They're big enough to get some data by trying things.
failure is a data point. As a scientist, I would say, arguably the more important data are failures. Consider it an exercise in falsification.
i agree that failure is an important data point, but i don't understand why that means amazon would want to build the fire phone. their reasoning couldn't have been "well this seems like a good idea, and if it fails, at least we'll have a data point!" I am curious what caused them to think the fire phone was a good idea.