Chips and Dips
Thus far, it is a typically volatile "triple-witch" option-expiration week. The NASDAQ100 traded in a 3% intra-day range Monday and 4% yesterday and today. At the same time, the sellers are generally outgunning the buyers as the NASDAQ100 has dropped 2% thus far this week. The selling remains most intense in the "big cap" technology area, with the Morgan Stanley High Tech and Semiconductor indices dropping 3%. Recent losses have reduced year-to-date gains for these two key indices to 11% and 43%. Selling remains tepid elsewhere, with the small cap Russell 2000 and the S&P400 Mid-cap indices each declining less than 1% so far this week. The Street.com Internet index has declined about 1% and the NASDAQ Telecommunications index has dropped 2%. The Dow is largely unchanged while the S&P500 has declined about 1%. The Morgan Stanley Cyclical index has dropped 1%, while the Transports have posted a slight decline. The Morgan Stanley Consumer index has added 1% and the highflying Utilities have jumped another 3%, increasing year-to-date gain to an astounding 36%. The AMEX Security Broker/Dealer index has increased slightly as its year-2000 gain sits at 55%. With Chase Manhattan sinking on the news of its purchase of JP Morgan, the S&P Bank index has posted a 1% decline. The gold stocks have had a disappointing week, declining 4% thus far.
It has been a positive week thus far for the credit markets. Treasury yields have generally dropped about 2 basis points, although 30-year bond yields have been the exception rising 3 basis points. And with the view that there is now little possibility of GSE legislation this year, this week investors have not only been buying the stocks of Fannie Mae and Freddie Mac, but they have also purchased mortgage-backs and agency securities. Mortgage-back yields have generally declined about 6 basis points while agency securities have seen yields sink 5 basis points. Likewise, the 10-year dollar swap spread has declined 4 basis points to 125. The dollar rally continues, with the dollar index adding about 1% so far this week. Energy prices continue to be extraordinarily volatile, with crude oil posting a slight increase this week.
The semiconductor industry is in a unique position in the business world. The chip business is where Wall Street meets Main Street. While this notion may seem odd as semiconductors are among the most advanced products in the world, the chip companies depend on both "streets" to work together. Semiconductors make-up the "brains" behind all the electronic goods we purchase, from PCs to cell phones, DVD players and game consoles. If Main street is in a feel-good mood, chip companies are feeling really goodAt the same time, semiconductor technologies are continually evolving. This forces the industry to constantly upgrade and expand fabrication plants. These "fab" investments are expensive, so semiconductor companies have an insatiable need for capital. These two forces working together (or against each other) make the semiconductor industry extremely dynamic. The recent environment of extreme consumer spending coupled with easy financing could not have been better for the semiconductor industry. Recent euphoria, however, has historically proven to be peak in the business.
The last 20 years of the semiconductor business has been rocked with tremendous booms that turned chip companies into proverbial printing presses for cash. However, these booms always have a downside. These downsides have been just as eventful with companies going out of business and industry experts wondering if the industry would ever recover. Some history of previous boom-bust cycles is necessary in order to fully grasp not only how cyclical the business is, but also how erratic. We will start with a boom that might still be memorable. In 1979 the Atari game machine captured the attention of the American public. As sales quickly grew, semiconductor shipments broke though 10 million units for the first time, 25% above 1978. A capital spending boom followed. As game machines became more of a fad than a trend in consumer electronics, the semiconductor market stalled in 1981. Just as the Atari was loosing steam, however, the Personal Computer came on the market.
The 1982 to 1984 boom was one of the biggest, if not the biggest the industry had ever seen. Chip shipments almost doubled from the 1981 low to 26.1 million. Capital spending followed suit, climbing to an unprecedented $8 billion in 1984. It was during this building boom that the Japanese semiconductor industry began its drive to dominate the industry. In the 1980s, of course, Japan had a soaring stock market and a very favorable lending environment. In 1984, the Japanese semiconductor industry spent approximately $3 billion on expansion. This compared to the $2.3 billion invested by the American semiconductor industry. The major difference between the two country’s chip industries is how the companies are structured. Throughout Japan, the chip companies were part of large conglomerates. This was in direct contrast to the American semiconductor companies, which were much smaller "merchant" firms. These "merchant" firms tended to produce a very limited line of chips, choosing to be more research intensive and innovative. The Japanese companies cared more about market share and supplying chips to their other divisions, namely consumer electronics. As such, the Japanese firms were able to subsidize the semiconductor business with other businesses and basically created an "arms race." All the companies were forced to expand, or risk losing business. Not surprisingly, as capacity increased the price for semiconductors fell. Needless to say, it is very difficult to run a profitable business when the underlying product is commodity-like in nature, and your competitors have little concern for profits and are, at the same time, large enough to maintain an unprofitable business strategy almost indefinitely. The end result of this "arms race" was the Japanese increasing market share of the DRAM market from 42 percent in 1979 to 98 percent in 1985, while creating enough excess capacity to last almost a decade. In fact, US chip shipments did not return to peak levels until 1988. By this time, Japan had not only surpassed its 1985 peak, but also had more than doubled it. Japan surpassed its 1985 peak two years later, and only suffered a 5 percent decline in shipments during 1986 compared to the 30% decline experienced by the US semiconductor industry.
During the next year the industry consolidated. However, it was not easy for the sellers to find buyers. Several semiconductor companies were almost given away so companies would not have to deal with the aftermath. To illuminate how bad a downturn can be, John Cornell, the head of Harris Corp.’s semiconductor unit stated: "We used to depend on good times to more than make up for the bad times. Now, though, I don’t think there’ll ever be another great year, with sold-out production and wonderful profits. Pricing will be tough all the time." The ensuing consolidation revealed the extent to which companies wanted out of the business. AT&T purchased Honeywell’s chip subsidiary, Synertek, for less than half of book value. Thomson-CSF paid $70 million for Mostek, a division of United Technologies. United Technologies had Mosket on the books at $500 million. The managing director of Soloman Brothers, who managed the United Tech. deal summed up the market "these deals are a bitch." He went on to shed light on the mindset of the US semiconductor industry at the time, "No U.S. company wanted Mostek, so we had to go the Europe. There really isn't much of a market." Thomas Kurlak fretted over Texas Instruments’ DRAM business saying "how long will TI keep banging its head against the wall in D-RAMs and wasting stockholders’ money?" This question was answered after the next bust when Texas Instruments sold its DRAM business to Micron Technologies in 1998.
The next significant semiconductor boom started in 1992, with PC sales driven by Windows 3.1, Intel's 486 chip and a recovering US economy. From 1991 to 1994 the number of PCs sold boomed from 28.7 million to 50.8 million. In 1995, PaineWebber forecasted that sales growth would continue in the low to mid teens, reaching 87.4 million units in 1998. These forecast ended up being conservative as 1998 units shipped topped 92 million. The PC boom, of course fueled tremendous chip demand. From 1992 to 1995 semiconductor shipments increased over 140%, reaching 144 million. These trends were quickly extrapolated. During 1995 Dataquest had expected the industry to double by the end of the decade to 234 million chips shipped, and the SIA was forecasting a slightly less 233 million. The most aggressive forecast came from Robertson, Stephens & Co., which predicted shipments to hit 300 million by 2000. Fast forward to now and we can see that total chip sales barely increased to $149mm in 1999, representing 3.5% growth from 1995's peak. Back then, Wall Street was salivating over the chip companies. After Micron Technologies share price doubled in 1994, its' 1995 gain was even more spectacular. Micron Technologies' share price gained more than 300% at its' peak in September of that year. Wall Street was enamored with the high margins and expanding capacity and extrapolated 1995 into the future. During September and October of 1995, Wall Street was aggressively raising earnings estimates. CS First Boston raised 1996 estimates to $6.27 from $5.07, while keeping 1997 estimates at $7.29. Bear Sterns was even more bullish as it raised 1996 estimates to $6.50 from $5.00, and introduced its' 1997 estimate at $8.50. All attributed continued surging demand for DRAMs and stable pricing. There was some concern about pricing, but gross margins were still modeled for 52%-53% for 1996 and around 50% for 1997. Jumping ahead so not to keep you waiting, Micron reported EPS of $1.38 and $0.74 for 1996 and 1997 respectively.
But if PC demand was stronger than forecasted, how did the semiconductor industry go into recession after 1995? It goes back again to what sank the industry ten years earlier: overcapacity. Based on lofty forecasts calling for the chip industry to double by the end of the decade, capital spending went through the roof, jumping from $12 billion in 1992 to $43 billion in 1996. The capital markets had no problem supplying the additional money for investment. In 1992 the Federal Reserve aggressively lowered short-term interest rates, which created a very steep yield curve. This sparked the speculators to borrow cheap US dollars that were then invested everywhere. Much of this "hot money" was directed to the emerging markets, namely the Asian Tigers. According to IMF data, capital flows in 1993 to emerging markets climbed 36%. The growth in Asian capital flows was even more impressive, climbing 165% to $57.6 billion. This accounted for 35% of all private capital investment into emerging markets. After the Mexican bailout in 1995, the "hot money" quickly determined that the "rules of the game" had gotten even better. They were now free to invest in the emerging markets with the protection of the US Government. In bailing out Mexico, the US Fed created a "moral hazard" where suddenly investors could reap rewards for investing in the emerging markets knowing that there would be downside protection. This is perhaps the first issuance of the "Greenspan Put." Capital flows to Asia exploded. Private capital flows into Asia soared in 1995 to $95.8 billion, 48% higher than the previous year and accounted for half of all private capital invested in emerging markets. Capital continued to flood into Asia through 1996, reaching $110.4 billion, or about 7% of GDP. Total private investments to Asia between 1993 and 1996 were an astonishing $330 billion.
Highlighting the ease of securing financing during the Taiwanese boom, Chris Kao of Macronix International described his financing plan: "We will borrow half from banks, one quarter will come from profits, and the remainder from the sale of additional stock." Pretty simple. The Director of the Hsinchu Science-Based Industrial Park commented that "I’ve never heard of a CEO complain that he couldn’t get financing for a fab." In fact the financing was not the most difficult aspect of building a fab. Instead, it was trying to buy the land to put it on. Reportedly, it took up to five years to negotiate the purchase of plots, yet everyone wanted to get into the chip business.
During the boom chip companies basically have a license to print money. The semiconductor industry made $2 billion a month during the 1993-1995 boom. These profits were driven from the higher prices the chip companies were able to demand when supplies ran tight. In 1994, Taiwan Semiconductor Manufacturing Company (TSMC) reported a 45 percent profit margin. Such high profits seduced companies in other industries to try their hand at chip manufacturing. Nan Ya Plastics entered into an agreement with Oki Electric Industry to build a $700 million fab in 1995. Additionally, the governments of the Asian Tigers desperately wanted to get into the game. These governments saw the semiconductor industry as something that could drive their economy forward creating profitable exports and high-tech jobs. Singapore’s government helped establish the chip industry with a $710 million war chest. The government took a 26% equity stake in a TI venture and a 30% share of a new Hitachi plant. The government also helped train the workforce by paying 50% of chipmakers’ training cost. The state-run universities developed crash programs to educate engineers and researchers.
With all this increased capacity, the semiconductor market hit a brick wall when the Asian crisis developed.
As worldwide chip demand declined, chip capacity reached a new record. Since the fixed cost is so high for fabs, it is most economical to run production as high as possible. This in turn causes production to be higher than otherwise economically prudent, which leaves the price of semiconductors vulnerable to collapse. This creates a vicious cycle. During 1996 the price of 16Mb DRAM went into free fall, declining more than 80%. This drop was particularly problematic for Micron. Micron's margins fell to 40% in 1996, 28% in 1997, and bottomed out at 9% in 1998 as Micron posted a $230 million loss. As investors started to recognize the problems Micron's stock price plunged 57% to close out 1995 just below $20. In the first half of 1996 the stock continued downward until dropping below $9, 81% below the 1995 peak.
The following bust was almost purely caused by overcapacity. Semiconductor shipments only declined 5% off the peak. As the bust played out, it quickly reverted back to a boom due to the booming US economy and consumers' insatiable appetite for electronics. This quick turnaround is exemplified by looking back at the planned capital expenditures of Taiwan Semiconductor Manufacturing Corporaton. (TSMC is the largest foundry. Foundries contract out their fabs for chip companies to use.) In 1998, due to slowing revenue TSMC reduced its' capital spending plans for 1998 and 1999. Spending plans for 1998 were reduced to $920 million from $1.3 billion, and for 1999 planned spending was reduced to the $800-$900 million range from $1 - $1.1 billion range. A year later in 1999, spending was revised upward to $1.13 billion for the year. This not only was more than last year's revised spending, but more than original estimates before being revised down. Spending for 1998 only ended up being $830, almost $100 million below the revised forecast. During the fourth quarter of 1999 TSMC raised its' capital spending plan to $2.5 billion. This estimate was again revised in June of 2000 to $4.7 billion. Commenting on TSMC's expansion plans, Chairman Morris Chang attempted to dismiss the boom-bust cycles that the semiconductor industry is known for:
"(It's) The reason we adopt this policy, you know the semiconductor industry; when the business is good, everyone expands like mad, when the business in bad, everyone retrenches. I want to stay away from that. Being human, even TSMC's behavior is modified somewhat. But I have to say that essentially, we have followed a linear expansion policy. In 1998 we modified that to 20 percent but in 1999 we quickly recovered from our error. We not only resurrected our expansion rate to 25 percent, we've increased to 30 to 40 percent for the next two or three years. Right now, we have kind of a broken curve. That's what we are following right now. This year expansion (capital expenditures) will be $4 billion [before recent increase] and next year it will be $3.6 billion."
It is easy to see that the boom is back. The Semiconductor Industry Association current chip forecast calls for the industry to double over the next four years to 312 million semiconductors shipped. This is more than double last year's shipments of 149.4 million. As TSMC details above, chip makers are again increasing capacity. The Financial Times ran a story in July headlining "Chip industry in $60bn race to meet demand." Within the article it mentioned Robertson Stephens estimate of $80 billion being spent on equipment alone in 2001. Some recent headlines that highlight individual company's capital spending plans shed light on the current spending frenzy:
- -On September 4th, Sony announced plans to build its first 300-mm plant at an estimated $943 million. Sony will increase spending this year by 16 percent to $660 to begin construction. Sony specifically mentioned Digital cameras and camcorders as key drivers of chip demand, along with LCDs used for data projectors and rear-projection TVs.
- In May, Toshiba said it will increase capital spending by $368 million to a record $1.56 billion, citing DVD players and mobile phones driving demand. Additionally, it will expand the production of chips for Sony's PlayStation.
- Last April, Intel increased capital spending for 2000 from $5 billion to $6 billion.
- On Monday, Micron announced its plans for a $200 million expansion at its Boise plant.
- Samsung Electronics increased its original $2.6 billion capital investment plan over 50% to $4.1 billion.
- UMC is now planning for $2.5 billion in capital spending this year, up 47 percent from $1.7 billion in 1999.
- Chartered is planning to increase its capital spending by 102 percent to $1.4 billion this year, from $693 million spent in 1999. The foundry expects to spend $1.68 billion in 2001.
- Samsung will invest a total of $6.88 billion in the business in the year. Starting next year it will invest $5.16-billion in new facilities up until 2003. The company forecasts chip-related sales for this year will reach $19.18-billion and $44.26-billion by 2005.
These spending plans are a function of the forecast made by the SIA and others regarding chip demand. The current driver of chip demand is the increase of cell phones and internet devices. STMirco’s Jean-Philippe Dauvin forecasts 1.8 billion cell phone subscribers worldwide in 2005 up from 510 million now. Additionally, he sees 260 million households connected to the Internet by 2005 from 60 million currently. To meet this predicted growth it seems virtually everyone will have a cell phones. When assessing these forecasts, keep these statistics in mind. If current demographic continue, 31% the world’s forecasted 6.4 billion people will be younger than 15, and 21% of adults will be illiterate. These two facts reduce by half the world's cellular phone buying population. Also, about half of the world’s population lives on less than $60 a month. Plus there are at least 500 million homes without electricity and two-thirds of the world population has never used a phone. But lets give the benefit of the doubt to Wall Street for a minute. (I promise not any longer!) U.S. wireless companies have spent approximately $71 billion over 15 years to secure about 90 million subscribers. This works out to be around $800 per subscriber. If there were to be 1.8 billion subscribers (currently there are about 500 million) it would require roughly another $1 trillion, yes with a "T", of capital expenditures.
All the same, once again Micron Technologies is trading north of six times book value. Just a couple weeks ago it was close to nine times book. During the previous boom, Micron traded at almost 8 times book during the peak in September 1995. Moreover, analysts are also raising estimates again. Today, PaineWebber increased next year's estimate to $4.92, which is in-line with I/B/E/S mean estimate of 4.97. Analysts forecast earnings to grow an additional 18% during the year ending 2002, with the mean estimate of $5.86.
The semiconductor market depends on the consumer electronics market and the capital markets for survival. The boom-bust cycle is more pronounced for the chip business. Not only because it supplies the economy with the primary parts for goods, but its' capital projects are so large it is only during economic expansions that they are able to secure large amounts of financing. The last twenty years have been extraordinary for both consumer electronics and the capital markets. Especially during the past five years the American consumer has been more flush with cash than ever before. This has not only driven demand for electronic goods, but has supplied a tremendous amount of liquidity to the capital markets. Additionally, the last twenty years have been riddled with distortions to the capital markets that facilitated massive amounts of money to flow into the semiconductor industry. Unfortunately, these euphoric expectations will prove too high and the money spigot will turned off and demand will collapse. These huge investments will then prove uneconomical and for years serve as powerful reminders what can happen when economies and capital markets become disconnected.