OpenAI has just revealed its first custom chip, Jalapeño, designed to power its large language models. This new chip is an ASIC, meaning it’s specifically designed for AI inference. Unlike GPUs, which are used for training models, Jalapeño will handle the processing of user requests to run agents like Codex or offer responses from ChatGPT. By reducing its reliance on limited Nvidia GPUs, OpenAI is taking a step towards greater control over its AI infrastructure. Why it matters: OpenAI’s new chip is a significant move towards developing more efficient AI systems, which could have far-reaching implications for industries that rely on large language models.