[OpenAI Claims Self-Developed Jalapeno Inference Chip Outperforms NVIDIA GB300 in Testing, Plans to Deploy Within the Year] According to Bloomberg, OpenAI stated that its self-developed Jalapeno AI chip, created in collaboration with Broadcom, outperformed NVIDIA's GB300 in testing, leading in metrics such as processing power per unit of energy consumption and response speed. The chip consumes approximately 700 watts of power and is primarily designed for AI inference rather than model training. OpenAI plans to begin using it later this year. During public testing, OpenAI utilized one of its smaller open-source models, as well as third-party models DeepSeek and Dark Side of the Moon, with a more pronounced advantage observed on the Kimi model. However, Jalapeno has yet to be tested against NVIDIA's next-generation Vera Rubin chip, which has just begun shipping. OpenAI noted that the second-generation chip is in the late stages of development and is expected to complete tape-out in the coming months, while emphasizing that NVIDIA remains an important chip supplier for the company.