I find it interesting that so few are working on the actual IoT devices but so many are working on software to support it.
I think you know the answer: generally, hardware is harder than software for a startup. There was a post about a hardware kickstarter project that failed: "Kickstarter project spent $3.5M to finish a prototype and ended in disaster" [1] and a list here: List of Failed Kickstarter Hardware Projects [2]
And not just that hardware is harder but also the IoT market is estimated at $19 trillion [1] creating a gold rush of sorts.
But yes, there is no shortage of IoT platforms (xively, dweet, nitrogen, data.sparkfun, ...) and I'm partly in involved as well. What I found interesting about the MS program is that recently they were giving away boards for you to try. So it seems they are in a position to provide the hardware and the platform.
I think you know the answer: generally, hardware is harder than software for a startup. There was a post about a hardware kickstarter project that failed: "Kickstarter project spent $3.5M to finish a prototype and ended in disaster" [1] and a list here: List of Failed Kickstarter Hardware Projects [2]
[1] https://news.ycombinator.com/item?id=8016201
[2] http://www.kickspy.com/browse/all-failed/hardware