During the “bubble”, PE multiples expanded on top of inflated earnings expectations, with sell-side analysts declaring the “end of the cycle”.
Years later, investors have become somewhat bored with the sector as evidenced by the relatively low volatility apparent in this chart of the widely followed Philadelphia Stock Exchange Semiconductor Index [SOX]:
Actually, the volatility is quite high (range of over 50% over past several years, Beta of 2.3), but the index has been in a holding pattern for several years now. As one can see in the bottom panel, this group have underperformed the NASDAQ since early 02 and still remains below the lows of that year. The overall industry is clearly cyclical, though it continues to grow:
It is important to understand some of the key factors that impact the industry and affect the companies and their stocks. Perhaps most important is the supply dynamic that results in declining ASPs. It is a very tough business whereby if a producer can’t lower their costs of production, then they become uncompetitive. While falling average selling prices is quite pervasive, there are a few exceptions. This requires companies to potentially make significant capital expenditures. As a result, many companies have outsourced their production, focusing on R&D or marketing (known as “fabless”).
Another factor to consider is the markets that the companies serve. Analysts break down those markets differently, but one that I like is: Computing, Consumer, Networking, Industrial and Cell Phones. In recent years, Consumer applications have proliferated. Computing has been stagnant, though some (not me) expect Windows Vista to turn the tide. Cell Phones have been and should continue to be a strong end market. Industrial applications have been strong, driven by strengthening global economies (and emerging growth in Asia and Eastern Europe). Networking demand hasn’t yet met the high expectations regarding cellular networking upgrades. A final consideration worth mentioning is that many of these companies have stellar balance sheets, reasonable valuations and strong business models. Not surprisingly, there was a large LBO (FreeScale), with continuing mention. While this list isn’t exhaustive, it serves as a good starting point to differentiate the companies in the industry.
A Look at the SOX and SMH
The SOX, the most widely followed index, consists of 19 companies, four of which actually make equipment used by the industry (Applied Material (AMAT), KLA-Tencor Corporation (KLAC), Novellus Systems, Inc. (NVLS), and Teradyne, Inc. (TER)). In terms of composition, the index doesn’t reflect the economics of the industry at all since it consists of one share of each of its members. Thus, KLAC is the largest (almost 12%) despite the much larger market-capitalizations of Intel (INTC), Texas Instruments (TXN) and AMAT. Similarly, Sandisk (SNDK) is over-represented.
SMH, the ML Semiconductors HOLDRS, is the most actively traded by far. While there are some others (XSD, IGW, USD), they are very thinly traded. The composition of SMH differs greatly from SOX and from economic reality as well. TXN, INTC and AMAT, all between 10 and 20%) make up about ½ the security. While INTC is by far the largest company, TXN is represented slightly more. As one can see in the charts below, the overlap of SMH and SOX is quite high (15 of 19) and the two have moved similarly:
SMH has underperformed by about 20% since the end of 2000, most likely a result of the composition weighted towards larger companies (particularly AMAT and INTC). The big drop in early 2006, for instance, was directly related to the weakness in INTC.
For those not comfortable investing directly in companies, SMH is a reasonable proxy. While I don’t particularly care for the prospects for INTC (Vista ramp slow, computers are not “growthy” the bulk of the names in SMH look attractive. There are many attractive names outside of the members of SMH (I own SiRF Technology Holdings Inc. (SIRF), for example), but there are plenty within. Here is my quick-take on the 19 members of SMH (in descending percentage):
Disclosure: Long SIRF and TXN and ALTR.