OpenAI and Broadcom unveiled Jalapeño on June 24, an artificial intelligence inference processor designed specifically for large language model operations.

Jalapeño represents OpenAI’s first custom-designed silicon accelerator. The chip went from initial design to manufacturing tape-out in just nine months, believed to be the fastest ASIC development cycle in advanced semiconductors.
The processor is architected specifically for inference, the process of running pre-built AI models in response to user queries. The specialized design reduces costs and improves efficiency compared to general-purpose GPUs.
Broadcom handled silicon implementation while OpenAI’s engineering teams drove architectural decisions. The companies used OpenAI’s own AI models to accelerate parts of the design and optimization process.
Jalapeño is the first step in a multi-generation compute platform. OpenAI and Broadcom plan to deploy gigawatt-scale data centers beginning in 2026, with deployments continuing across subsequent years.
The chip represents OpenAI’s push to control its compute destiny. Developing proprietary silicon reduces reliance on Nvidia and improves margins on inference workloads that dominate ChatGPT usage patterns.



