In some sense, it "matters" as much as dozens of other "fringe" language projects (Ceylon, F#, Kotlin, etc.). It brings a different way of thinking to the table. If that way of thinking is one you like, great. If not, move onto the next one.
But beyond that, I think the whole reason Pharo matters is that its team is doggedly bringing Smalltalk out of the historical context. Pharo has libraries for things like XML, JSON, OAuth, Redis, and OpenID so that you can interact with other tools. It has support for command-line applications, comes with an easy-to-use C FFI, and sports a fast JIT so that it's usable for real applications. It has rewritten libraries for speed, breaking backwards compatibility where necessary, and has completely overhauled the UI, while keeping all the pieces that made Smalltalk great. So part of why Pharo matters is that it makes Smalltalk a real, modern language, that can trivially integrate with the massive ecosystem that is existing software. No longer is Smalltalk living in its own little world; it now plays well with others.
As for Smalltalk itself: it still has valuable lessons, even today. Smalltalk is the only system I know where I can trivially save the running app to disk and open that running app, complete with all of its state, on another system to debug what went wrong. It still has some of the best debugging and development tools of any environment I've used for any language. Smalltalk in general, and Pharo in particular, is still one of the most solidly turtles-all-the-way-down environments I have used in my life, where there's no meaningful distinction between your app, and the core system libraries.
Smalltalk may not be for you, but Pharo makes it a genuinely viable choice for real-world development.