OpenAI has unveiled Jalapeño, its first custom AI chip designed with Broadcom. The chip is built for inference, which means running AI models after training. That matters because inference powers daily usage, from chatbot answers to coding tools, AI search and enterprise software.

This is not only a Broadcom headline. It signals that AI infrastructure trade may be moving from scarce GPUs toward custom chips, lower power use and more control over the AI stack.

Winners

Custom AI silicon and design partners

Broadcom is the clearest winner because OpenAI chose it as the design partner for Jalapeño. This supports Broadcom’s custom AI accelerator story and shows that major AI companies may want chips designed around their own workloads, not just standard GPUs.

Marvell may benefit from the same theme. This specific c ...