The Semiconductor Industry Association (SIA) published its latest semi-annual forecast for global chip sales on Nov. 7. It projects chip sales rising just 6% in 2002, then about 21% in both 2003 and 2004. It will be 2004 before the level of sales in 2000 is surpassed, by this reckoning. The present pace is lackluster for an industry that grew at a 17% average rate over the past 40 years.
But a little perspective is in order here. The industry has long been subject to a typical three- to four-year year business cycle, so a long pause after the boom in 2000 may be considered normal, even if it is breathtaking in its scope. If the pattern of cycles continues, there may be a big year sometime out in the 2004-6 horizon.
The SIA's November 2001 forecast for total semiconductor worldwide sales (billion US$) in 2000-2004, and percentage change from the preceding year:
SURVEYING THE DAMAGE. In the summer of 2001, the semiconductor industry was about as far down as it could get based on sales and order patterns. Backlog was used up, so order cancellations could not get worse. End-markets for most electronic goods including PCs, cell phones, and especially wireline telecommunications equipment were all sour. The abysmal fundamentals created some hope for investors on the theory that if things can't get worse, they must get better ... eventually. Something of a rally in chip stocks evolved after the July earnings reports.
However, economic forecasts were ratcheted lower after the September 11 terror attacks. Note that for many chip companies, the September troubles made for simply a few lost days of operations, rather than creating lasting damage. But for companies hoping for a mild comeback based on strong back-to-school sales and Christmas sales, the slide in consumer sentiment did disappoint. The weaker consumer sector may have delayed recovery by a quarter or so for some chipmakers.
Wafer fabrication plant utilization rates fell to a level resembling a cycle low in the summer. The SIA numbers show industrywide plant utilization was 72.7% in the second quarter of 2001, compared to a cycle-top level of 96.4% in the third quarter of 2000. The rate likely fell further in the third quarter of 2001, given a large number of planned plant shutdowns as part of cost-saving campaigns.
IN THE MIX. Utilization rates at the big chip foundries in Asia were pushed even lower than the industry average noted by the SIA, as they represent the marginal capacity for the industry. The data tell a tale about the strength of different end-markets for chips.
The third quarter capacity utilization rate fell to 41% at Taiwan Semiconductor, which has a mix of customers including chipmakers serving the PC (49% of sales) and consumer (25%; think X-Box, DVD players, digital cameras, etc.) markets. Another Taiwan-based foundry, United Microelectronics Corp., saw capacity utilization drop to 36%. United Micro's mix includes 37% computer and 25% consumer.
On the other hand, Singapore-based Chartered Semiconductor had a third quarter capacity utilization rate of 22%, much lower than at Taiwan Semi and UMC. Why? The product mix included only 32% computer and 12% consumer. Chartered's greater exposure to communications customers (45% of September quarter) meant slower business because communications markets, especially wireline telecom equipment, are some of the weakest end markets at present.
BEST BETS. According to the notion that wireline communications is a sector that may take a year or longer to recuperate, whereas automotive, aerospace, military, and industrial markets for chips are relatively steady, and markets for cell phones, PCs, video games, and home broadband may rebound by late 2002, we at Standard & Poor's have rated many chipmakers aiming mainly at communications markets as 2 STARS (avoid) and 3 STARS (hold). Among the avoids are Applied Micro Circuits (AMCC
), Cypress Semiconductor (CY
), Integrated Device Technology (IDTI
), PMC-Sierra (PMCS
), and Vitesse (VTSS
The sole 1 STAR (sell) recommendation at present is Micron Technology (MU
), a reflection of a truly dismal outlook for memory chip pricing. Micron is the low cost producer, but even they are losing money. There is too much global capacity for DRAM. Competitors are fleeing the business but not rapidly enough to support prices. If loss-plagued Korean memory chipmaker Hynix would ever quit the business, memory pricing would improve, but so far it looks like Hynix has become too big to fail -- witness the recent bailout of the company by the Korean government.
Companies that we like are aiming at markets that are broader and may come back relatively early in the cycle. These include the high-end analog chipmakers Linear Technology (LLTC
), rated 5 STARS, or buy, and Maxim Integrated Products (MXIM
), rated 4 STARS, accumulate; the DSP and analog players Texas Instruments (TXN
), 5 STARS, and Analog Devices (ADI
), 4 STARS; graphics chipmaker NVIDIA (NVDA
), 4 STARS; and the pure play in microcontrollers, Microchip Technology (MCHP
), 4 STARS.
The stock market is quite lively at present and, of course, our recommendations are subject to change as valuations fluctuate. As I have tried to indicate, a turnaround in the chip sector is underway, but the comeback will be a long process marked by different recovery periods for different chip categories. The best of the bunch: the analog and DSP categories. Smith is director of technology research for Standard & Poor's