I think, partially, Smalltalk has never reached a kind of critical mass to attract widespread attention from non CS geeks.
Part of the problem is also the slight lack of easy deployability, as already mentioned. However, I think the seeds are already sown, e.g. for browser accessed-apps that can be dynamically tinkered with.
Imagine a Croquet world accessed through today's web browser where everything is editable in Smalltalk... I'm sure that would be something of a killer app. I think the Second Life etc models don't have nearly as clean or open an architecture to rival it.
The other problem is that Squeak has been behind the curve (probably because of available resources!) to interface with "hot" topics e.g. 3D rendering, video, databases. There have usually been more mature and readily available libraries in less elegant languages.
I think the whole "reinvention of computing" effort is ripe for fruition, though - particularly as smaller, cheaper mobile devices become ubiquitous. Smalltalk could easily exploit it's agility in these areas, and I think this is indeed happening. We need a Smalltalk phone now, dammit! It would only be a matter of time before the brighter kids are hacking in Smalltalk, and then the world is set for a revolution.
Ryan