OpenAI unveils its first custom chip, built by Broadcom

What happened

OpenAI and Broadcom on Wednesday unveiled Jalapeño, OpenAI's first custom 'Intelligence Processor' optimized for LLM inference. The chip was designed end-to-end in just 9 months with help from OpenAI's own models.

Context and impact

This is a direct strike at Nvidia's AI hardware dominance. According to Broadcom CEO Hock Tan, the accelerator delivers roughly 50% cost savings vs typical AI GPUs. OpenAI plans to deploy at gigawatt scale through partners including Microsoft.

Details

• First chip in a multi-generation joint compute platform
• Designed end-to-end in 9 months — possibly the fastest ASIC cycle ever
• Celestica provides board, rack and system expertise
• Targets initial deployment by end of 2026, gigawatt scale from 2027 onward
• Broadcom stock climbed on the news