Tag
OpenAI and Broadcom have announced Jalapeño, a custom ASIC designed for large language model inference in data centers. The chip promises substantially better performance per watt than current state-of-the-art.