Here's my guess:
- 9.4 THs
- 1300 watts
- $1100
- Accepting orders on June 15th
- Shipping by June 30th
How did I come up with these wacky numbers? First, I figured they'll try to double the output of the S7, shooting for 10 THs but not making it (just like the S7 didn't make it to 5 THs). This assumes they've either got a 16 - 14 nm process chip, or can pull off some last ditch trick (like ASIC-boost) to squeeze more out of their current process.
The power consumption number is because I think they want to keep the per unit consumption similar to what they have now with the S7. Their goal would be to replace the S7 units in their farm with S9s by swapping boxes, rather than adding more power supplies, using more physical space, and increasing the power to their DC.
The price is based on the trend I've noticed from Bitmain where the ROI is generally in excess of 200 days. If I remember, the S7 was around 240 days at 3% difficulty (which was much lower than reality, as we all know). Anyway, for the S9 I projected out 250 days at 3%, then cut the earnings number in half since it's unlikely that any S9 will be in customers' hands and mining before the halving.
Of course, they could just as easily price batch 1 as double that amount hoping people won't factor in the halving. There will almost certainly be a large number of idiots willing to pay that. But if the initial price really ends up being $2200, I won't be one of those idiots this time.
Regarding my guess for availability, they will want to fully deploy S9s in their DC before the halving (or their mining revenue stream is cut in half) and with enough time to work out any kinks in the manufacturing process, and to physically move the S7s out as part of a "used" sale. So let's say they're shooting to be deployed by mid June at the latest. At that point they'd probably have enough confidence to offer them for sale, with a couple of weeks for delivery, because everyone knows that you can do anything in two weeks
Also, I checked for Chinese national holidays around this time. There's a three day holiday on June 9th for the Dragon Boat Festival. I doubt that three days will make much difference either way, especially if they don't start accepting orders until after the festival.
Maybe one way we can possibly verify the above predictions early will be to see if we can spot a sustained hash rate increase of roughly double Bitmain's estimated mining operation sometime in early to mid June. That should be noticeable enough to be visible, and not just variance.
Hopefully Bitfury and BW have already bulked up for the halving. Because if they haven't, then between them and Bitmain the difficulty might very well explode in June.