The Double-Peak Nature Of The Semiconductor Cycle, Revisited Hans C. Mosesmann
Classic semiconductor cycles, driven by over- and under-investment of capital in fabrication facilities (a.k.a. fabs), have an interesting periodic characteristic: Upcycles show a double-peak revenue-growth dynamic—with revenues peaking every three to four years—and downcycles have double-valley patterns of revenue growth.
Classic Semiconductor Cycles Are Characterized By Double-Peak, Double-Valley Revenue Growth. The oversimplified reason why these cycles last three to four years is because it takes at least a year for semiconductor companies to realize that they need to add capacity when business is improving on a sustainable basis, and it takes two years to actually build the fab(s) to add this capacity. Invariably when multiple semiconductor players add capacity, an overcapacity situation results—and lasts some three to four years until the next upcycle begins.
Extreme Macroeconomic Conditions, However, Can Result In Nonclassic Cycles, With No Second Peak. Now, there are instances when macroeconomic conditions are so severe that classic semiconductor dynamics do not apply. Instead, we see a period of hypergrowth, followed by a period of hyperdecline—which in turn leads to a cycle “reset.”
The 1983-84 period of hypergrowth (characterized by 135% year-over-year revenue growth) was followed by a period of hyperdecline in 1985 (when revenue growth fell 17%), driven by the severe recessionary dynamics at the time. This basically led to a situation that disrupted the second peak of the expected double-peak pattern of the mid-1980s. And it led to a resetting of the double-peak dynamic between 1986 and 1988.
That’s What We Appear To Be Seeing Today. This nonclassic cyclical pattern that the industry experienced in the mid-1980s was very similar to what we’ve been seeing in 2001 and should continue to see in 2002—i.e., severe macroeconomic conditions (in this case, a recession combined with huge overspending by the telecommunications industry in 2000) coincident with a mid-upcycle trough. We suspect that scenario once again to lead to a
resetting of the double-peak dynamic.
If We’re Right, Investors Can Expect A New Double-Peak Upcycle To Begin Within TheYear. Our analysis suggests that year-over-year semiconductor revenue growth will likely prove to have bottomed during the third and fourth quarters of 2001, and we expect a recovery—forming the first peak of a new, classic, double-peak upcycle—to occur over the next year (i.e., by the fourth quarter of 2002).
Accordingly, Cyclical Semiconductor Plays Are The Stocks To Buy. There has been a high level of correlation between semiconductor stock-price performance and semiconductor revenue growth on a year-over-year basis over the past eight years or so (Figure 3). Given that correlation, we believe that investors have a great opportunity to buy semiconductor stocks now, near the current bottom.
Our Favorite Cycle Plays Are Intel… We believe that Intel Corp. (INTC—$34.08; rated Buy and moderate risk) is now in its best competitive position in years, as it has successfully pushed its P4 architecture into the mainstream and is far ahead of its competition in terms of clock speed. Our investment thesis remains that it’s best to buy this stock just as margins are bottoming and the competitive dynamic is shifting back into Intel’s favor. The company’s
aggressive plan to ramp production on 0.13-micron process technology and 300 mm (millimeter) wafer should significantly improve Intel’s cost structure throughout 2002. We believe that Intel’s focus on its core competencies as well as its drive to succeed in selected emerging markets should enable it to achieve continued-healthy growth for years to come.
…LSI Logic… We view LSI Logic (LSI—$16.49; rated Buy and high risk) as perhaps the best diversified cycle play, with exposure to almost every high-growth end-market without commodity memory exposure. The company’s broad proprietary product offerings and diverse end-market exposure are solid predictors for the semiconductor industry, given that custom chips (which is what the company specializes in) have the longest lead times. When true end-demand turns around, LSI tends to get the early phone call for orders, as it typically takes approximately two months to build a product. LSI stock led the semiconductor group out of the 1998 recession and was the S&P 500’s top performer over the following year. We expect the stock to perform quite attractively coming out of this downturn as well.
…And Micron Technology. We view leading memory manufacturer Micron Technology (MU—$31.75; rated Buy and high risk) as the ultimate cycle play. We believe that it will be one of the few players left standing when the DRAM industry is finished consolidating. We view the Japanese retrenchment from the DRAM market and Hynix’s financial troubles as healthy for the industry over the longer term as incremental capacity continues to come off-line. The trends of:
1) increasing memory per box (driven by XP, P4, and the price elasticity of demand),
2) fragmentation of the DRAM market (into segments such as SDRAM, DDR, DRDRAM, EDO, etc.), and
3) the capital required to build fabs (representing a high barrier to market entry) in order to keep pace on the process-technology front all should benefit Micron going forward. |
