OpenAI, Broadcom unveil Jalapeño chip for AI models

What happened

OpenAI and Broadcom unveiled Jalapeño, OpenAI's first custom chip built from scratch for language model inference. The two companies are building a multi-generation platform together, with Broadcom handling manufacturing and networking, and Celestica managing boards and system integration. The design cycle took nine months, which OpenAI says is the fastest ASIC development cycle for high-performance semiconductors it is aware of.

Why it matters

OpenAI argues that controlling the full technology stack from chip to product allows it to run models faster, more reliably, and at lower cost. This move signals OpenAI's shift from focusing only on models and products into custom hardware—a strategy that may let it reduce dependence on existing chip suppliers and potentially improve the economics of running its AI services at scale.

What to watch

Early tests showed performance per watt that OpenAI claims is "substantially better" than current state-of-the-art hardware, though these are self-reported numbers that have not been independently verified and a technical report is expected to follow. The first deployment is planned for late 2026 at gigawatt scale, together with Microsoft and other partners.