United States-based OpenAI and U.S. semiconductor company Broadcom have unveiled Jalapeño, OpenAI’s first custom-designed artificial intelligence chip, developed to accelerate inference workloads such as real-time ChatGPT responses. The application-specific integrated circuit (ASIC) marks OpenAI’s most aggressive move yet to reduce its dependence on Nvidia’s dominant general-purpose GPUs and to move toward a more vertically integrated, full-stack technology strategy. Unlike Nvidia’s chips, which are designed to handle both training and inference for a wide range of AI applications, Jalapeño is architected from the ground up solely to run inference, the real-time computing process that turns user prompts into model outputs. United States industry analysts say the focus on a single task allows Jalapeño to deliver substantially better performance-per-watt than current state-of-the-art hardware, enabling complex language processing with significantly lower energy use than traditional multi-purpose graphics cards. OpenAI and Broadcom completed the journey from blank-slate design to manufacturing tape-out in about nine months, a timeline the companies attribute to deep hardware–software co-development and the use of OpenAI’s own advanced AI models to help automate and optimize elements of the chip design and verification workflow. The launch highlights a broader shift in the technology sector from general-purpose computing hardware toward highly specialized silicon tailored for specific AI workloads.
Prepared by Jonathan Pierce and reviewed by editorial team.
No left-leaning sources found for this story.
No right-leaning sources found for this story.
Comments