OpenAI and Broadcom launch Jalapeño AI chip for LLMs

OpenAI and Broadcom unveiled Jalapeño, a custom AI chip for LLMs, offering 50% cost savings over GPUs. Mass deployment starts in 2026, with a roadmap for future versions and major data center expansion.

OpenAI and Broadcom have announced Jalapeño, a custom AI chip designed specifically for large language model (LLM) inference tasks, such as those used in ChatGPT and Codex. This ASIC chip is optimized for dedicated workloads and is expected to deliver around 50% cost savings compared to traditional AI GPUs, according to Broadcom. Jalapeño was developed in just nine months and is being manufactured by TSMC. Mass deployment is scheduled to begin in late 2026. The partnership includes a roadmap for future chip generations, with the next version planned for 2028. OpenAI and Broadcom aim to deploy 10 gigawatts of custom AI accelerators across OpenAI's data centers by 2029. The announcement led to a 2% rise in Broadcom's stock and a slight dip in Nvidia's, signaling a strategic shift in the AI hardware landscape.