OpenAI Unveils Its First Custom AI Chip, Jalapeño

OpenAI has officially unveiled its first custom-designed artificial intelligence accelerator, named Jalapeño, developed in collaboration with semiconductor giant Broadcom. The new processor is purpose-built for large language model (LLM) inference rather than training and is designed to power ChatGPT, Codex, the OpenAI API, and future AI products. The company says it delivers improved performance per watt while optimising compute, memory, and networking efficiency.

Engineering samples of Jalapeño are already running machine learning workloads, including GPT-5.3-Codex-Spark, at production target levels. OpenAI stated that the chip was designed jointly with Broadcom and manufacturing partner Celestica, covering everything from chip architecture to data centre deployment. The company expects Jalapeño to become the foundation of its next-generation AI infrastructure starting next year.