Why would any company spend so much on R&D for the A3 knowing that it will flop?
I don't think they care too much about how successful the miner will be in terms of ROI. They turned $13M worth of miners in 20 minutes... and if they do the same for batch 2, 3, 4, 5 etc, that alone will probably be worth way more than what they spent on R&D.