OpenAI and Broadcom unveil Jalapeño AI chip

OpenAI and Broadcom have introduced Jalapeño, a new AI accelerator designed specifically for large language model inference. The chip is aimed at improving speed, reliability and accessibility for AI systems, with a focus on better performance per watt than current leading AI chips.

Developed in just nine months, Jalapeño is built to support current and future large language models, including GPT-5.3-Codex-Spark. OpenAI plans to deploy the chip at gigawatt scale in data centers with partners such as Microsoft, while Broadcom is contributing silicon and networking expertise for large-scale production. The launch advances OpenAI’s push to build a full-stack AI platform spanning models, infrastructure and hardware.