OpenAI debuts Jalapeño, its first custom AI inference chip developed with Broadcom

OpenAI announced its first custom AI chip, Jalapeño, built in partnership with Broadcom and designed for AI inference workloads. The move signals OpenAI’s effort to secure more dedicated compute capacity as demand for its models and services keeps rising.

Custom silicon could help OpenAI reduce reliance on scarce third-party GPUs and optimize performance for its own model-serving needs. The partnership reflects a broader trend of major AI companies moving deeper into vertical integration to control cost, supply, and product differentiation.