thinfilmmfg.com Subscriber Index DRAM market weakness

WWW thinfilmmfg.com

20 November 2001

DRAM market faces structural weakness

• More than just another cycle
• Consolidation and diversification aren't so easy

For years, emerging economies have seen the semiconductor market as key to the development of local technology industries. Semiconductor manufacturing both encourages and demands a stable infrastructure and a well-educated labor pool. It requires relatively few raw materials, creates relatively little pollution, and yields an extremely valuable and portable product. It supports a wide array of supplier businesses, from cleanroom maintenance to electronic system design. Japan, South Korea, and Taiwan have all built powerful technology sectors on their domestic semiconductor industries. Mainland China, Malaysia, and other nations are positioning themselves for similar leaps.

The memory segment has been especially lucrative for new semiconductor powers. First Japanese giants like NEC and Toshiba, then South Korean newcomers like Samsung and Hyundai (now Hynix), thrived on the PC market's voracious appetite for memory. Emerging semiconductor manufacturers often focus on memory as well.

Yet, since 1995 the memory market has been a lousy place to do business. From an all time high of US$53.5 billion in 1995, total memory revenues shrank to $23.0 billion in 1998, and only recovered to $49.2 billion during the boom year of 2000. For 2001, the WSTS predicts only $24.9 billion in memory sales. In other words, this "lucrative" segment of the semiconductor manufacturing industry shrank by more than 50% from 1995 to 2001.

Source: WSTS

Economists say that declining revenues are a clear sign that supply exceeds demand. Yet Micron, Samsung, and others have maintained or increased manufacturing capacity during the current downturn, and several manufacturers will continue to do so in 2002.

Why? Why do otherwise rational businesspeople invest so much money in a shrinking market?

Back to top

More than just another cycle

The first answer is that drivers of memory demand are still strong. More and more electronic systems require more and more data space. Nikkei Market Access predicts that memory demand will increase in bit terms by as much as 36% in 2002. The same study predicts that capacity will grow by 52%, however.

The demand-driven view holds that a macroeconomic recovery will cure all that ails the memory market. Yet the boom year of 2000 still fell short of 1995's record sales. The PC market, the largest consumer of memory, is maturing. Many of the most interesting new applications are limited by bandwidth, not memory.

A second answer may be that manufacturers view the last five years as an anomaly. The price collapse of 1996 followed nearly three years of stable memory prices and rising profitability. When those days return, the argument goes, companies with large amounts of leading edge capacity will reap huge rewards. This argument is circular, though. If all companies continue to build more fabs, then supply will continue to outpace demand and memory revenues will continue to fall.

Anomaly theory proponents observe that the Asian financial crisis of 1997 and 1998 distorts dollar-based comparisons. Or does it? Because the US dollar is a relatively stable currency, it is the currency of many international transactions. Currency devaluations in Asia eroded the purchasing power of the entire region, including buyers of electronic systems and capital equipment for semiconductor manufacturing.

Moreover, the Asian financial crisis has been blamed in part on questionable lending practices, and the semiconductor industry was as guilty of those practices as any other. Banks financed enormous capacity investments based on the assumption that DRAM prices would remain high. After memory prices collapsed in 1996, semiconductor manufacturers were left with debt-to-equity ratios of several hundred percent, threatening their own solvency as well as the solvency of their lenders. The threat of massive bank failures helped to undermine currencies in the region.

Intervention by the International Monetary Fund helped stabilize the currencies, and was supposed to lead to significant banking reforms. It worked to some extent. Fabs were sold and the most troubled manufacturers merged. At least one overly ambitious construction project folded after it lost financing. And yet, just a few years later, memory manufacturers are still investing vast sums to gain share in a shrinking market.

Many observers blame current conditions on Hynix, which has restructured its debts twice already this year. The most recent agreement calls for a US$2.1 billion debt-to-equity swap, restructuring of an additional $1.9 billion in debt, and the sale of $750 million in new securities. The package includes financing for substantial technology upgrades. Critics note that Hynix's creditor banks are partially owned by the South Korean government, which routinely injects public funds into domestic banks. In other words, critics claim that the restructuring of Hynix's debt was carried out with public funds and therefore constitutes a government subsidy of the troubled manufacturer. Hynix complains that it is being singled out unfairly, claiming that it has reduced capacity while rivals like Samsung and Micron have not. Samsung retorts that its capacity increases are merely a natural result of aggressive linewidth shrinks.

There lies a third part of the problem. When DRAM demand exceeded supply in the early 1990s, Moore's Law worked in favor of manufacturers. Shrinks allowed them to cut costs, while high demand allowed them to maintain prices and collect huge profits. Those profits fueled further investment and attracted more companies to the sector. Now that demand has softened, Moore's Law forces suppliers into a race for the bottom. Advanced technologies are essential if manufacturers are to maintain a competitive cost structure, but investment also makes the global memory glut worse. Though government supports of various kinds did not create the problem, they make it worse by impeding consolidation.

To summarize, the DRAM market's present woes point to structural weaknesses rather than just another cyclical downturn. Continual overinvestment by established companies seeking to minimize costs contributes to systemic overcapacity. New suppliers, often driven by nationalistic rather than business motives, create still more capacity. Existing suppliers are often willing to license process technology to newcomers, which reduces the technological barrier to entry. Licensing also reduces differences between competing memory products, forcing suppliers to compete on price alone.

Back to top

Consolidation and diversification aren't so easy

The situation cries out for consolidation, and in fact merger rumors have been flying in all directions for months. Toshiba has announced it will leave the memory market. Infineon, rumored to be a potential buyer of Toshiba's memory business, has also been linked with several Taiwanese memory suppliers. It's not clear that such relatively minor mergers will fix the industry's underlying problems. Governments that have subsidized manufacturers in order to create domestic semiconductor industries are unlikely to allow those manufacturers to fail or be purchased by foreign interests. Nor are highly subsidized manufacturers likely to back away from their expansion plans during economic rough spots. In fact, any short term price stabilization is likely to encourage still more investments.

Several companies are reportedly considering filing antidumping complaints with their respective governments. It's worth noting that Micron has yet to do so, even though the US-based memory manufacturer has pursued antidumping claims very aggressively in the past. Several countries have suggested they will raise the issue of government subsidies for memory manufacturers with the World Trade Organization. Such complaints may be politically problematic, however, as both consumers and electronics systems manufacturers benefit from low memory prices.

Some DRAM manufacturers have attempted to diversify their production capacity by offering foundry services as well. These efforts have met with only limited success. The high-volume, low product mix production required for profitability in the memory market is directly opposed to the foundry market's low-volume, high-mix business model.

It seems likely, then, that the DRAM market will remain precarious for some time. A macroeconomic recovery will help increase demand and stabilize prices, but is unlikely to fix the industry's structural weaknesses. For a long term solution, companies must be willing to abandon the market, and governments must force them to face the consequences of poor investment decisions.

Back to top