Are we stuck in a GHz barrier?

Yes but ghz is just one part of the overall package.

Phenom 2s already can get to over 7 ghz with liquid nitrogen cooling, and the intel chips aren't far behind. That proves that it can be done, and 10 ghz might be within reach soon but still not for daily users for a while yet  

What they have been doing, instead of increasing the speed of the core, is increasing how much it can get done in a single cycle. The architectures have been improving and becoming more complex to improve the instructions per clock cycle.

Part of the reason we dont see higher frequencies has to do with the technology we are building on, because transistors are so small and because of the capacitances built in them they dont perform as well at high frequency which is why higher OCs require higher voltage, it provides more of a difference between 1 and 0 to make it easier to tell apart.

The extreme overclocking efforts use higher leakage chips which put out more heat than a good chip at the same speed and voltage, but allows for higher stable overclocks. If you have a good chip design that runs well at 3GHz, you could add more transistors to improve its performance or you could increase its clock speed, but since increasing the speed likely requires an increase in voltage, choosing the clock speed route rather than adding more transistors will create more heat and will cut back on your overclocking potential.

A Pentium 4 ht 8GHz on Liquid Nitrogen.

As for the barrier, I don't think we are. Most of the current CPUs can easily be clocked at 4GHz stock. Look at the Core 2 Duo E8600. Most people who overclocked them were hitting 4.5GHz on air.

We will see 10GHz some day but right now it is about improving the CPUs IPC.

Right now, people are focusing a lot more on more cores and more efficient architectures. As the manufacturing process goes down, stock speeds will go up, as will oc capability likely. Also, CPUS are generally ahead of the game right now. A lot of games are GPU bottlenecked, normal activities are vastly exceeded by most modern computers, and only the most intensive video editing strains cpus. What is happening though is a merger between CPU and GPU, as NV always bragged how their GPU was 13x faster than the fastest Intel CPU, or something like that. CUDA and IGP are putting more and more stuff on that small little chip, or using it more in tangent with the GPU, and therefore I suspect we wont really see a vast GHz increase, atleast until more of a new technology replace CPU's.