So that being addressed, you can satisfy the core business needs (mining with hardware for BTC) with math. Anytime hardware arrives and is not need to maintain the hash rate buffer, it can be sold. The price should be set in a way (which again is math against supply and demand) that ensures all surplus units are sold. This would additionally allow a certain level of SLA to be maintained with the channels producing the hardware. My understanding is this hardware (didn't I read it costs around $100 USD to make) can be produced and priced competitively to compete with the other vendors who seem to be coming online.
Anyhow, that's how I would do it.
Edit: grammar and clarity
Everything you've said sounds reasonable, which makes me wonder if that really is the problem, as it seems like friedcat and his team would have taken steps to ameliorate it if it was the issue (OK, you can't just snap your fingers and make a data centre appear, but you still must be able to acquire rack space in a city the size of Shenzhen at short notice).
I've also read that the hardware costs about $10K per THash. AM are supposed to have about 50THash on hand that hasn't yet been deployed.