Let's call the current Bitcoin network 200 Th. So in order to "take over" the network they would need to buy/build at least 200 Th right now.
However, since the network is growing it may be 300 Th before they have their hashing power ready. Meaning at that point they would need at least 300 Th to be 50% of the network.
But there are much easier ways they can attempt to "break" Bitcoin. Simply making it illegal comes to mind. However, if they do that then the price will sky rocket (see drugs, alcohol, other things they have tried to prohibit).
It wouldn't be hard at this point for a government or large organisation to put in a mass order for ASIC chips to build their own devices and start to gain a large amount of hashing power for itself. As the main cost in ASIC production is for the initial investment in the masks used in the photolithography of the wafers. Still at the current rate of expansion of the asic market and more manufacturers looking at smaller node sizes this becomes an increasingly more difficult attack vector.
But still, you'd only want to gain hashing power if you wanted to control block creation and be able to double spent. At this point it would be much simpler to flood the network under a ddos attack or to build many cancer nodes which send invalid transactions out to nodes which then get 'clogged' trying to validate and see if those transactions are real before relaying them. This would be a much easier way of stopping money transfer instead of going after hashing power.
https://en.bitcoin.it/wiki/Weaknesses#Cancer_nodesThis estimates the current number of nodes at 158833, that's a lot to attack especially as you'd probably need more than 1 cancer node for every real node. But this still isn't far out of the reach of any government or very large organisation.
http://getaddr.bitnodes.io/52/
