OpenAI and Broadcom Unveil Jalapeño Chip for Large-Scale LLM Inference

OpenAI and Broadcom have announced a new chip named Jalapeño, specifically designed for large language model inference in data centers. The companies plan to deploy the chip at large-scale data centers and emphasize that this is the first generation in a long-term project aimed at continuous refinement. The chip targets the growing computational demands of AI workloads, particularly for models like those powering ChatGPT and Codex.