How the term "R-squared" in VIF(variance inflation factor) is different from normal R-squared calculation?

In normal calculation of R2 , more the value of R2 , it indicates variable represents more variance across the dataset. But in the calculation of VIF (variance inflation factor), higher the value of R2 , more will be the multi-collinearity present in the variables, hence more unstable the dataset is. My question is whether R2 is the same on both cases or will it be different calculation? If different, how? If its the same, why R2 value treated as good and bad in both cases? Can someone explain this for me? Thank you!

It is calculated in the same way. Instead of the response variable $y$ of the original model with numeric regressors $x_1, \dots, x_p$, the response of the VIF-model is $x_i$ and the regressors all other $x_j, j \ne i$.