OpenAI’s Jalapeño chip matches Nvidia Blackwell — and it drops late 2026
openai revealed jalapeño, an asic inference chip designed with broadcom that matches nvidia’s blackwell performance, aiming to cut gpu reliance and boost power efficiency by late 2026.
openai revealed jalapeño, a custom application-specific integrated circuit designed for ai inference and built in partnership with broadcom. the chip targets current and future large language models and, according to broadcom ceo hock tan, matches the performance of nvidia’s blackwell chips and google’s tpus. early testing indicates it delivers substantially better performance per watt than current state-of-the-art hardware, though final measurements are pending. this marks the first step in openai's multi-generation compute platform expected to deploy by the end of 2026, aligning with similar moves by microsoft, meta, and amazon to mitigate reliance on nvidia.