Cerebras builds wafer-scale computers and cloud services for AI training and inference, using radically larger chips to make intelligent applications respond faster.