OpenAI and Broadcom have unveiled their first joint custom artificial intelligence chip, named Jalapeño, marking a significant milestone in the ChatGPT maker's infrastructure strategy.
The announcement, made Wednesday, represents OpenAI's inaugural entry into AI silicon design, fulfilling a partnership agreement announced eight months ago.
The chips are manufactured by Broadcom and will be deployed by OpenAI to support its growing compute requirements.
The move underscores OpenAI's ambition to control more of its technology stack, reducing reliance on third-party hardware providers for its core model training and inference workloads.
By co-developing custom silicon with a specialized semiconductor partner like Broadcom, OpenAI aims to optimize performance and efficiency for its specific AI workloads.
This vertical integration is increasingly common among large AI labs seeking to manage the escalating costs and complexity of training next-generation models.