On June 24, OpenAI launched its first custom AI chip developed in partnership with Broadcom, marking a step forward in the ChatGPT maker's efforts to reduce its reliance on Nvidia chips and take greater control over its AI infrastructure.
The chip is named Jalapeño, and it is built only for inference, the stage where a trained AI model actually answers a user's question. Jalapeño is part of OpenAI's broader plan to build the full system behind its models and products, marking its entry into AI silicon technology and going up against rivals such as Google, Amazon, and Groq, who already run their own custom chips.
Check out everything about OpenAI, Jalapeño, the AI chip industry and the Broadcom partnership shaping the future of AI.
What Is OpenAI's Jalapeño Chip?
Jalapeño is a custom application-specific integrated circuit built for specific AI tasks, joining the list that already includes Google's TPUs, Groq's LPUs, Amazon's Trainium, Microsoft's Maia 100 and Tesla's AI5 chips.
Unlike Nvidia GPUs, which handle both training and inference, Jalapeño does one job only. OpenAI says the chip's architecture reduces data movement and balances compute, memory and networking resources to push utilisation closer to the hardware's conceptual peak.
How Broadcom and OpenAI Built It in Nine Months
OpenAI said the chip was designed from end to end in nine months, drawing on its understanding of LLM fundamentals along with its own roadmap of models, kernels, and serving systems.
OpenAI also used its own AI models in the design process, after which Broadcom and Celestica helped industrialise the platform through chip implementation, board and rack system integration, networking, and production systems.
Richard Ho, who once worked on Google's TPUs and now leads OpenAI's hardware program, said the team optimised the chip around the patterns that matter most for frontier models.
What Does Jalapeño Mean for ChatGPT Users and The Future of AI?
Early testing shows Jalapeño delivers performance per watt substantially better than current state-of-the-art chips, though OpenAI is still measuring final figures, and a full technical report is expected in the coming months. If those numbers hold, everyday ChatGPT users could see quicker replies, and AI tools could become cheaper to run at scale.