Correct, I will reformulate that: if batch #3 arrives with difficulty higher than lets say 35m, they won't even break-even IMO.
Even at 35 million they'd break even in
46 days (4-unit with extremely conservative power costs as well).
Your calculations are completely wrong.
1) I paid 76.46btc for a 3 modules unit, thus you should put $10,000 as "rig cost". The point is to break even in btc, not in usd
2) the crucial point is that you are considering a constant difficulty of 35m during 46 days, which won't happen. Difficulty retargets every 15 days, so after 30 days of receiving your unit diff may well be 70m.
Summing up, you are not calculating the profitability decline during the analyzed timeframe, which is a huge mistake.
Use bitcoinx.com/profit and insert "0.01" in the "yearly profitability decline" field, which means that in 1 year difficulty will grow x100, and repeat your calculations.
As a rule of thumb: currently you should aim to break even in 1 month if you want to make some good long term profit. If you break even in 2 months it's ok, but if you don't is very likely that you won't b/e at all because of the huge grow in diff. we are witnessing.