It's all over the map for the Asia Pacific outlook
11/01/2001
Debra Vogler, Senior Technical Editor
Trying to discern a one-size-fits-all industry outlook for the Asia Pacific region is like trying to follow the logic in an M.C. Escher drawing. Just when you think you've found where the path is going, you tumble head first in the opposite direction.
The year 2000 was an up market, but 2001 is down. Foundry capacity was saturated in 2000, but now there's plenty — yet more foundries are being built. Japan, Korea, and Taiwan have been in the doldrums, while China is ramping up for what could be the mother of all growth spurts.
 Figure 1. DRAM price/bit (based on a three-month moving average). |
The topsy-turvy view is not confined to capacity, either. While the price of DRAMs has been falling like a lead balloon, foundries like Anam and Hynix have been buying old memory fabs. Even though the price of LCDs has been falling, many fabs in Japan are operating at full capacity to meet expected demand. So what can be made of all of this? I asked a group of industry analysts to comment on some of the issues facing the Asia Pacific region, including Japan and Korea.
The participants in this discussion were: G. Dan Hutcheson, president, VLSI Research Inc.; Philip Koh, Dorothy Lai, and Hee-chan Lee, senior analysts, Gartner Dataquest; George C.T. Lin, VP, Asia Pacific operations at Semi; Brian Matas, VP of market research, IC Insights Inc.; and Tia-Min Pang, managing director, institutional research, S.G. Cowen.
LCDs and DRAMs
With prices of LCDs decreasing, thereby making it virtually impossible to profit on these commodities, why are Japanese fabs operating at full capacity to make them and why has Japan continued to push for more LCD fab capacity? It appears that this sector is not following the current drop in demand.
 Figure 2. The Asia Pacific total semiconductor market. |
"LCDs come in many flavors," stated S.G. Cowan's Tia-Min Pang. "There are several different form factors. Instead of just the format we are used to in a laptop computer, there are several different ones for use in PDAs, airline video systems, low-profile desktop systems, and so on. Not all LCD processing tools are able to handle all form factors; that's why talking about raw capacity adds is not telling the whole story. Of course, this can be pushed too far, and, if prolonged, can lead to other problems like liquidity."
"Fabs are usually built to projections of forward demand, not what exists today," continued Pang. "The belief is that long-term demand for devices using various forms of LCDs will continue to grow. Japan, in particular, has shifted focus very much away from PCs to consumer devices, which utilize LCDs. Thus, the types of LCDs being made in Japan are very much targeted at specific end markets that may be quite different than elsewhere."
When Pang and Brian Matas, IC Insights Inc., were asked about the DRAM debacle, they were not shy about their assessments.
"The DRAM market? Ugh!" said Matas. "A shakeout in this field would be helpful. Perhaps as few as two or three major players in this market would restore some semblance of pricing sanity here. But when five 'big' players and a host of second-tier companies all want 20% market share — and they build fabs to back up their claims — it prevents the industry from moving forward."
 Figure 3. Capital spending vs. worldwide semiconductor market growth (1999-2005). |
Pang noted that the DRAM market has never really recovered from the massive overbuild in capacity that occurred in the mid-90s and is going through what he refers to as a process of natural selection. "Because of the Asian crash of 1997/98, several chip companies found themselves financially strapped and unable to invest fully," said Pang. "Those that were able to continue doing so, have really only exacerbated the situation since the underlying capacity is still in excess. Three years after the Asian situation has ostensibly passed and the industry has lived through one of the strongest demand periods for semiconductors, DRAM pricing is still weak and liquidity tight (Fig. 1). Leading companies like Hynix are voluntarily shutting down fabs for extended periods of time. However, it's going to take more than just Hynix to do that to change the situation materially. I believe we will see further consolidation and evolution here until only two or three players remain."
Hee-chan Lee of Gartner Dataquest thinks that the most important factor is how soon the global/US economy recovers along with the demand for electronics. "Although some DRAM suppliers plan to cut output lately, it is difficult to regard this as a long-term solution to the basic problem — recovery in demand," noted Lee. "DRAM prices have been declining steeply since 4Q00 due to sluggish demand, excess inventory, and oversupply. The DRAM industry is suffering from price declines brought on by falling PC demand and oversupply and is still searching for the bottom."
Japan and Korea
With all the doom and gloom market data, could there be hope? For Japan, the increasing demand for silicon in cellular phones — especially those with merged interconnects and packaging — broadband applications, and robots, combined with the new fab light business model might be the answer. Korea is still going strong in FPD, particularly Samsung and LG Philips, but recovery will probably be focused more strongly on the large-area TFT LCD market because of the growing demand for cellular phones with color displays.
 Figure 4. Total semiconductor equipment shipments and bookings (based on a three-month average). |
"Driven by the expanding global demand for electronic equipment, South Korea's economy staged a rapid rebound in 2000," observed Lee. "However, overinvestment in the electronics sectors and the demise of many dot-com companies has damaged the worldwide economy since the 2H00. South Korea is suffering from a serious recession and the market remains volatile, and thus, difficult to predict with much certainty. Gartner Dataquest expects that the American economy will rally by the end of 2001, and the excess inventory of PCs and cellular phones will be reduced by 2H01. We'll see a stronger rally by the 4Q01, and South Korea will regain its momentum in electronics production by the end of 2001."
"Japan seems like it has been stuck in a recession for at least five years," commented Matas. "Nothing new for it on the horizon. I'd not expect much out of Japan, Korea, or Taiwan for the next six months. It will take at least that long for these countries — as well as North America and Europe — to recover from the IC industry downturn. I don't think they're in any better position than any other region of the world to recover faster when business conditions improve. In fact, it may take longer for these countries to improve."
China
China is, of course, seen as the shining light in a sea of downturned tech stocks. "China is one of the hottest growth areas for new fabs because of the market potential and the access to good, low-cost engineering talent," observed Hutcheson, VLSI Research Inc. "The latter is probably the most important benefit, considering the persistent shortage of engineering talent in most semiconductor-producing areas."
Matas also thinks China is a bright spot, but cautions that even there, IC business is not booming as strongly as it was. "Still, China's economic engine is strong enough to support local IC market growth in the next six months, growth of perhaps 5-6% annually."
Senior analyst Dorothy Lai at Gartner Dataquest thinks the global downturn helps China. "The slowdown causes many foreign companies to think or rethink how to cut production costs," noted Lai. "One solution is to increase their investment in China to take advantage of the low-cost labor and government incentives and cut production elsewhere. Although global capital spending has slowed and that affects the overall foreign investment in China, in the long run, as China is preparing its entry into the World Trade Organization (WTO), foreign investment in China should increase."
"The Chinese PC market still had a stable growth of 28% last quarter [Legend (HK Stock Exchange) report released 8/14/01]," commented Lai. "The trend is echoed by our [Gartner's] PC and mobile phone research. The macro indices also show that China is still doing well." Lai observed that, while the semiconductor market in China is suffering, it is not as bad as in other countries in the Asia Pacific region (Fig. 2). "The reason is that China produces only approximately 10% of the semiconductors it needs. It needs to depend on imported sources. So basically, you have the volume, people are buying, but because of the global slowdown, foreign companies like to dump their products (not just semiconductors) in China. That means prices for most high-tech products have gone down."
Semi's C.T. Lin also thinks foreign investment will play a big role in China. "Perhaps the most interesting new development is the recent announcement by TSMC of its intent to look seriously at investment in mainland China projects," noted Lin. "As recently as six months ago, both TSMC and UMC downplayed the idea of building any operations on the mainland, but the investment climate and tax breaks offered by the Chinese government, coupled with the expectations that the Taiwanese government will relax its US$50M cap on mainland investment have brought a change of direction."
In comparing the market outlook among various regions, particularly major changes in status, Hutcheson noted that the question can always be answered by "who's got their capital pedal down in the downturn and who's let off....In the eighties, it was the US that let off the gas and Japan emerged. Japan let off the gas in 1989-1992 and Korea emerged. In 1998, Korea let off and Taiwan emerged. This time, it's Taiwan that's let off and China and the IDMs are going full bore. So expect China to emerge and the IDMs to come back."
 |
The foundry business model: Still viable?
When analysts were asked to comment on what new foundries expected to gain by starting up during the downturn, not everyone had something encouraging to say. "This is the first year since the foundry concept was introduced that foundries will lose share to IDMs," noted Hutcheson (Table 1). "I doubt that the new foundries will be successful because the market has matured and the business model has changed significantly. The business model no longer centers on capacity. Service is the issue. This is why, after almost 15 years, there are only three foundries of significance."
"The fabless/foundry model has been very successful in the past and should continue to be a major factor going forward," observed Pang. "The outsourced foundry model is very much the way the electronics industry is going. Just look at frontend foundries, backend test and assembly houses, contract manufacturing of electronic boards and assemblies, etc. Foundries account for about 20% of total capacity today, twice that of just a few years ago. But even here, weaker players disappear or are swallowed up by stronger companies. There is another intangible driver in the growth of foundries to date.
 |
"Many developing nations, wanting to join the global technology fraternity, have identified chip manufacturing as a relatively straightforward way to do it," continued Pang. "The companies that are thus spawned are usually government-funded and . . . not too concerned about making money, just establishing a good foothold."
"I don't think the foundry model needs to change so much as there just does not need to be as many participants in the business at this time," stated Matas. "It's an overkill situation. I don't have an explanation for why a company would be coming to market with additional foundry capacity when foundry utilization rates are at or near all-time low levels. What can they offer? What competitive advantage do they have that TSMC, UMC, or Chartered et al., don't already provide?" Matas also does not believe the IDMs will take back center stage, preferring instead to have their hands in both cookie jars (their fabs and access to foundry capacity).
 |
Philip Koh of Gartner Dataquest believes that the foundry market has changed over the last five years and some of that change has been brought about by the fabless sector. "Shorter product life cycles are the biggest challenge for foundry providers, whether existing or new players," said Koh. "Time-to-revenue and product life-span have been changing. Today, time-to-revenue is estimated to be about 9-12 months. Five years ago, time-to-revenue was 1.5-2 years. Product life-span today is probably 2-4 years, whereas five years ago it was between 5-7 years."
Koh further notes that, because a large part of the foundry business comes from the fabless companies that require rapid product evolution and fast changes in management and operations to meet market requirements and fluctuations, foundry providers need to provide flexibility and excellent services to respond to them (fabless companies).
Regarding the race between established foundry players like TSMC and UMC and the newcomers, Koh brought up a good point. "The two 'big boys' have announced new products at 0.1µm and pilot production availability for 300mm fabrication," commented Koh. "They are definitely gaining the upper-hand in terms of technology capability and are able to support large-volume capacity. But in the long run, how are they going to manage their older fabs with older process technology? This probably is a good sign for the new foundry players!"
Koh cautioned that the new foundry players, particularly in Malayasia and China, are facing a bigger challenge. "With the current global economic slowdown, the new players are seen to be coming in at the wrong time, in terms of gaining production capacity," said Koh. "However, for long-term success, the timing of the fabs' opening may be ideal. These fabs will start with a slow ramp initially and have some breathing room to purchase more equipment. When foundry market growth resumes, these companies will then be ready to reach full production."
While Koh thinks that competing with the big boys may not be easy, he notes that, "These new entrants still have a good chance of gaining a foothold and quickly establishing themselves in the niche marketplace. Continuing to partner and entering joint ventures with fabless companies or IDMs will definitely help during the start-up phase and building up the initial customer base. For long-term success, the newcomers must expand their customer base in order to reduce the risk of one company's success or unexpected situations. The earthquake-safe geographical locations, cheaper labor costs, and stable power supply generation are some additional attractive factors for these newcomers."
What about capital spending and equipment?
According to IC Insights' mid-year update report (p. 58), volatility in capital spending trends will continue (Fig. 3). It notes that, "Volatility is evident in both the good and bad years. For example, in 2000, capital spending increased 85% compared to 37% growth in the semiconductor market and in 1998, capital spending declined 25% while the semiconductor industry registered a more moderate 8% contraction." The research firm projects that the same pattern will continue throughout 2001-2005.
 |
The outlook on capital equipment spending does not look much better than for the industry itself. In its forecast in September, Semi, (Fig. 4) showed the 3-month average trend for worldwide capital equipment. At Semicon West, it looked like bookings had leveled off, with April and May 2001 figures very close ($1531.8 million and $1510.3 million, respectively). However, the June 2001 bookings figure was even lower at $1429.9 million. Lin does not see much over which to rejoice. "As we move through the 3Q01, forward visibility remains poor," said Lin. "While there have been some encouraging signs, there is still no indication of when the next up-cycle will actually begin. Analysts and industry leaders alike are as divided as they ever have been about the timing of the next upturn, with forecasts ranging from the 2Q02 to 1H03. We really will need to see signs of improvement in the global economic picture before we can have a better idea of when a return to growth in the semiconductor and equipment and materials industries might be." Tables 2 and 3 show equipment sales for 2000 and 1H01.
Beating the roadmap
The equipment sector should be encouraged by IC manufacturers' willingness to stay ahead of the industry roadmap. Dual damascene, especially deep trench fill, ultrashallow junctions, shallow trench isolation, and high-k gate dielectrics (ALCVD) are going to keep R&D budgets going in all directions. Equipment suppliers will have to be on the cutting edge of process development to keep up.
 |
Matas takes a reserved approach. "I think what appears to be beating the roadmap is actually the industry looking for competitive advantages through technology advancements, a rather typical move during a downturn," said Matas. "There is too much capacity on-line and enough new fabs are available to serve the industry for another year or longer. Companies will trim capital expenditure budgets for new fab shells and equipment during the next 12-18 months. They will, however, consider ways to gain a competitive advantage through technology improvements. This does not play a significant role in getting the industry through a downturn."
"The chip industry has always been a road race," said Hutcheson. "The nodes on the roadmap have become the industry's finish line and no self-respecting chipmaker ever wants to be the first loser. Being first has often led to great financial rewards. Just look at how AMD and IBM have arisen in recent years because of their technology." As long as chipmakers follow the industry roadmap, and not one conceived by Escher, there is hope. n
Debra Vogler is a senior technical editor at Solid State Technology magazine, ph 408/774-9283, e-mail [email protected].