OpenAI and Broadcom unveil Jalapeño, their first custom AI chip
OpenAI and Broadcom unveiled Jalapeño, OpenAI’s first custom-designed processor, on June 24. The chip is an inference accelerator — built to run already-trained models in response to user requests rather than to train new ones — and OpenAI says it is optimized for the workloads behind its real-time products, including coding assistants.
The companies, whose partnership was announced in October 2025, took the design from concept to manufacturing tape-out in about nine months, which they describe as one of the fastest cycles for a chip of this complexity. OpenAI says it used its own models to accelerate parts of the design work. Early testing reportedly shows performance per watt meaningfully better than current state-of-the-art parts, though final numbers are not yet set and no firm production date was given; initial deployment is targeted for late 2026.
The move follows the pattern set by Google’s TPUs and Amazon’s Trainium: large AI operators designing their own silicon to control cost and supply and to lessen reliance on Nvidia, whose GPUs dominate the market. By owning the chip, data-center, and product layers together, OpenAI is betting that tight hardware-software integration will lower the cost of serving its models at scale.