OpenAI and Broadcom unveil Jalapeño chip designed in nine months
- Source
- OpenAI
- Time
- 1:19 PM
- Weight
- 95/100
OpenAI and Broadcom have announced the development of Jalapeño, a custom-designed inference chip specifically optimized for large language models (LLMs). The project was completed from initial design to production tape-out in nine months, a timeline the companies claim was significantly accelerated by the use of OpenAI’s own AI models.
The chip is intended to serve as the foundation for a multi-generation compute platform aimed at making AI workloads faster, more reliable, and more energy-efficient. Architected from the ground up for inference rather than general-purpose computing, Jalapeño is designed to maximize performance per watt by reducing data movement and balancing compute, memory, and networking resources.