A few words about the semi industry

March 29, 2006

View from Silicon Valley- A few words about the semi industry.

As the certainty of a strong 2006 for the semi business builds, we thought it would be useful to step back and look at some numbers. (I know, I know, so what else is new?) Are the numbers really up? Which numbers? How do the top-line numbers compare to the bottom? (Does anybody remember the bottom line is supposed to be the point?)

Before we embark, I just want to warn you this is not MBA-level material. We are, however, going to exercise the heck out of our seventh-grade algebra skills. Following in the footsteps of the joke about, "How many psychiatrists does it take to change a light bulb?," the reader has to "want to see the numbers."

Unlike the rocket-science calculations allegedly used by Wall Street pros, it's not hard to see if you really want to look.

First up, Semiconductor Industry Association (SIA) sales results over the last few years:

2000 $204M

2001 $139M

2002 $141M

2003 $166M

2004 $213M

2005 $227M

Reading the press, nobody seems to remember sales fell -32% in 2001. That 2001 sales represented only a 10% gain in three years from 1998 and nearly zero gain from 1997.

Not to worry, the industry touts and pundits proclaim, the 2000 peak was all dot-com boom money with no real profits or business plans. The 2004 and 2005 numbers are really sustainable. Since January is a seasonally weak month for SIA sales, 2006 will be even stronger. "This time it's different!" No, really!!

Time will tell. More importantly, these are only top-line numbers. What about the bottom line? Say what you will about the quality of the customers in 2000, SIA members saw across-the-board fat bottom lines during this allegedly "false" boom. How do the bottom lines in 2004 and 2005 compare?

There are a number of services out there which claim to provide this info. (A couple of whom seem to offer terribly self-serving results.) For better or worse, we subscribe to none of them. Denali surmised in late-2004 memory supplier profits were topping out at roughly half those of 2000 with declines in 2005 but an exhaustive study is beyond the bandwidth of a free site. (We will share a post if readers want to chime in with their own estimates...) We understand the P/E of SOXX components is running quite high and is 50x+ if you exclude Intel and Texas Instruments.

What we can explicitly measure are square inches of wafer produced quarterly, expressed in millions. This data tells an interesting story:

Sales MuInch.^2 $/Sq.Inc

1997 $138M not available

1998 $126M " "

1999 $149M " "

2000 $204M 5,552 $36.74

2001 $139M 3,940 $35.28

2002 $141M 4,681 $30.12

2003 $166M 5,147 $32.25

2004 $213M 6,263 $34.01

2005 $227M 6,645 $34.16

Most readers remember 2002 and 2003 were THE two most miserable profit years in the history of the semi biz. While our crude calculation does not allow for capital depreciation, among the prime killers in those 2002 and 2003 losses. It also does not allow for stock buybacks nor any of the other financial maneuvers routinely employed to fluff up earnings so that management and employees can cash in their options at the expense of shareholders. (The latter's impact on the debt load of some of these companies is significant.)

In short, "dollars per wafer square inch" is crude but it's our metric for the day. Let's run with it.

Our recollection of events implies 2000's $36.74 was wildly profitable while 2002's $32.25 is deep in the red. Halfway in between the high and low is $34.50 (or $33.43 using 2002, or $33.96 if you average them). For the purposes of today's "fun with numbers" and showing off our seventh grade algebra, hold your nose while we average those three numbers and somewhat-arbitrarily call it breakeven at $33.96.

Now let's break down the last few quarters of SIA sales and wafer shipments:

Sales Inches^2 $/Inch^2

1Q2004 $48.86M 1,529 $31.95

2Q2004 $53.51M 1,619 $33.05

3Q2004 $55.55M 1,629 $34.10

4Q2004 $55.10M 1,486 $37.08

1Q2005 $55.08M 1,465 $37.60

2Q2005 $53.87M 1,606 $33.54

3Q2005 $58.67M 1,748 $33.56

4Q2005 $59.86M 1,826 $32.78

If $33.96 is breakeven, the semi industry wasn't very strong the last two years. Even if you can barely get up from rolling on the floor laughing at our metric, you have to admit sales per square inch has clearly fallen the last three quarters.

Also relevant to note is all eight quarterly wafer area shipments exceeded even the highest total during the 2000 boom. The last two quarters in 2005 represent successive all-time high in wafer area shipments.

Another point which the above comparisons do not capture is the fab process migration. The "sweet spot" in 2000 was .18 micron while today it's at least down to .13 micron (a 28% increase in die per wafer or "effective inches" of wafer area). An argument can be made that all "new" wafer capacity of late is .11 micron (39% increase) or 90nm (a 50% increase in chips per unit of wafer area).

In this week's tribute to Wall Street "analysis," I closed my eyes, clicked my ruby-red-slippered heels three times and thus decree this works out to a 1% per quarter net increase across the industry. (Anyone recognize our methodology? Anyone think 1% is too conservative?)

Adding another column to our table, you get:

Sales Inches^2 $/Inch^2 $/Eff Inc^2

1Q2004 $48.86M 1,529 $31.95 --

2Q2004 $53.51M 1,619 $33.05 $32.72

3Q2004 $55.55M 1,629 $34.10 $33.43

4Q2004 $55.10M 1,486 $37.08 $36.00

1Q2005 $55.08M 1,465 $37.60 $36.15

2Q2005 $53.87M 1,606 $33.54 $31.94

3Q2005 $58.67M 1,748 $33.56 $31.66

4Q2005 $59.86M 1,826 $32.78 $30.63

Now combine this trend of declining dollars per wafer square inch to the avalanche of even more new 300mm wafer capacity coming on-stream over the next several quarters. (Scan the titles of the articles on the web site at: http://www.viewfromsiliconvalley.com/id114.html if you haven't noticed the stream of announcements.)

Conclusion:

In today's era of EBITDA or cash-flow or special-exception earnings (and sometimes all three and two more besides), it's wickedly difficult to tell what's really going on at the typical semiconductor company. Regardless of what price per square inch of wafer space you feel is breakeven, the above exercise opens a window for measuring the strength of the business. Are three consecutive quarterly declines in dollars per square inch a signal?

* * * * *

The above is not intended as advice to buy, sell or hold any stock, bond nor any other financial product. Invest at your own risk.