuous--- at best.  John Chambers paved the way for this dodge in 2000 /2001.  Most technology companies have millions, and even billions, of dollars in resources tied up building products for customers. Effective utilization of these resources (ROI, ROE, etc., take your pick) is dependent on visibility into the market. Managers of large companies who really do not have visibility are replaced by the board of directors with someone who can give the board visibility.  (Assuming the BoD is doing its job.)

The reality is these managers do have visibility but what they "see" is bad news.  They are waiting for visibility to "improve," which in this context means hoping "good news" can be found.  Once they can find, or concoct, some they will heroically report improved visibility.  At some point, investors will look back and be amazed the visibility comments were accepted and allowed to slide by.  (End of sermon, sorry...)

OK, even if there is all this inventory, so what?  Won't these companies just work off the excess over the next couple quarters?  Or take a one-time write-off with minimal impact on the value of the stock? Then growth and earnings will ramp right back up?  Right?

Unfortunately, all of the above comments and estimations, however valid or invalid, miss the "big picture."  This "big picture" has little to do with this quarter's forecasts, sales, inventories or visibility. (Although the inventory figures are a clue.) It has nothing to do with the next move by the Fed, the latest consumer confidence poll or which party wins the White House. (Although interest rates play a role.) The "big picture" involves stepping back from this "noise" and looking at semiconductor supply from a wider angle.

There are several data points not being accounted for in Wall Street's bullish case for semis. Strategic Marketing Associates(#), self described as "the semiconductor industry's fab information source since 1992"(##), and Semico were recently quoted as follows:

"Right now we're in a big fab boom, almost as big as... 2000"; (Whether it's being driven by cheap money or projected skyrocketing demand, does anyone still remember how the last fab boom turned?)

"In 2003, 500,000 200mm (8") wafers additional capacity was brought online, and the end of 2004 will have brought twice as much online. Those are 5% and 10% (global capacity) increases (respectively)."

"I think (the fab boom) is going to last for more than a year." "Eventually, it's going to be too much capacity, but not until the end of 2005."  (SMA's forecast was issued back in July and their inventory data at the time was only through March, 2004. As of March, 2004, SMA showed zero(!) y-o-y inventory build. This makes the current inventory accumulations all the more startling.)

"(Eleven) new 300mm sites (will) ramp in 2005, adding capacity to the 15 existing 300mm sites in production." (SMA shows 300mm/12" wafer capacity will grow from 450K/ month in May, 2004 to ~1,055K in 2005. A 605Ku 300mm wafer increase is equivalent to 1,360Ku new 200mm wafers, which is 4x+ the 2004 capacity increase!!)

"I think the fabs in China will price their products to gain market share, not necessarily to make a profit, and that's going to hurt the pricing structure of the industry."

"(M)any refurbished 6 in. and 8 in. lines are moving to China." (Unique to this build cycle, "old" fab equipment is not staying off-line after it is replaced by upgraded gear.)

Notice these capacity forecasts are based are based on projects already under way. Money is already being spent. I missed mention of any of this information by any of the mainstream press.

Conclusion:  Inventory is the canary in the coal mine. It tells us current global capacity already exceeds demand.  On top of this, there is an avalanche of new semiconductor wafer capacity coming on-stream in the immediate future. You tell me where semiconductor ASPs and profits are headed in the next few quarters. And what this means for Silicon Valley...




For more insight into what's really happening in Silicon Valley, visit www.viewfromsiliconvalley.com



(c) copyright View from Silicon Valley, 2004.  All rights reserved.