Here are several different approaches to projecting the XP requirements beyond level 80.
There are reasonable arguments to be made for each of these fitting approaches, I wouldn't call any one of them more predictive than the others.
The most conservative one (green) fits a quadratic to the log of the XP requirement and projects a total cost at level 100 to be ~600M
The best-fit line of the log (orange) of the XP produces the most extreme projection, at about 3.3B needed.
If we just draw an exponential curve (yellow) through the two endpoints (lvl 10 and 80) it predicts 2.2B needed.
If we best-fit an exponential directly (blue) it highly favors reducing error around level 80, at the cost of terribly inaccurate predictions for the earlier levels. It predicts a lvl 100 cost of about 1.2B.