Hi Jens.
There are a few things at play here, the first is the JESD51-7 thermal numbers. These are typically much more conservative thermal coefficients than we typically see on our own EVBs, so it’s possible for older parts that only have JESD numbers listed that we can actually find more realistic thermal coefficients that will be more reflective of your application (based on our own Eval Board measurements).
The second is to first understand how much power you can dissipate through your package at your given ambient temp before you go over the recommended junction temp operating condition. For your application you can dissipate (125°C-105°C)/(55°C/W) = 364mW over your package. Your application however would ultimately dissipate closer to 753mW–since this is higher we have to use the lower of the two and you’re only allowed closer to 364mW given a max junction temp of 125. So this means at face value that you wouldn’t be able to supply your full current limit without hitting a thermal limit in the device.
However, this device itself doesn’t go into thermal shutdown until 150°C at the earliest. So if we go back in and plug 150°C into the previous equation it lets us know that we can actually dissipate 818mW across this package and the above application of 5V@2A should be ok.
What’s important to note is for these parts we didn’t characterize the behavior between 125°C and 150°C, which is why we recommend operating with the junction only up to 125°C. The temperatures between these ranges are probably fine, but we only tested typical operation up to 125°C (not counting the thermal shutdown of course).
This is all assuming a very conservative θJA value as well, when it’s probably less than what’s listed in the JESD51-7 section of the datasheet.