OpenAI revealed its first custom artificial intelligence chip named Jalapeño in partnership with Broadcom on June 24, 2026, aimed at accelerating AI inference workloads [1, 2, 3]. The chip focuses exclusively on inference, the task of processing data to generate AI model outputs, rather than training [1, 2, 3].

Jalapeño was designed and developed in about nine months, a notably fast turnaround in chip manufacturing [1, 2, 4]. OpenAI engineers used AI-assisted design techniques to speed parts of the development process, reflecting a growing trend in chip production [1, 5].

The architecture is optimized for frontier AI models, with a focus on kernels, memory movement, networking, and serving patterns critical to large language model (LLM) performance, said Richard Ho, OpenAI Hardware Lead [4]. He added, "It will be performant on, we think, all kind of future iterations of LLMs" [1]. Broadcom CEO Hock Tan said the chip matches the capabilities of Nvidia's Blackwell GPUs and Google's tensor processing units [1].

Early engineering samples are already running test workloads in OpenAI’s labs, including models like GPT-5.3-Codex-Spark [1, 4]. The chip is expected to deliver roughly 50% cost savings compared to typical AI GPUs, while providing substantially better performance per watt than current accelerators [1, 6]. Broadcom supplies the silicon implementation, networking, and connectivity, while Celestica handles board, rack, and system integration [1, 7, 4].

The Jalapeño chip fits within OpenAI’s broader multi-generation compute platform strategy to own more of the AI technology stack from software to hardware. "By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI toward broader access," said OpenAI President Greg Brockman [2].

The first finalized chips will integrate into large data centers from OpenAI partners, including Microsoft, later in 2026 [2, 6, 5]. OpenAI plans to spend tens of billions of dollars on Broadcom chips and related infrastructure as part of its AI expansion efforts [6]. Following the announcement, Broadcom shares rose about 2%, boosted by the company's prominent role in the generative AI boom [2, 6].