Prevent EPROM powering from IO pin

Why not just do what the datasheet suggests as the Application example for an USB-to-I2C-converter? Which is having the USB-chip powered via USB and the pull-ups via the VCCIO which is driven by the USB-chips 3V3OUT? In that case the pull-ups would not be active when the USB-device is not powered and there should be no problem I think...

This is all fairly rough but based on your numbers and some past experience:

• 40 mA for the GPS receiver is probably a good figure to use for a start, although for most GPS receivers the maximum current is running at the maximum update rate which looks like 5 Hz for that module. You can probably save some power by setting it for a slower update rate even though you might still want the update rate to be higher than the transmission rate to capture maximum speeds for example.

• You'll probably find the GSM module may spend about five seconds in transmit mode per minute or say 12% of it's time at 200 mA which would give an extra 24 mA on average. Maybe round that up and say the GSM module will take 50 mA. GSM modules are always a little trickier because it can also depend on how many retries are required and the transmit power needed to reach the basestation.

• One problem I see is that page you've linked says while running an Arduino Uno the current is 46.5 mA without a load, the I/O current limit 0f 40 mA isn't related to that so you need to add on that current required by the Arduino itself. But the Nano is simpler and after a quick look at the schematic 20 mA would probably be closer the mark.

So I'd estimate as 40 + 50 + 20 = 115 mA but don't forget to take into account any conversion losses if somewhere along the line you're using a switching converter for the power supply. It's probably better for a prototype to add in say a 50% extra margin and then do some actual testing on current draw. If you don't have much in the way of test equipment just using a battery of a fairly known capacity should give you a good idea.