OpenAI unveiled Jalapeño, a custom AI inference chip co-designed with Broadcom for optimized data movement, marking OpenAI's entry into custom silicon and following integration with LLM accelerators.

OpenAI has unveiled its first custom-designed chip, developed in partnership with Broadcom. Dubbed "Jalapeño," the semiconductor marks the company's initial step toward reducing its dependence on external chip suppliers. Broadcom CEO Hock Tan and president Charlie Kawwas presented the accelerator to OpenAI's Sam Altman and Greg Brockman this week.

The chip represents a "blank-slate design" for inference hardware, optimized specifically for AI workloads. Its architecture is engineered to minimize data movement while balancing compute, memory, and networking resources—critical constraints in AI infrastructure. The project moved from initial design to manufacturing tape-out in just nine months, an accelerated timeline for custom silicon development.

"Jalapeño is part of our long-term full-stack infrastructure strategy to make compute more abundant, resulting in AI which is faster, more reliable, more affordable for people and businesses, and can be used to solve more important problems," Sam Altman said in a statement. "By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI toward broader access."

The partnership, announced in October, reflects OpenAI's broader ambition to own more of its compute infrastructure rather than relying exclusively on commercial solutions. The chip design addresses fundamental challenges in AI inference—the phase where trained models are deployed at scale—by optimizing the flow of data through the system to reduce latency and power consumption.