We’ve all been admiring of the seemingly unstoppable high rate of growth in the computer industry, despite tech-busts and dot-com blowouts. The view of this miraculous growth was mythologized by Gordon Moore, co-founder and CEO emeritus of Intel. Though Moore was accurate in his prediction on technological advances in microprocessors, he was lucky in having hit on a business principle fundamental to almost every industry. The implications of Moore’s Law and the principles behind it are especially relevant in this time of economic slowdown, because in periods such as these, profit growth strategies should focus on controllable elements, namely reducing costs and maximizing productivity.
The oft quoted Moore’s Law is sometimes mis-stated as computer speed doubles and price halves every 18 months. Actually in 1965, Gordon Moore initially asserted that “The complexity [of an integrated circuit] for minimum component costs has increased at a rate of roughly a factor of two per year…over the long term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least ten years.” This observation, that the number of transistors on a chip would double about every year, was quickly dubbed Moore’s Law, and with extension and slight modification has become the mantra and benchmark for computer innovation for the past 35 years.
With his prediction that the number of transistors per chip would double, at first every year, and later about every 18 months, Moore alluded to the fact that cost should halve in the same period of time because transistor density is so closely related to cost. However, the proper denominator for cost halving is not time, but accumulated experience. Moore was lucky that in the period he commented on, accumulation of experience correlated very closely with time.
Although the rapid growth and technological innovation of the computer industry has made it look glamorous, it follows the pattern predicted be the experience curve. Codified by The Boston Consulting Group (BCG) in the 1960s, the experience curve or e-curve is the traditional way of illustrating the potential to manage costs down with accumulated experience. According to Bruce Henderson, founder of BCG, “Cost of value added declines approximately 20 to 30 percent each time accumulated experience is doubled.” This decrease in cost is driven by advances in learning, specialization, scale and technological innovation and is nearly universally observable.
Moore essentially acknowledged the importance of accumulated experience in 1975 when he changed his timeframe from doubling every year to about every 18 months. He attributed this change, which is predicted by the e-curve, to having “no room left to squeeze anything out by being clever.” In recent years, transistor density has been growing at a rate less than doubling every 18 months. This is simply an issue of taking longer to accumulate a sufficient volume of experience to drive cost down. The computer industry e-curve, with a 71% slope, predicts that cost halves when accumulated experience quadruples (71% X 71% = 50%). The last time cost halved was over two years ago, and the next time that it will halve seems likely to be in four years. Intel itself has recognized this shift in timeframe. Andy Grove (former CEO of Intel) predicted that by 2011, Intel will produce a microprocessor with 1 billion transistors. Extrapolating out from the 9.5 million transistors on the Pentium III processor in 1999, Moore’s Law predicts about 2 billion transistors per chip in 2011, or twice the density that Intel thinks it can manufacture.
Though the technological advances in the computer industry are staggering, it is not fundamentally different from most other industries. It follows the same e-curve as the traditional examples of crushed and broken limestone and steam turbine generators, both of which have slopes between 70% and 80%. Surprising as it may seem, one of Henderson’s early examples for the e-curve was semiconductors, and, through today, the actual growth in the computer industry fits the e-curve remarkably well with accumulated experience predicting actual cost decreases with 98% accuracy.
The greatest unit cost reductions in the computer industry are found in the leaps to next generation technology. When a next generation chip is introduced, transistor density increases exponentially and unit cost per transistor plummets.
Chip size, and therefore transistor density, is the key cost driver in computer processor production. Area on a silicon wafer is a fixed cost of about $1 billion per acre (So to cover an acre—say the area of a football field or floor plan of Bill Gates’ house—with silicon wafers would cost $1 billion, while you could do the job in gold foil for a mere $750 million). Increasing transistor density is all benefits; smaller chips are cheaper, faster and use less energy than larger chips with the same number of transistors. Therefore a next generation chip with twice as many transistors in roughly the same area has a unit cost per transistor of approximately half that of the previous generation.
Marching down the e-curve and taking advantage of the potential to reduce costs, requires substantial increases in accumulated experience. The traditional ways to increase experience are to drive scale and capacity utilization. Activities such as acquisitions can increase scale, but to achieve reduced cost benefits of the e-curve, you need to streamline your business to fully utilize all of your capacity. The computer industry on the other hand increases experience through technological innovation.
Although it may seem like product innovation is the key driver in the computer industry, in fact it is just a major determinant of cost. The importance of the e-curve by no means contradicts the recognized relevance of time-to-market. Faster time to market gives the innovator many cost advantages, but most critical is the increased accumulated experience gained through: volume achieved initially, higher market share, and volume earned by deterring the entrance of competitors
Initial reactions in the post September 11 world have run counter to this wisdom. There has been much discussion about the avoidance population centers which would make efficient large-scale operations difficult. Additionally, an increased interest in redundancy increases costs and makes full capacity utilization nearly impossible.