OpenAI and Broadcom have announced the design of Jalapeño, their first jointly developed AI accelerator chip. The processor targets large language model inference and marks OpenAI's entry into custom silicon production.
The companies revealed the chip on June 24 through an OpenAI blog post. Jalapeño is described as an Intelligence Processor built around OpenAI's vision for LLM inference. The firms had planned their collaboration in October 2025.
OpenAI stated that Jalapeño offers performance per watt substantially better than current state-of-the-art chips. Final performance testing remains underway, with a detailed technical report expected in the coming months. Initial deployment in data centers is slated to begin in late 2026.
The chip was co-developed from initial design to manufacturing tape-out in nine months. OpenAI and Broadcom called the effort the start of a multi-generation compute platform aimed at making advanced AI faster, more reliable, and more accessible.