They say they have a 65nm chip. Which would mean BFL.
Either way, their prices are totally fucking insane. If their hashing comes online immediately it might be a good deal, but $350k is a pretty huge investment.
If they already have their units online, they may simply be charging whatever it is they think will make them more profit over self-mining. People are selling Avalon units for $100Btc. If you have chips on hand, it only makes sense to sell them for more then what you'll get if you self-mine over the next few months.
If they did end up getting a boatload of chips from BFL, the pricing starts to make sense...kinda...
Tying into CoinHoarder's post above:
Everyone seems to be forgetting that they host the equipment for no extra costs. This saves a lot of $$$ on HVAC, power costs, and time spent. HVAC/power costs would cost a small fortune for a farm in the range of 5-10Th (especially if these are built on Avalon's power hungry chip design.)
So let's say
they bought BFL chips and didn't get any sort of bulk discount. To get up to 1TH/s, they'd need 250 chips, costing them $18,750 (not including shipping).
With that as a base cost, they'd still need to get the boards, custom casing (presumably), and cooling in place, which all that alone could double their cost. I have no idea how much these sort of things could cost since it's largely space / efficiency oriented, but let's call it $20k.
On top of that, the chips get 3.2 watts per GH/s...3200w for the 1TH/s. Let's shoot for a normal-(ish) $0.15 per kWh price (which would probably be quite high for a datacenter), which would put the cost per month at $345, or $4147 a year.
With that in mind, we're sitting on a cost of $42,897 per terahash per year as it is, not including labor/maintenance, insurance, etc. Throw in a 10% profit margin and I can see how we'd
start to approach the $65,000 mark.
I wonder though...do they still charge you $65k per 1TH/s if you wanted to continue mining for another year? If so, then that's just massive profits for them.