You could spend hours digging into each company's product line and determining its potential uses, or just pick the popular stocks that are consensus winners. On the hardware side, data centers filled with thousands of GPUs (graphics processing units) and other computing devices are needed to help process calculations and train AI models. On the software side, data sets are used to train AI models, which can then be integrated into software and sold as an AI product from which the end user benefits.