RAM first, CPU second for data science.

Yes, prioritize 32GB RAM over a faster CPU for most data science work. You can’t compute on data that won’t load.

RAM is the hard limit. If your dataset spills into swap, everything grinds to a halt — even the fastest CPU can’t save you. Most data science tasks (pandas, numpy, scikit-learn) are memory-bound, not CPU-bound. You’ll notice the difference between 16GB and 32GB far more than the difference between an i5 and an i7.

Exceptions exist. If you’re training deep learning models on a GPU, the CPU is mostly a glorified waiter — it just feeds data. In that case, RAM matters for dataset size, but CPU speed only matters for data loading bottlenecks. If you’re doing heavy number crunching on the CPU (like monte carlo simulations or complex matrix math), then a faster CPU might win. But that’s not typical data science.

For the common workflow of loading, cleaning, exploring, and modeling medium-sized data (say 5-20GB files), 32GB RAM is the first upgrade. Spend the extra CPU money on an SSD instead. That’ll make more difference than a few hundred MHz.

Future you will thank present you when you don’t have to slice your data into quarters just to fit.