They'd want to be making returns north of 20% per year just to offset the fees alone if the numbers abalone outlined are correct.
Edit: to elaborate on how I calculated that:
1. We don't know transaction size so I just focused on the 0.05% transaction fee. Obviously if transaction sizes are small, the fixed element of fee is higher as a proportion of the amount held for 1-2 days so the required return to break even is much higher.
2. If the transaction fee is 0.05%, $1 transferred turns into $0.9995
3. Square is sending $1 to the recipient. Therefore to break even, it must turn 0.9995 into $1.00 in the 1-2 days it holds the cash for.
4. To do that in 1 day, it must earn a return that is roughly equal to 20% annualized. So the annual return is (0.9995 x 20) = 0.1995. The daily return is 0.1995/365 = roughly 0.0005.
5. Adding the return of 0.0005 to the $1 brings you back to the $1 that Square sends on to the recipient. So they break even if they are earning 20% a year on the cash they hold before it gets sent on to the recipient.
Note the required return to break even is lower if they hold the cash for 2 days. However I'd guess they don't because one of the banks along the way probably hold it for at least half that time. Also, even if they do hold it for 2 days, you still have to overcome the fixed cost, so that moves the required return back towards my 'north of 20%' figure.
PS - since the Square guys are obviously smart, I'm sure they've done the above math so I'd question whether they really are paying these kinds of fees on each transaction. However if they are, Abalone's dotcom days comment is entirely correct.