Bitfury is not spending $600m on hardware. Some fortunetellers estimate $600m will be spent based on how much btc can be mined. They also say "The cost of deploying a petahash ranges from $1.5 million to $2 million" which is way too much. A month ago Hashratios deployment costs were $800k/PH and it will only get cheaper.
http://www.datacenterknowledge.com/archives/2014/07/15/bitcoin-infrastructure-may-grow-600m-second-half-2014/
A 1 terahash/1 KW rig with 0.02 power (which i do not believe exists anywhere) between today and 5/23/2015 will produce $1,080 of BTC @ $600, 16%, 12 days per bump. At 850 Watts, it produces $1,100 worth of BTC.
I don't see the business case for large new farms to go into production with new capital investment with current efficiency. I can see already committed capital investment being put into production. I don't see the business case for a new generation of chips unless they can achieve an order of magnitude improvement in power.
Then labor costs come into play. A one petahash farm at 850W/THS and 0.02 power burns $12K in power per month. That is roughly 1000 servers. Even if labor to service that is only $3,000 - it decreases the time to go negative.