OpenAI took a major step toward hardware independence on Wednesday, unveiling Jalapeño — the company’s first custom artificial intelligence chip, developed in partnership with Broadcom. The chip is purpose-built for large language model inference, the compute-intensive process that generates responses from AI models like ChatGPT, and marks a strategic shift away from heavy reliance on Nvidia GPUs.
“Jalapeño is part of our long-term, full-stack infrastructure strategy to make compute more abundant, resulting in AI which is faster, more reliable, more affordable for people and businesses, and can be used to solve more important problems,” OpenAI President Greg Brockman said. “By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI toward broader access.”
The chip is an application-specific integrated circuit (ASIC), less flexible than Nvidia’s general-purpose GPUs but cheaper to produce and highly optimized for specific tasks. OpenAI designed the chip in just nine months and already has early versions running in its labs, including with GPT-5.3-Codex-Spark. The company claims Jalapeño delivers more computing power per watt than current leading AI accelerators, though it has not yet published benchmark results.
Broadcom CEO Hock Tan framed the collaboration as foundational for the next decade of AI infrastructure. “By co-developing our industry-leading silicon directly with OpenAI, we are enabling the deployment of gigawatt-scale data centers with Microsoft and other partners beginning in 2026,” Tan said. Broadcom’s stock rose about 2% on the news.
The unveiling comes after months of reported tensions between OpenAI and Nvidia. Reuters previously detailed that OpenAI sought alternatives amid frustration with the latency of Nvidia’s latest chips on certain inference workloads and was eventually targeting a custom chip to handle roughly 10% of its inference needs. A proposed $100 billion investment by Nvidia in OpenAI was ultimately scaled back to a $30 billion equity stake, with Nvidia CEO Jensen Huang noting the larger figure was “never a commitment.”
Despite the friction, OpenAI CEO Sam Altman stressed the company’s ongoing relationship with Nvidia, posting on X: “We love working with NVIDIA and they make the best AI chips in the world. We hope to be a gigantic customer for a very long time.”
Jalapeño is positioned as the first of a multi-generation compute platform. Alongside Broadcom, OpenAI plans to roll out racks of the chips later this year and expand to support gigawatt-scale infrastructure. The chip is expected to handle a meaningful share of inference tasks once deployed at scale, potentially reducing operating costs and latency for the millions of users of ChatGPT and other OpenAI products.
While the immediate impact is confined to the AI hardware landscape, the announcement could ripple into the crypto sector’s decentralized AI projects, where access to efficient inference hardware is increasingly critical. As custom silicon lowers barriers to powerful AI, AI-focused blockchain networks may eventually benefit from cheaper, faster model serving, though no tokens are directly affected at this stage.
