It's an intriguing, albeit risky move. Existing smoke detectors are pretty ugly with bad UI. There is little to no competition for the 'high end' retail market. With that said, current smoke detectors can be had for $10-20 bucks, and my guess is Nest will charge significantly more than that. Of course they can add a number of extra sensors/functionality to justify the costs. A CO sensor would be a good addition, maybe motion sensor (for security), temperature sensors (to provide more granular feedback to the thermostat), and maybe some other sensors to monitor indoor air quality (VOC's, radon). I am sure there are a few other things that can be thrown into the device to make it more useful (wifi repeater?).
All things considered, this is a pretty genius move. In the push to create 'smart homes' most of the attention has been on lighting, kitchen appliances, and entertainment systems. But one thing that has always bugged me was how do you get a refrigerator at the back of the house to communicate seamlessly with the wireless router in the front if the signal doesn't reach that far. I was thinking that mesh networking would play a big part, but even in an appliance saturated home, there could still be some dead spots. Well, smart smoke detectors might solve that, as they are in almost every room in the house! One issue I can forsee is that not all homes have hard-wired smoke detectors (which means power would be an issue, thus limiting what they can do with it).
EDIT: Seems like everyone is focusing on the UI comment. While I think that that is still an issue, I think the more exciting possibility is that a 'smart detector' could also provide nodes for a mesh network (or something similar), essentially providing a communication layer for appliances/devices in the home. If they go this route and they succeed, they could put themselves in a really good position competitively, as they would 'own' the networking layer.