tech 1 min read

OpenAI and Broadcom Unveil Custom Jalapeño Chip for LLM Inference at Scale

OpenAI and Broadcom have collaborated on an ASIC called Jalapeño, optimized for large-scale LLM inference in data centers, promising substantial performance per watt improvements in early testing.

2026-06-25T08:58:30.126Z

OpenAI and Broadcom Unveil Custom Jalapeño Chip for LLM Inference at Scale

OpenAI and Broadcom announced Jalapeño, a custom ASIC built from scratch for large language model inference in data centers.

The chip’s design, informed by OpenAI’s own roadmap and completed in nine months, is the first generation in a long-term project aimed at iterative refinement.

Early testing suggests Jalapeño delivers substantially better performance per watt than existing state-of-the-art systems, though a full technical report is pending.

Topics OpenAI Broadcom Jalapeño LLM inference ASIC performance per watt