I strongly disagree, but I’m thinking you probably just have a really narrow idea of “most things” for example try coding an agent based model in R, that will be slower than molasses. Or try solving PDEs for each time step in a time series model or write a model with a delay differential equation. Or do an optimization for each week in a 20 year time series… Etc
I worked on a model where after each week of collecting new data we sampled a Bayesian model for a few thousand samples and then ran an optimization the choose a portfolio of bets to maximize an expected outcome at the end of the week. It would have been ridiculously painful to do in R.