OpenAI has announced a new custom chip called Jalapeño, developed in partnership with Broadcom. The chip is purpose-built for large language model inference, marking a shift toward specialized hardware for running AI models rather than training them.
According to the announcement, Jalapeño is intended to improve performance, efficiency, and scale across AI systems. The collaboration with Broadcom suggests a focus on production-grade deployment, though the source provides no technical specifications or availability timeline.
This is a single-source announcement, so there are no independent assessments to compare. The stated goals are clear, but details on how the chip achieves these improvements remain undisclosed.