OpenAI has officially unveiled Jalapeño, its first custom inference processor developed in collaboration with Broadcom to optimize performance-per-watt. This strategic move aims to reduce reliance on external hardware by tailoring silicon specifically for the company’s own machine learning workloads.