Notes on exponential revenue growth
One of the more fascinating aspects of a speculative market is the classic enthusiasm for new issues, and the emergence of growth “darlings” whose long-term exponential growth is taken as a sure thing. This movie has played so many times in the historical data that we’ve practically memorized the lines. Near the end of the tech bubble, I got myself a nice bit of criticism on CNBC when Alan Abelson of Barron’s Magazine published my projections for Cisco, EMC, Sun Microsystems and Oracle – all in the range of about 15-20% of the prices where they had recently changed hands. Those projections actually turned out to be slightly optimistic.
Given that we’re seeing some similar behavior today, it’s probably a good time to discuss the dynamics of growth. Consider a very large, untapped market for some product. We can model the growth process in terms of how quickly that product is adopted by new users, whether there are any “network” effects where new buyers are attracted to the product because other people already use it, how frequently existing users replace their products, whether late-adopters come in more slowly than early-adopters because of budget constraints, how quickly the untapped market grows, and a variety of other factors.
Whether you do this sort of modeling with a spreadsheet or with differential equations, you’ll get essentially the same results. Specifically, growth rates are always a declining function of market penetration. Most strikingly, the growth rates begin to come down hard even at the point that a company hits 20-30% market penetration. Network effects accelerate the early growth, but also cause growth to hit the wall more abruptly. Replacement helps to accelerate the early growth rates too, but ultimately has much more effect on the sustainable level of sales than it has on long-term growth. In fact, if the replacement rate (the percentage of existing users that replace their product each year) is less than the adoption rate (the percentage of untapped prospects that are converted to new users), it’s very hard to keep the growth rate of sales from falling below the rate of economic growth.
The chart below gives the general picture of various growth curves and the effect that different factors can exert. The paths are less important for their actual growth rates as they are for their general profiles (below, I’ve assumed that 15% of the untapped market adopts the product each period). It may seem odd that you could get a growth rate below the adoption rate. But notice that with an adoption rate of 15% and a total potential market of 1000 units, for example, you’ll sell 150 units the first year, but the next year’s sales will only be 15% of the 850 remaining untapped prospects, so growth will actually be negative unless you have other factors contributing, such as discovery, replacement, network effects, and so forth.
To see how all of this has played out in the actual data for past market darlings, let’s take a look at several extraordinary growth companies that can now reasonably be viewed as having reached their “mature” level of market penetration: Microsoft, Cisco, Intel, Oracle, IBM, Dell and Wal-Mart. The chart below presents the combined scatter of historical revenue growth and penetration data for these companies. Again, the key feature is that growth rates are a rapidly decreasing function of market penetration.
With regard to the elephant in the room, which is Apple, my impression is that what appears to be endless exponential growth is actually the overlay of three separate logistic growth curves – one for the iPod, one for the iPhone, and one for the iPad. These are great products. Still, in order to maintain even a constant level of sales, every unit sold in a given year has to be matched by a replacement the next year – year-after-year – or it has to be matched by a new adoption, and adopters of used products don’t count. Simply put, even zero growth demands that every dollar existing users spent on Apple products last year has to be spent again this year, or matched by some new user this year, and then again next year, and again the year after that, ad infinitum. Of course, it’s reasonable to expect that late-adopters (e.g. those who have to save in order to afford the product) will have lower replacement rates, which will need to be offset by even greater adoption. Yes, there are billions of people in developing countries without an iPhone. Unfortunately, most of these people are also without running water.
We’ve seen very rapid adoption rates, very high replacement, and very strong network effects in Apple’s products. All of this is an extraordinary achievement that reflects Steve Jobs’ genius. I suspect, however, that investors observe the rapid adoption and very high recent replacement rate of three very popular but semi-durable products, and don’t recognize how improbable it is to maintain these dynamics indefinitely. Despite great near-term prospects, within a small number of years, Apple will have to maintain an extraordinarily high rate of new adoption if replacement rates wane, simply to avoid becoming a no-growth company. That’s not a criticism of Apple, it’s just a standard feature of growth companies as their market share expands. It’s something that Cisco and Microsoft and every growth juggernaut encounters. Apple is now valued at 4% of U.S. GDP, but then, Cisco and Microsoft were each valued at 6% of GDP at the 2000 bubble peak. Not that things worked out well for investors who paid those valuations. There’s always the hope that this time it’s different.