Hamza Yalcin, different scholars/books suggest different thresholds. There is however consensus that VIF > 10 will most likely be problematic. Unusually high VIF values will eventually cause SE of regression coefficients to be inflated. Thus, you should keep an eye on the standard errors as well as following the suggested thresholds. Incidentally, PLS-SEM is no different from multiple regression as far as VIF is concerned.