As we were coming out of the post-Dot.com recession, I formulated a theory as to why that particular downturn was so especially hard on the IT industry. I came to the conclusion that it was an instance of a perfect storm: three bad trends peaking at the same time. The first of these was simply that there was a general economic downturn. The second was the historically unique Y2K event, which caused about two or three years worth of systems upgrades to get front-loaded prior to the deadline, with a corresponding fall-off in 2000 and after.
The third effect was less clearly defined, but may be of more interest, because I expect that it will continue to affect the IT industry. Technically, Moore’s Law refers to the doubling of the number of transistors that can be placed on an integrated circuit, with this doubling predicted to occur every two years. Most of us are more familiar with the resulting implication: computers, cell phones, and other electronic gadgets tend to double their performance (usually at a lower price) about every 18-24 months.
However, “doubling of performance” means different things in different contexts. Transaction processing does not account for all computer usage, but it does drive the market for large, networked systems purchased by large corporations and government agencies.
In a transaction processing environment, a major (though not only) measure of performance is the time that it takes to process a transaction. In pre-computer days, manual systems might take up to a month to process a simple invoice or purchase order. The first computers cut these times drastically. Even prior to ubiquitous networking of computers, a new generation of computers could halve the time needed to turn around a particular type of transaction.
So, if the first generation of computers cut turnaround time from a month to a fortnight, everyone needed to get one, for reasons of competitive advantage. If the next generation cut the time from two weeks to one, then one week is now the industry standard—and the hardware vendors are happy to provide the forklift upgrades.
However, what happens when transaction processing time, for certain types of transactions, are measure in seconds, rather than weeks, days, or hours? At some point, the halving of transaction processing time stops being a huge competitive advantage—particularly when the cost to achieve it is very high.
So, if your system takes, say, 8 seconds to process a transaction, it is unlikely that you will lose customers to a competitor solely because the other firm can promise a 4 second processing time. True, 4 is better than 8—but now people look closely at the cost of those marginal 4 seconds, as the halving is no longer a matter of competitive necessity.
Once this threshold is passed—and it will vary by application, by transaction type, by vertical industry, and by other factors—demand for new hardware by no means goes away. However, it will start to bear a closer relationship to the more gradual increase in number and complexity of transactions, rather than the steeply rising curve driven by Moore’s Law and competition.
This third effect, if combined with the first two, could go a long way towards explaining the depth of the post-Dot.com doldrums. However, a more interesting aspect is the idea of being able to spot a similar inflection point in other areas of IT. Anyone who can figure out how to do this—and, hey, we’re open to your ideas on this point—can wind up making a fortune. Assuming, that is, that you have the resources and nerve to short companies that have hitherto been following an upward exponential curve.