Where will you be building these?

I think in this case it makes sense to sell units, and also generate feedback about the product as early as possible

So you're going to use end customers as your test base? That's pretty gutsy. Especially when they're shelling out $10K and ripping out their existing (and working) HVAC system to do this. Don't use the car analogy here, there's no backup if it doesn't work.

Unless you predict 100% flawless execution and satisfaction on that aspect, you're going to get screwed. Your customers, more so. A few fucked up installations, or performance below expectations, and it's over.