Developing new materials often starts with understanding their electronic structure, or how electrons are arranged and behave within them. Calculating these electronic structures typically requires expensive simulations on supercomputers. That's led researchers to train artificial intelligence models on millions of simulated data points to make faster predictions.