Will Android be Profitable?
by Mark W. Hibben
Last week came the twin announcements of Google's Q4 financial results as well as the departure of Eric Schmidt as CEO in April of this year to be replaced by Larry Page. The uptick in Google's Q4 income came as a pleasant surprise, but the departure of Schmidt was probably less pleasant for most. Both surprises lead me to wonder, yet again, about the basic profitability of a key enterprise of Google, Android OS. It is an unusual business model for a commercial OS product: no licensing fees, with income derived solely from the additional ad revenue presumably generated for Google by Android devices. As is customary for Google, they offered no breakdown of revenue or costs by product line, and at the analysts' conference call on January 20, all questions probing Android profitability were gently but firmly deflected. Indeed, Android received scant mention in the prepared remarks of the CFO Patrick Pichette and the SVP for Product Management Jonathan Rosenberg. This only further inflamed my suspicion that Google's overall profitability might mask Android's basic unprofitability. I decided to attempt a deeper analysis of the Google's financial results in an effort to discern how Android does as a stand-alone enterprise.
Market Share and Android Population
I have a couple of times printed a graph based on Gartner Research data purporting to show that in 2010 Android OS mobile OS market share exceeded iPhone OS market share worldwide by 17.7% to 15.4%.
I accepted these numbers uncritically and assumed that they confirmed the explosive nature of Android OS growth. It appears that Gartner grossly overestimated the relative size of the Android device to iPhone OS device populations at the end of 2010. Although Google has declined to specify the exact number of Android devices in existence, Apple has been a bit more forthcoming. At the Apple earnings conference call on January 18, Tim Cook stated that Apple had delivered 167 million iOS devices, which I assume included iPads. Total iPad deliveries, according to official Apple announcements, totaled 14.8 million in 2010, so the total iPhone OS deliveries now stand at 152.2 million. Total deliveries isn't the same as current population however, since some iPhones will be bought to replace other, older model iPhones. I assume for both Android and iPhone that this self-replacement rate is low, since the devices haven't been on the market that long. Somewhat arbitrarily, I'm pegging the self-replacement rate for these smart phones at 10%. This means that the current worldwide population of iPhones stood at the end of 2010 at 90% of 152.2 million or 137 million. If Gartner's numbers are to be believed, then the population of Android phones stood at 1.15 * 137 M = 157.5 million Android phones.
The problem with Gartner's data is that at the activation rates that Google has claimed, there's no way for the Android population to achieve the Gartner derived population of 157.5 M. Even if you assume an activation rate of 300K/day for all of 2010, that's only 109.5 M activations for the entire year, and in fact, Google management have only been claiming a 300K/day activation rate since the Q3 earnings report in October. What was the starting number of Android phones at the beginning of 2010? Well, it wasn't 48 M. The most reliable number I've been able to find is that the Android OS device population stood at 11.6 M in April 2010. Based on this starting value, and modeling the Android population growth using a variable activation rate that ramped up from 100K/day in April to over 300K by the end of the year, I've concluded that the total Android OS population stood at 83.1 M by the end of 2010. The model results are shown graphically . Note that I don't expect the activation rate to increase indefinitely. From Google's statements, it appears to already be leveling out, and the population model reflects this.
Android Revenue and Cost Model
Getting a handle on Android population is essential to understand its role in the revenue and cost structure of Google. As an advertising company, Google's revenue (and to a large extent, its costs) are proportional to the number of devices that can host Google ads or sites
If we take Google's quarterly revenue and divide by the total number of devices that can host Google ads or sites, we can obtain a mean value for the revenue per device:
Device Revenue ($) = Total Revenue / (Number of Devices)
We can also apply a similar logic to estimating the cost per device for Google's Revenue Cost, which is mostly cost associated with traffic acquisition and is therefore proportional to revenue itself:
Device Revenue Cost ($) = Total Revenue Cost / (Number of Devices)
Revenue cost is Google’s single biggest cost, and just amounts to the old adage, “you have to spend money to make money”.
The other types of costs that Google outlines in their financial reports, General and Administrative, Sales and Marketing, Research and Development, don’t necessarily scale with device number. Nevertheless, I found I was able to apply as a first cut approximation the same principal of device scaling to these costs as in revenue cost. Basically, it works like this: as the device population increases, Google’s revenue and revenue cost increase. As revenue increases, Google gets bigger, so their other costs also go up. Likewise, for a small business segment such as Android (based on revenue), management tend to scale spending for the segment in G&A, R&D and Sales and Marketing proportionately.
I'm proposing the device scaling principle as a means of estimating the portion of the business (both costs and revenue) that falls under the Android umbrella. The approach isn't perfect, but it does give some feel for the relative contribution of Android to Google's overall business, as well as how Android might perform as a standalone business. The key fact to keep in mind is that Android must be a tiny fraction of Google's overall business, at least from a revenue if not from a cost standpoint. This is because the number of personal computers worldwide completely dwarfs the number of Android devices, and even dwarfs the combined number of Android and iOS devices.
As of the end of 2010, there were approximately 1.42 billion personal computers, growing linearly at the rate of 900K/day. In the estimated income chart, I show model results based on device number scaling for operating income (revenue – costs) for estimated income from PCs and Android OS, compared with the Google actuals for 2010.
The chart is subject to a number of simplifying assumptions. The first is that at the beginning of 2010, Google income was driven by advertising on PCs and that mobile was insignificant. This allowed me to estimate the income/device based just on the worldwide PC population. This comes to US $1.92/device per business quarter. The model also assumes that the income/device remains constant over time and is the same regardless of the device type, which can be only approximately true at best. However, at this stage, given the lack of information, I have no way to adjust the income/device number over time or for device type. It's very possible that Google's earnings improvement towards the end of the year came from improved revenue/device performance as a result of improved Google site design, improved ad quality, and other features that were added such as Google Instant. Also, the region where actual earnings fell below the PC-only model (Q2-Q3) clearly indicates that not all costs are scalable with device number. Furthermore, the growing use of the Internet by mobile users certainly contributed to the growth of income in Google's Q4.
Final Words of Caution Regarding Android Profitability
The above results indicate that Android was profitable in the sense that operating income was always positive for the modeled Android revenue and costs. While I’m fairly confident about the revenue model for Android, I’m much less confident about the cost side of things. Since the Android revenue is relatively modest, being only about US$ 430 M in Q4 by my estimate, perturbations in Android costs could easily push Android “under water” with negative net income. R&D especially, doesn’t scale well with device population. Google could be devoting a much larger fraction of their R&D budget to Android development than the current Android population would imply by the scaled cost model. Given that the Tablet-ready version (Honeycomb) is due to appear shortly, much of the latter half of 2010 was probably devoted to Honeycomb in the Android development segment of Google’s business. The Android device-scaled R&D cost model estimates a cost of US$51.5 M for Android R&D in Q4 and US$112 M for the year. The real development cost for Android last year could have been a few times this number.