The following is a reprint of my January 16, 2007 RealMoney column

In a volatile market, investors tend to gravitate toward companies and investments that provide stability. As crazy as this may sound, I think that stability can be found in a semiconductor company – namely, Taiwan Semiconductor (TSM). I think the table below shows just how stable.

Taiwan Semiconductor Cash Flow Generation ($U.S. Billions)

Sources: Taiwan Semiconductor, Yahoo! Finance, William A. Trent estimates

Taiwan Semi operates in an unsexy part of the semiconductor industry known as “foundries.” It sounds as exciting as a blacksmith shop, and that isn’t far from the truth. Foundries don’t design any of the products they manufacture. Instead, they make the chips that other companies design. Their expertise isn’t in technology so much as process and efficiency.

Because they don’t design the chips themselves, Taiwan Semiconductor and other foundries such as United Microelectronics (UMC) typically get lower gross margins. The design profits fall to their customers. TSM’s expertise in manufacturing and economies, however, are much needed by customers who are often too small to absorb the enormous costs of building a chip fabrication plant.

Such customers include many fabless semiconductor companies and systems companies such as Altera (ALTR), Broadcom (BRCM - Annual Report), Marvell (MRVL - Annual Report), nVidia (NVDA), Qualcomm (QCOM) and VIA Technology, as well as integrated device manufacturing companies such as Advanced Micro Devices (AMD - Annual Report), Analog Devices (ADI), Freescale, and Philips (PHG).

Many small customers have given the company a balanced sales base. By end market, 40-45% of sales are communications-related, about 30% are to the computer market, 15-20% go to consumer electronics and the rest serve the memory and industrial markets. In 2006, the largest customer represented 10% of company sales, and the top ten amounted to just over half of sales. The lack of concentrated exposure to any customer or end market is one of the reasons TSM can generate stable cash flows.

The largest customer related risk factor may be that three quarters of sales are to customers in North America, and thus may impact the company if there is a U.S. recession. However, the global end markets for technology suggest that the true end customer is more widely dispersed geographically.

As the cash flow table shows, it seems fairly safe to say TSM will generate about $2.5 billion in cash flow. In some years, such as 2006, the cash flow may be unusually high. But even the industry downturns in 2004 and 2007 did relatively little harm. Given that the current enterprise value for Taiwan Semi is about $42 billion, it is offering a free cash flow yield of just under 6%.

If I had $42 billion that I wanted to invest safely, I might choose between buying TSM outright or investing it all in 5-year U.S. Treasuries. The Treasuries are currently yielding about 3.0%, so I would get $1.25 billion in interest each year from my investment. If that were my choice, I think I would go with the $2.5 billion in cash flow offered by Taiwan Semi.

It’s true that as a small investor owning a portion of TSM I would not be able to access all of the free cash flow. There is some risk to the comparison, since I am hoping the company invests any cash they hold onto wisely. But the company does pay two thirds of the cash flow as a dividend. Unless things change, that is still a 4.0% yield taxed at 15% compared to a 3.0% yield taxed at my marginal income tax rate.

How Bad Can it Get?

As stable as it may appear, I also have to acknowledge that TSM’s cash flow is not guaranteed. However, I think 2007 probably marked a fundamental bottom for the semiconductor industry – or at any rate that things won’t get much worse.

Consider, for example, the pricing environment. The Bureau of Labor Statistics reported that semiconductor prices declined 16.9% in December compared to the year earlier. That number was a modest improvement over November’s decline, which was the worst on record. Even the depths of the Internet bust were better times for semiconductor pricing. The fact that the pricing environment is so extraordinarily bad suggests to me that it probably won’t get too much worse.

Year/Year Change in Semiconductor Prices (PPI Data)

Source: Bureau of Labor Statistics

Furthermore, as I have written in other columns, I think the turnaround in semiconductor fundamentals is within sight. Pricing is a function of supply and demand, and since March of 2007 demand (semiconductor revenues as reported by the Semiconductor Industry Association) has been growing at a faster rate than supply (bookings for new semiconductor equipment as reported by Semiconductor Equipment and Materials International).

I think the industry’s recent restraint in adding new capacity will soon become evident in stronger pricing even if there is an economic slowdown. If I am right, what already looks like a solid and stable cash flow level could soon look even better.

Disclosure: William Trent has a long position in SMH.

My December 4, 2007 RealMoney article:

Recent data points are shaking my confidence in the near-term outlook for semiconductor stocks.

After being bearish for more than a year I had turned cautiously bullish earlier this year because it looked like potential supply (which I measure as orders for new semiconductor equipment) was coming back in line with demand (which I measure as the year/year change in semiconductor revenues. Unfortunately, the latest data show this trend weakening faster than I thought it would.

On Monday, the Semiconductor Industry Association (SIA) released their sales report for October, saying worldwide semiconductor sales rose to $23.1 billion, an increase of 5 percent over the $22 billion reported in October 2006 and 2 percent higher than the $22.6 billion reported in September of this year.

That 5.0% sales increase year/year was a slight decline from the 5.8% year/year growth reported a month ago but was still the second-best growth reported since January. Taken alone, I wouldn’t consider this report troubling in terms of the supply/demand balance as it shows stable if not slightly improving demand trends.

However, on November 15 Semiconductor Equipment and Materials International (SEMI) released their October book/bill report for semiconductor equipment orders and sales. The bookings figure was flat with the final September 2007 level of $1.24 billion and 16 percent less than the $1.47 billion in orders posted in October 2006.

The fact that demand growth (up 5%) was greater than supply growth (down 16%) is generally supportive of positive performance for semiconductor stocks. The performance of the SOX index during the last five periods in which such conditions prevailed is presented below.

Sources: SIA, SEMI, William A. Trent

In each of the periods other than 2001/2002 the excess demand growth relative to supply resulted in positive returns for the SOX. Unfortunately, the current period to date most resembles 2001/2002. That similarity is also noticeable when looking at the size of the relative peaks and troughs in supply/demand balance.

Sources: SIA, SEMI, William A. Trent

As the chart illustrates, there is typically a fair degree of symmetry between subsequent peaks and troughs, which is only natural because over time one would expect balanced supply and demand.

That 2001 peak in excess demand, however, occurred quickly and was shallow relative to the long, drawn-out period of excess supply that preceded it. So far in 2007, the chart is looking very similar. If the relationship continues to hold, it could be some time before semiconductor stocks again experience the normal cyclical upswing.

There are still some reasons for cautious optimism. For one thing, the semiconductor industry data are sometimes subject to large revisions. With that in mind, I’m not going to get too hung up on the data released in a given month.

Furthermore, recent forecasts have continued to show a cautious approach to adding capacity. Gartner Dataquest forecast that 2008 capital investments by the four largest foundries will decline 9.6% year-on-year. The largest foundries are Taiwan Semiconductor (TSM), United Microelectronics (UMC), Chartered Semiconductor (CHRT) and Semiconductor Manufacturing International (SMI).

Also, according to a Friedman Billings Ramsey analyst, capital spending in the DRAM sector is expected to fall by more than 30 percent in 2008. Leading players in this market include Samsung (SSNLF), Qimonda (QI), Hynix (HXSCF), and Micron (MU - Annual Report).

Together foundries and DRAM have been responsible for a good deal of the total capex and their caution ahead increases the chances of supply and demand returning to balance.

However, given the current state of the economy and the seasonal factors that should have helped demand in October and November, I’m glad I have my long position in the Semiconductor HOLDRS (SMH) offset by a put option on equipment maker LAM Research (LRCX).

In general, I favor the semiconductor makers like Intel (INTC - Annual Report) over the equipment makers like Applied Materials (AMAT - Annual Report) or KLA-Tencor, due to the fact that more capex cuts will be needed to restore the supply/demand balance.

Disclosure: William Trent is long SMH and holds put options against LAM Research (LRCX)

Note: Be sure to use real time stock quotes and pick the right stocks for your portfolio.

Disclosure: William Trent has a long position in SMH.

Lest you think we were going soft, we hereby balance our earlier enthusiasm for semi stocks with our more customary caution. The five reasons to avoid semiconductor stocks right now include:

1. The fundamentals will get worse before they get better. While supply indications grew slower than demand in April, the turn followed 16 months of too much capacity being ordered. As that capacity comes on line, the inventory situation will worsen and margins will get hit more. It is not at all certain that estimates reflect this.
2. It is May. Sure, sell in May and go away is a cliche. Things often become cliches for a reason.
3. Demand? What demand?
4. Valuations are too high because investors are hoping for more premium buyouts. They will happen, but not to every name in the sector.
5. The last bear may no longer be standing.

Food for thought.

Disclosure: William Trent has a long position in SMH.

A reader complained yesterday that we have been too negative. While we aren’t going to go crazy and have a whole positivity day, we will take the time to outline the bull case for the industry on which we have been most negative: semiconductors.

1. The bad news is known. When we started harping about oversupply, it was the farthest thing from anyone’s mind. Like Heisenberg’s uncertainty principle, the act of observation can alter the experiment.
2. The market is ignoring the fundamentals. Related to point 1, the market knows about the bad fundamentals and doesn’t care. Often this means that the bad news is sufficiently well known to be priced in. This is of course the weakest reason, as the market ignored the fundamentals in 2000 as well.
3. Demand may be ready to pick up. Double-digit growth from a tech distributor for the first time in a long time should not be ignored. The Vista hoopla has passed, now the nuts and bolts work may be beginning.
4. Supply and demand will soon realign. For the first time since 2005, orders for new equipment grew at a slower rate than semiconductor end demand. The longer this situation continues, the healthier it will be for future industry sales, pricing and profit margins.
5. The game has changed. Forget private equity buyers. For the first time a semiconductor management team decided it was more important to take capital out of the industry than to add more. This is a sea change in semiconductor management-think, and the strong positive reaction from investors ensures that the wave will continue to build.

There. That wasn’t so hard, was it? Stay tuned for our five reasons NOT to buy semiconductor stocks today.

Disclosure: William Trent has a long position in SMH.

Despite generally strong orders for semiconductor equipment, we have maintained for some time that overcapacity building by the semiconductor manufacturers would ultimately result in order push-outs or cancellations. The order push-outs are one reason why some equipment makers never seem to be able to spot the downturn, even though experience should tell them otherwise.

Case in point: Texas Instruments (TXN - Annual Report). Even though bulls frequently argue that the long lead times necessary to build a semiconductor fab dictate the order cycle Texas Instruments offers a case study of order push-outs in action.
TI’s RFab not for sale - Semiconductor Fabtech

There has been much speculation over the future of Texas Instruments’ shuttered new 300mm fab in Richardson, Texas - dubbed RFab - since the chip manufacturer announced that it would stop internal manufacturing of digital CMOS devices in favour of foundry agreements.According to a story by Mark LePedus at EETimes, Kevin Ritchie, senior vice president of TI’s Technology and Manufacturing Group, said in an interview that KFab was not for sale, though the timing for first phase tool installation had been put back ‘a year-and-a-half.’

A brand new multi-billion dollar facility that just isn’t going to operate - for now. Perhaps TI will eventually use it, or perhaps they will sell it. In the meantime, however, don’t count those semiconductor equipment bookings until they are billed.

Disclosure: William Trent has a long position in SMH.

We have been talking about oversupply of semiconductors for some time. Now, in the last two weeks two research firms have cut forecasts according to Semiconductor Fabtech:

IC Insights has drastically cut its semiconductor growth projections for 2007 citing severe pricing pressures in the NAND flash memory market as well as the continued decline in microprocessor prices. However, the market research firm has added that a major DRAM price collapse has also started and will also affect market growth this year. As a result, the firm has lowered its forecast to 2 percent growth compared to its previous forecast of 7 percent growth for 2007.Only two weeks ago, Semico Research lowered its semiconductor forecast for the second time this year citing poor prices even though unit demand remained strong, and now projects only 1.8 percent growth for 2007. In March, Semico had projected growth of 5.8 percent compared to a projection at the beginning of the year that the semiconductor industry would grow by 7 percent.

While everyone plays catchup we will refer you back to our October 2006 comments, which have not needed any revisions:

The chart below shows the year/year growth rate in semiconductor sales for each month going back to 1998. Data is courtesy of the Semiconductor Industry Association (SIA).

This tells us a few things:

• The semiconductor sales growth rate has been more consistent since 2005 (in the mid-high single digits.)
• The forecast calls for that to essentially continue for two more years.
• The chart tells us that is pretty dadgum unlikely.

Sales growth is likely to be either much higher or much lower than 8.6% next year. The million dollar question (or however much you may have at stake - for us it’s more like a couple thousand) is which direction. We’re betting it is lower.

For one thing, this is the longest the semi industry has ever gone without a year/year decline. By itself that doesn’t mean much - due to the fabless/foundry model and general tech industry maturity sales growth should be less volatile.

However, when you combine low volatility in sales growth with huge orders for new manufacturing equipment you end up with oversupply. Oversupply in a cyclical industry means sharper than normal price reductions. If the prices fall faster than unit demand rises - you get a decline in sales.

The other reason we expect semiconductor sales to be less than the industry predicts next year is summed up in the following chart, which tracks the total sales (rather than growth) over the preceding twelve months. There has been a definite change in the overall growth rate, going back to about 1995 or 1996, depending on where you want to draw the lines. For simplicity, we’ll just use the last 10 years (September 1996 - August 2006). Over that time frame, the average growth rate has been 6.3%. However, it is easy to see that that rate is toward the top of the new range (the trendline represents resistance in this case.)

So, given overcapacity, volatility and resistance we think 8.6% growth in 2007 is on the optimistic side. Perhaps even the wildly optimistic side.

We’ll have to wait and see how much of a shakeup there is in 2007 before we’d be willing to comment on 2008.

Disclosure: William Trent has a long position in SMH.

When Semiconductor Equipment and Materials International (SEMI), the industry trade organization for semiconductor equipment makers, reported the January book to bill ratio for chip equipment, we were concerned.

Unfortunately, just when it looked as if things might be set to turn the semiconductor companies re-accelerated their pace of equipment orders over the last two months.

Until orders for semiconductor equipment start growing at less than the roughly 10% growth in demand for semis, there will continue to be the brutal pricing environment we have seen recently. The decent guidance and calling of bottoms are pipe dreams.

But now it appears there could be another reason for the datapoint we found so strange. According to Reuters.com:

Global sales of microchip-making equipment in February rose 16.6 percent from a year earlier on demand for tools to make and process silicon wafers, an industry group in Japan said on Friday.Sales of gear used to make semiconductors rose to $2.72 billion in February, the Semiconductor Equipment Association of Japan (SEAJ) said.

The group also restated sales figures for January, saying sales rose 17.2 percent year-on-year to $3.49 billion, instead of a previously stated rise of 34.5 percent to $4.01 billion.

A member of Semiconductor Equipment and Materials International, a California-based industry group, had given the wrong sales numbers, the SEAJ said in a statement.

Then SEMI reported adjusted figures. Here is what the numbers looked like in February as originally reported:

And here is the new and improved data:

The March numbers will be reported Thursday. Given that the supply/demand imbalance is the main contributor to our bearishness, a continuation in this trend would mean we are much closer to adopting a neutral/bullish outlook for semis. We can’t wait for Thursday’s release to find out.

Disclosure: William Trent has a long position in SMH.

Having given you our take on the semiconductor sales data yesterday, We had to laugh when we saw this article from Tech Trader Daily - Semis: Feb Sales Data Look Weak; Are We Nearing A Bottom?

The Street this morning agrees on the obvious conclusion of the latest SIA data: it was weak. “Weak February results,” “shipments weaker than typical,” “weakness in both units and pricing,” “data appears generally weak,” “weaker than normal seasonality,” said analysts this morning at J.P. Morgan, Wedbush Morgan, Robert W. Baird, Lehman Brothers and UBS, respectively.But there is also a growing consensus that we are at or near a bottom in semi fundamentals, and that a recovery is just around the corner; it is a theory tied to the notion that the recent inventory correction in the sector is nearly completed.

Oh, really? Here’s a chart we’d like the analysts Eric quoted to look at, which is a total semiconductor sales on a trailing 12-month basis since 1995, based on data from the Semiconductor Industry Association:

So, Christopher Danely, analyst at J.P. Morgan, Craig Berger, of Wedbush Morgan, Lehman’s Tim Luke, and Uche Orji, of UBS: Can you tell me why this looks like a bottom to you?

Disclosure: William Trent has a long position in SMH.

The Semiconductor Industry Association released global semiconductor sales data for the month of February, 2007, and it did not look good.

Worldwide sales of semiconductors of $20.09 billion in February were 6.5 percent lower than January when sales were $21.48 billion, the Semiconductor Industry Association (SIA) reported today. February sales increased by 4.2 percent from the $19.28 billion recorded in February 2006.“While seasonality clearly contributed to the 6.5 percent decline in worldwide chip sales month-on-month, declining unit shipments and lower average selling prices (ASPs) in several key market segments were a factor,” said SIA President George Scalise. “Both unit shipments and total sales of microprocessors and DSP chips experienced sequential declines in February. Unit shipments of NAND flash increased sequentially while total sales saw a double-digit decline, indicating very competitive market conditions.”

It reminds us of something we said in February:  Although end demand for semiconductors has experienced fairly steady growth, excess supply has sometimes caused inventory to grow to unsustainable levels. For example, “by 2000 most of the capacity was in place, churning out chips. Since end demand was growing at a slower pace, inventory built up. The falloff in 2001 wasn’t so much a drop in demand, but the fact that the demand could be filled from that existing inventory.” We even backed it up with this chart:

We followed up last week, saying “Next we turn to capacity utilization, which has clearly started to fall. Excess capacity means the potential for even more inventory to be produced, which tends to put downward pressure on prices. Since the majority of semiconductor manufacturing costs are fixed, low utilization means lower profits - either because prices have to be reduced or because the per-unit costs are higher when production is cut.” It sounds very similar to what Scalise described in today’s SIA press release:

“Year-on-year, we see evidence of the fiercely competitive market conditions – across the board unit sales in key products increased, while ASPs declined. Unit sales of microprocessors were up almost 8 percent while ASPs declined 15 percent, and NAND flash units grew by over 40 percent while experiencing a nearly 50 percent drop in ASPs. These products tend to be indicators of conditions in important end markets such as personal computers and consumer devices,” Scalise continued. “Personal computers and consumer products now account for approximately 60 percent of semiconductor sales. Both competitive conditions and product mix issues appear to be affecting revenues of these key components.”

SIA noted that overall capacity utilization declined from 88.9 percent in the third quarter of 2006 to 86.8 percent in the fourth quarter. Most of the decline was in foundry utilization, which fell from 91.5 percent in the third quarter to 80.9 percent in the fourth quarter. The reduction in capacity utilization in the fourth quarter addressed inventory builds in the semiconductor supply chain and selected end markets which are expected to show growth consistent with GDP performance in key world markets in the coming months.

The problem for us is, we don’t see the reduction in capacity utilization as having “addressed inventory builds.” Rather, we see it as the logical conclusion of excess capacity expansion that has built up now for more than a year:

At the time, we said:

The good thing about the downward revision, and also the decline in February, is that it restores some balance to at least the trend in equipment orders relative to end demand for semiconductors. Although supply (chip equipment orders) is still growing much faster than the roughly 10% growth in semiconductor demand, at least the rate at which the capacity is growing is starting to slow down again. Furthermore, the billings (which represent what is actually installed rather than orders, which may prove too optimistic) have been running at a slower rate than orders. The 22% growth of installed equipment is still well higher than what is needed, but has less far to fall.

Unfortunately, today’s release also blows a hole in any hopes we had that slowing orders would restore balance. Since we measure the imbalance as the differential between dollar capacity growth and dollar sales growth, the slowdown in semi demand to 4% year/year from “roughly 10%” means that February’s decline in chip orders did not improve the potential supply imbalance. Things have farther to fall than we thought.

Disclosure: William Trent has a long position in SMH.

More capacity.

UMC breaks ground on $5 billion 300mm fab - Semiconductor Fabtech

UMC has started construction of its third 300mm fab at an estimated cost of $5 billion US Dollars, the company has said in a press statement. The new fab will be located in Taiwan’s Tainan Science Park, adjacent to its existing 300mm facility, Fab 12A. UMC has another 300mm facility in Singapore. The new fab is expected to have a capacity of 50,000 wafer starts per month and is expected to be completed by the end of 2007. First phase tool install is expected in the first quarter of 2008.

There is plenty of that already.