AI for protein folding refers to machine-learning systems chiefly deep neural networks that infer a protein’s three-dimensional structure from its amino-acid sequence and related biological data, producing atomic-level models far faster than traditional experimental techniques. Over the past 10–15 years the field moved from modest, feature-based predictors to end-to-end deep learning: community benchmarks such as CASP documented steady gains until the dramatic leap at CASP14 (2020), when modern AI architectures outperformed conventional methods and proved the approach’s practical value.