OpenAI and Broadcom have introduced Jalapeño, a custom-designed AI accelerator chip optimized for large language model inference. This Intelligence Processor aims to improve efficiency and performance for current and future models. Early testing indicates Jalapeño delivers substantially better performance per watt than existing state-of-the-art hardware.
The development process from design to manufacturing tape-out took only nine months. OpenAI reportedly used its own AI models to accelerate the chip's design cycle.
This chip represents the first step in a multi-generation platform for gigawatt-scale deployment. OpenAI plans to roll out the hardware with data center partners including Microsoft starting in 2026. The move establishes a full-stack strategy to control technology from software models down to the silicon.