OpenAI Unveils Custom Inference Chip Jalapeño Built With Broadcom
Original: OpenAI unveils its first custom chip, built by Broadcom
Why This Matters
Custom AI chips represent a critical shift in AI economics, reducing inference costs and vendor dependence for large model operators.
OpenAI announced its first custom-built inference processor named Jalapeño, developed in collaboration with Broadcom. The chip shows significantly better performance-per-watt than current alternatives and is designed to reduce dependence on Nvidia GPUs for AI inference operations.
OpenAI unveiled its first custom inference processor on June 24, 2026, named Jalapeño and designed in partnership with Broadcom. The chip was specifically engineered for OpenAI's inference systems, with the company's own AI models assisting in development. Early testing shows the processor delivers significantly better performance-per-watt compared to current state-of-the-art alternatives. The partnership was officially announced in October, though chip development plans had been rumored for some time as a strategy to reduce reliance on Nvidia's GPUs. Jalapeño is optimized specifically for inference—running pre-built AI models in response to user commands—with particular emphasis on low operating costs for real-time coding models. OpenAI President Greg Brockman stated the company sought to identify underserved workloads and build solutions to accelerate capabilities. While inference tasks will use Jalapeño, more performance-intensive tasks like pre-training are expected to continue relying on Nvidia hardware. OpenAI emphasized that developing custom chips represents the company operating across the entire technology stack, from chip architecture and memory systems to networking and deployment. This vertical integration allows each layer to be optimized toward making models faster, more reliable, and more affordable for users.