OpenAI, Broadcom unveil Jalapeño chip—first custom AI accelerator built for LLM inference, delivers better performance per watt than current state-of-the-art, developed in nine months, deploying at gigawatt scale with Microsoft and partners starting 2026

OpenAI and Broadcom unveil LLM-optimized inference chip

Early testing shows that the first-generation accelerator will deliver performance per watt substantially better than current state-of-the-artBuilt from the ground up for current and future LLMs across the industryDeveloped from design to production in nine months, accelerated by OpenAI’s modelsExpands OpenAI’s full-stack platform, from products to models and now to chipsTo be deployed at gigawatt scale with data center partners, over multiple generationsOpenAI and Broadcom (NASDAQ: AVGO) today ...