More extrapolation from Google IO numbers, this time looking at tablets. Google gave two 'charts' on this topic, shown below.
In addition, Google's developer dashboard gives a breakdown of active Android devices by screen size: in the first week of July they accounted for 12.6%, not including phablets. And since Google also, for the first time, told us the active base, 'over 1bn', we can calculate an active tablet base of at least 126m devices, rising to 138m if we take 'over 1bn' as 1.1bn.
Apple, almost uniquely in the market, reports tablet sales and shipments. In the last 12m (which I presume that "62% in 2014" bubble refers to) iPad shipment were 69.7m and sales were 73m. If Android tablets indeed had 62% of sales (the chart isn't clear if it refer to sales or shipments) then that would equate to a market of 192m units and 119m Android tablets sold in the last 12 months.
This would imply that Android tablets have an average active life of a little under a year. Or, that they have a life of more like two years (say) but half of them are inactive.
Conversely, it's pretty clear that the active life of an iPad is if anything too long, from Apple's point of view - at least two years and probably longer. Apple has sold 143m iPads in the last two years and 196m in the last three.
That, in turn, implies that each iPad unit sale is worth 2x more active devices over time than an Android tablet sale - before looking at actual usage.
There are, though, two problems with this data. The first is that it's not clear if that 'Android tablets market share' includes AOSP tablets in China, where the vast majority Android devices have no Google services and hence do not show up in Google's active user figure or the developer dashboard screen size breakdown. Sundar did say that this doesn't include Kindle and other Android variants, which would add 'a few percentage points'. So it probably doesn't attempt to include China.
The second is that it appears that a very large number of cheap Android tablets are used mostly or entirely offline, consuming video via SD card and playing games. So those might not be in Google's MAU data either.
A year ago, Google gave a more explicit set of numbers for Android tablet sales, in the chart below. If you eyeball the numbers, that equates to 50m tablet activations (i.e. not AOSP) in the 12m to July 2013.
(I'm not sure, incidentally, why this sources Gartner and IDC as well as 'Google internal data' - shouldn't Google know activations?)
50m tablet activations in that 12m period compares to 70m iPad sales in the same period (more or less), giving a total market (excluding AOSP) of 120m units and Android a market share of 41%, which is probably close enough to the cited 46%. Activations in the previous 12m look like about 15m: Apple did 53m, giving Google Android a market share of 22%, though, not 39%. Something has got lost in those numbers somewhere along the way.
One more thing. Google's internal data on tablets activations and device screen sizes is self-reported by the devices when they access play, using the Android screen profiles described here. But devices that are on the margin between two buckets can and do change what category they report from region to region and from one software update to another. This is a particular issue for phablets, where it's really a matter of opinion whether you should tell an app this is a normal or large screen, and whether it's xhdpi or xxhdpi.
This, incidentally, is one reason why I sometimes mark shares of tablet sales with terms like 'approximate'.
Finally, of course, all the available usage data of tablets shows iPads with around three quarters of total use, even in China.