OpenAI unveils AI chip Jalapeno

OpenAI and Broadcom unveiled Jalapeno, a custom AI accelerator chip designed for LLM inference. This purpose-built processor aims to run ChatGPT, Codex, and future agentic products faster and more cheaply than general GPUs from Nvidia or AMD. Developed rapidly using OpenAI's own models, Jalapeno supports current and future industry LLMs. While created for OpenAI's internal workloads and API, Broadcom indicates the chip could potentially be made available to other external AI firms.