Note however, that this value is simply calculated by multiplying the last difficulty by the average time per block since the last difficulty change divided by 10 minutes. Since the hashrate is growing steadily, the time per block is decreasing accordingly, which in turn makes the estimation from bitcoincharts.com increase all the time. This is by no means a reliable estimate of the next difficulty since the underlying model is much too simple (it is only accurate when difficulty does not change).
Having a glance at this graph
suggests, that even a linear extrapolation of the global hashrate is probably a good-enough estimate at the moment (what this means for the difficulty is left as an exercise to the reader
). For a more sophisticated analysis have a look at the very fine work of chodpaba