High Tech…the Boom/Bust Cycle of a Generation?

Orders for Computers, Communications Equipment, and Semiconductors

This reflects above all the rise in the market penetration ratio of PCs and related products to levels from which further rapid market penetration in percentage terms is no longer possible.

Accelerating supply meeting stagnating demand has resulted in the first---and very large---decline in average selling prices for PC's. As a result of this decline in average selling prices, the relative decline in PC prices per functional unit has accelerated. The key issue now is, what will be the demand elastic response to an acceleration in the relative price of PC performance?

To understand this process, we must understand the difference between short and long run price elasticities. The price elastic response we have been discussing to date reflects long run price elasticities. When PC performance fell in price in the past, the market expanded significantly on a permanent basis; the new sales level that was reached remained stable. This is not the case, however, for markets that achieve saturation. Take, for example, the automobile market. When Henry Ford reduced the average selling price of an automobile with new production efficiencies, he brought the automobile to the masses in a way that changed the world. However, today, virtually everyone who wants a car now owns a car. If the Ford Motor Company wants to reduce its inventory and introduces steep price discounts, it improves sales on a temporary, not a permanent, basis. Consumers, perceiving a bargain, buy tomorrow's needs today. Sales, in effect, are stolen from the future. Sales expand and, then, in the next time period, they fall below trend. This transitory sales response to a relative price decline in a mature product market is a short run price elasticity which must be distinguished from the long run price elastic response of demand during the phase of market penetration.

The key issue then is whether the surge in nominal orders I computers, communications equipment and semiconductors in the first two months of 1998 is a short run or a long run price elastic response. If the PC market is really saturated, then it will prove to be a short run price elastic response. Sales will be stolen from the future and sales will fall below a slow overall growth trend in coming quarters. This will be a disaster for the industry and for high tech investors.

Industry participants and investors have adaptive or backward looking expectations for industry sales growth. Because PC sales grew faster for longer in the 1990's, they believe they will continue to do so, despite the growing signs of market saturation. Therefore, they interpret every slowdown as temporary and every pickup in sales as a resumption of the new higher trend. The order recovery in January/February 1998 visible in the above table is being regarded a harbinger of resumption of the growth trend of the 1990's. If, instead, it is a short run price elastic response to drastic price cuts, coming quarters will produce a trend in unit sales that will fall below the single digit rate of 1997 despite huge product price cuts.

There is a potential in the PC related high tech sector for boom bust dynamics of unprecedented proportions. First, PC sales have clearly fallen substantially from a peak rate of 25% per annum. Yet capacity growth in the computer, office equipment sector, and components markets globally remains rapid. If strong positive capacity growth with its associated fixed charges is met with low nominal revenue growth, all other things being equal, the capacity utilization rate will fall and industry profitability should fall rapidly with it.

The Problems of the Industry Do Not End There

There are no end of storm clouds on the demand side. First, unit sales and overall industry revenue growth have fallen sharply during a period of accelerating economic activity. The recent well above trend growth of the US economy cannot be sustained. Typical of such boom periods are phases of above trend rates of accumulation of big ticket durable goods by households and firms. Cyclical conditions alone point to an eventual further fall in the growth rate of PC purchases, at least in the U.S..

Second, the pace of PC related high tech sales growth is, to some extent, a function of product improvements that stimulate demand. Apparently, the improvements that lie ahead over the next two years are marginal in nature. Processor speeds will be incrementally increased, as will memory capacity. But there will not be a new generation of processors for years. The next generation will not come to the market until late 1999, and then it will be applicable only in servers, which generate only 2% of the global revenues of the PC market. According to Andy Grove, this new generation will not have widespread use for five years.

It is widely argued that ongoing product enhancements provide little incentive for computer users to upgrade to new models. Again, we quote Fred Hickey:

"A PC World reviewer was forced by his editor to use a $797 Kingdom Pinnacle PC for a week. It had a 166MTTz Cyrix processor, with 32 Megs of DRAM, and a 1.7 gig disk drive. He had just bought himself a $3,000 Dell XPS D300 with a 399MTTz Intel Pentium II chip, 96 Megs of DRAM an 8.4 Gig disk drive, and lots of secondary cache memory. His conclusions after comparing the two PCs: "Surprisingly, the performance of the two machines felt similar." "For day-to-day productivity tasks (word processing, e-mail, web-surfing and balancing a checkbook), I saw little difference between them." "Did I feel buyers remorse? You bet."

At the same time, it is possible that ongoing manufacturing efficiencies and growing competition and oversupply may keep average selling prices in a steep decline. Microprocessor prices may collapse. AMD and Cyrix now have microprocessors that compete with Intel at virtually every level of performance. On a combined basis, their competitive microprocessors were produced at only a 12 million unit annual rate (in a 100 million unit industry) in the first quarter of 1998. Virtually all PC vendors have now introduced non-Intel based PC's. AMD, Cyrix, National Semiconductor, IBM and the largest custom fabs in Taiwan and Singapore now plan a production rate of more than 60 million non-Intel units by early 1999. In addition, other large and small potential competitors are seeking to expand into the microprocessor market. Comparable competitive pressures are reducing prices of the other components of the PC.

There is the possibility of a very unusual outcome in the PC-based high tech industry. In most saturated markets for new innovations, technological change, manufacturing efficiencies, and profit margins all mature as the market approaches saturation. As a consequence, relative product prices stabilize with market maturation, and so do industry revenues. But it is possible for technological change, manufacturing efficiencies, and profit margins to continue to bring relative product prices down rapidly at the moment of market saturation. Then, price declines can exceed long run demand elastic responses for units. Then the elasticity of demand falls well below unity and the industry experiences declining revenue.

It would be a great shock to Wall Street if this great "new era" growth industry actually experiences outright negative sales growth during an economic expansion. And, of course, declining industry sales revenue would be an extreme disaster for the industry. The dollar value of its capital spending and hence its future fixed costs would still be rising rapidly, and profitability might then collapse.

There are signs that something like this is happening in the computer industry. This tendency toward an elasticity of demand less than unity is most apparent in the semiconductor memory market. In 1995 a 4 megabit memory chip cost $15. Today a 64 megabit memory chip costs less than $10. The demand for bits of memory has increased rapidly, but not as rapidly as prices have declined. Global revenues for semiconductor memory were $40 billion in 1995. They were only $18 billion in 1997. The leading research houses predict a further 15%-20% decline in global revenues this year. Despite the lower prices, users simply do not need the additional memory needed to keep the elasticity of demand above unity. It is estimated that current production rates are sufficient to provide each computer now coming off the assembly lines with 80 megabytes of memory. Few users need so much memory regardless of its price.

With an almost doubling in microprocessor production in the pipeline it is probable that we will see declines in microprocessor prices that are so great that the elasticity of demand for this key PC component may also fall well below unity. It is not hard to imagine that prices of PC's will then fall so rapidly that the total revenues of the PC-based hardware industry will go significantly negative. If this occurs, the usual four year cyclical boom bust dynamics of the industry would be compounded by the worst market saturation dynamics that can ultimately affect markets for new innovations when these markets finally mature.