OpenAI and Broadcom introduced Jalapeño, a custom processor optimized for large language model inference. This hardware supports OpenAI’s strategy to build a full-stack AI infrastructure. The initiative aims to increase hardware control and reduce long-term operational costs.
Broadcom provided the silicon implementation and networking technology for the new platform. The partnership developed the first-generation chip in nine months. Early tests indicate the processor delivers significantly higher performance per watt than current technology.
The collaboration reinforces Broadcom's role as a critical provider of custom silicon for major AI companies. Data centers will begin deploying Jalapeño-based systems in late 2026.