OpenAI and Broadcom unveil LLM-optimized inference chip

OpenAI and Broadcom announced a collaboration on a specialized inference chip designed to optimize the performance and efficiency of large language model deployments. The chip, referred to as "Jalapeno," is engineered to reduce inference costs and latency for LLM applications. This partnership represents efforts to develop custom hardware that improves the practical deployment of advanced AI models in production environments.

Source: OpenAI | 2026-06-24