Ceva rolled out the second-generation AI processor architecture scalable from 2 to 100 TOPS and CDNN-Invite API, an interface that allows mapping the customer's AI accelerator into the same computational graph alongside Ceva's own AI engine.