OpenAI and Broadcom have revealed Jalapeño, their first custom AI accelerator chip designed specifically for running large language models. The chip represents OpenAI's initial move into hardware production, marking a significant shift from relying solely on third-party processors for ChatGPT operations.

According to OpenAI's announcement, Jalapeño delivers performance per watt that substantially exceeds current state-of-the-art technology. However, the company noted that final performance testing remains incomplete, with a detailed technical report expected in upcoming months. The rapid development cycle stands out—Jalapeño went from initial design to manufacturing tape-out in just nine months.

Data center deployment begins in late 2026, though the chip is already demonstrating the benefits of tight software-hardware integration. OpenAI claims controlling more of its technology stack will improve efficiency and expand ChatGPT accessibility for users.

The partnership frames Jalapeño as the foundation for a multi-generation compute platform aimed at making advanced AI faster, more reliable, and accessible to broader audiences. This hardware push positions OpenAI to compete more directly with Nvidia and other chip manufacturers in the AI infrastructure space.