Don't listen to Ballmer. It really is all about the software.
Flurry tracks developers' new project starts using Flurry Analytics. It's unclear whether there's systematic bias in the survey population that would undermine the utility of Flurry's measure (e.g., if we trust that one in five App Store apps uses Flurry, and one in five Android Market apps uses Flurry, can we derive any conclusion about Blackberry development?), but across the survey population it's apparent that iPad development has become significant.
The new project starts isn't a function of application availability on the various platforms: it's a function at least in part of the novelty of the platform. For example, Android is more novel than RIM's Blackberry and had more than four times the apparent development effort. However, lots of developers' applications have been on Blackberry for years – plainly obviating the need for a new project to make the developer's app available. The real story is that although Android is a very new platform, it was dwarfed overnight by iPad development. It should be unsurprising that RIM's more mature platform lost new project share while iPad development was just beginning, but it's interesting that RIM's drop of only 3% of share lowers it to 25% of its former share, and from that perspective is more drastic than Android's share drop from 18% to 10%. The iPhone's share drop from 78% to 67% is roughly similar in absolute share drop as Android's drop, but Android's loss of 8% in share is a 44% drop, whereas iPhone's 11% drop was a loss of only 14% of its former share.
Adding iPad's share to iPhone's gives Apple-vended platforms 89% of new application starts. (Flurry doesn't seem to be used by WinMo developers, so this number does have some recognizable limitations – and what of Symbian, WebOS, etc.? Are these even significant?)
Apple's dominance in the handheld market continues undiminished by recent developments among rivals.