OpenAI announces Jalapeño custom inference chip with Broadcom to reduce Nvidia dependency

According to TechCrunch, OpenAI unveiled Jalapeño, a custom ASIC inference processor developed with Broadcom, designed to power current and future large language models and reduce reliance on Nvidia's hardware dominance. The move joins other major technology companies including Google, Apple, and SpaceX in developing custom silicon to hedge against single-supplier risk and gain performance benefits tailored to specific inference workloads.