OpenAI Takes On Nvidia With New Jalapeno AI Chip Developed Alongside Broadcom: All Details

As the race to build more powerful artificial intelligence systems intensifies, AI companies are now increasingly looking beyond software and investing heavily in custom hardware. Today, the ChatGPT maker has taken a major step in that direction. OpenAI has officially unveiled its first custom AI chip, called Jalapeño, developed in partnership with Broadcom. Keep reading to know what the new chip does, why OpenAI built it and how it could help the company reduce its dependence on Nvidia.According to OpenAI, Jalapeño has been specifically designed to handle AI inference, the process where a model generates responses after receiving a user query. In simple terms, every time someone asks ChatGPT a question, inference is the stage that turns that request

into an answer.The US-based AI giant says the chip has been built to efficiently run large language models and support future generations of AI systems. OpenAI hardware chief Richard Ho told Reuters: "It will be performant on, we think, all kind of future iterations of LLMs."The chip was designed jointly by OpenAI engineers and Broadcom. Broadcom CEO Hock Tan even claimed that Jalapeño performs at a level comparable to Nvidia's Blackwell processors and Google's Tensor Processing Units (TPUs), two of the most prominent AI chips currently available.

The announcement comes as AI companies face growing pressure to secure enough computing power. Demand for AI infrastructure has exploded over the last two years, making access to advanced chips increasingly competitive and expensive. Building custom silicon could give the San Francisco-based AI company greater control over costs while reducing reliance on external suppliers.Reuters reports that OpenAI completed the chip's design in roughly nine months before sending it to Taiwan Semiconductor Manufacturing Company (TSMC) for production. Interestingly, the company said artificial intelligence itself helped accelerate parts of the design process.OpenAI also revealed that sample versions of Jalapeño are already running inside its labs and have successfully operated alongside its GPT-5.3-Codex-Spark model. The company plans to begin deploying the chip before the end of this year.This is only the beginning. OpenAI described Jalapeño as the first step in a broader multi-generation chip strategy. As AI infrastructure becomes just as important as AI models themselves, custom processors like Jalapeño could play a massive role in shaping the next phase of the industry.