OpenAI and Broadcom have unveiled "Jalapeño," a custom accelerator built specifically for large language model inference and the first chip in a multi-generation platform. The chip, OpenAI's first so-called "Intelligence Processor," was designed from scratch for modern LLM inference, with OpenAI handling chip design, Broadcom contributing silicon manufacturing and networking technology, and Celestica managing boards, racks, and system integration. Early self-reported tests showed performance per watt that's "substantially better" than current state-of-the-art hardware, though these numbers haven't been finalized and lack independent verification. Engineering samples are already running ML workloads, including the GPT-5.3-Codex-Spark model. OpenAI says the process from design to tape-out took just nine months. Broadcom CEO Hock Tan stated the first deployment is planned for late 2026 at gigawatt scale, together with Microsoft and other partners, and Broadcom has reportedly demanded that Microsoft guarantee it will buy 40 percent of the chips to secure the first phase.