I think it'll be at least another year before we start to see much movement on the J2ME front. I've been saying this since the summer. With the recent economic downturn, it could even be further out.
I say this for two reasons. First, aside from phones, wireless is more hype than substance right now. The infrastructure needs to grow a little more before serious wireless applications are supported. I think the unwired handheld market is about reaching capacity in the US, so wireless is the way to go.
Second, J2ME has a lot of issues. I give Sun a lot of credit for creating an "semi-open source" language, and getting buy-in from the commercial vendors. However, this works best when the market has already optimized the solutions and the
Java APIs are simply mimicing what's been proven to work. With J2ME, everyone is making it up as they go. We don't know that this is the right way to do it, we're just taking out bets guess. So I think J2ME still needs time to mature. (Contrast this with, say messaging/JMS, which had been developed over the last 10 years, before it was adopted into Java.)
The best skills you can have for J2ME is to understand the JVM and how it works. Being able to optimize for memory and processor speed will be key in the early days, and it's a skills applicable to any Java project.
--Mark