OpenAI unveils its first AI processor: Jalapeño

OpenAI has just unveiled a new “intelligence processor” chip for AI servers made in partnership with Broadcom. The chip, called Jalapeño, is designed to run current and future large language models. According to the advertisement Wednesday.

Jalapeño is an application-specific integrated circuit (ASIC), which means it was designed for a specific purpose: AI reasoning. With AI inference, models process a user’s request to run an agent like Codex or provide a response from ChatGPT, while AI training involves a model consuming massive amounts of data to guide its responses.

He – she Only nine months coming After OpenAI revealed that it will partner with Broadcom to manufacture its own chips, helping reduce the company’s reliance on Nvidia’s GPUs, which are in limited supply. in Interview with ReutersBroadcom CEO Hock Tan says it matches the performance of Nvidia’s Blackwell chips and Google’s Tensor processing units. Microsoft, deadand Amazon It is among other AI companies that have also recently launched custom-designed AI chips to power their servers for either training or inference, while still lagging Nvidia chips in overall performance.

OpenAI describes Jalapeño as “the first step in a multi-generational computing platform,” which it expects to deploy by the end of 2026. “While OpenAI is still measuring final performance, early testing shows that Jalapeño will deliver performance per watt that is significantly better than current performance,” OpenAI says.