OpenAI and Broadcom have teamed up to create a new chip called Jalapeño, specifically designed for large language model inference in data centers. This chip is the result of a nine-month collaboration, with Broadcom using insights from OpenAI to design the chip from scratch. The goal is to create a more efficient and specialized chip for LLMs, which could lead to significant performance improvements. Early testing shows promising results, but a full technical report is still to come. Why it matters: this development could be a major step forward in making large language models more efficient and accessible, which is crucial for the future of AI research and applications.