Assume half of AMZN's 2013 net shipping costs of $3,538 occurred in U.S. and profit generated thereon was ~10% of revenues (UPS 2013 operating margin was ~13%), AMZN has a guaranteed (and growing) chunk of change to go after. Not to mention the custom delivery options and probable long-term improved delivery metrics. Of course, this is a somewhat capital intensive business (relative to our estimates of the existing core businesses at least), but still could generate attractive long-term returns.
A much more interesting thought experiment is to do the same calculations on credit card exchange fee savings AMZN would generate if the company processed its payments itself. The savings would be into the billions of dollars and would be incomparably less capital intensive than any physical delivery system.