OpenAI and Broadcom to co-develop custom inference chip named Jalapeño

OpenAI and chipmaker Broadcom are developing a custom inference chip codenamed Jalapeño, according to a TechCrunch report on the company’s Equity podcast. The project is designed to reduce OpenAI’s exposure to single-supplier risk in the AI chip market, where Nvidia has been the dominant provider for years.

The collaboration is framed as a hedge rather than a full break from existing suppliers. OpenAI describes the effort as a way to gain more control over hardware tuned to its specific inference workloads, drawing a parallel to Apple’s transition from Intel processors to custom silicon.

Jalapeño joins a growing list of custom chip initiatives among major technology companies seeking to diversify their AI infrastructure. Google, Apple, and SpaceX have also pursued custom silicon to reduce reliance on a single vendor, reflecting broader industry concerns about supply chain concentration.