OpenAI unveils its first custom AI inference chip, Jalapeño, in collaboration with Broadcom — and its development was accelerated using OpenAI's own models
OPENAI AND BROADCOM UNVEIL THE JALAPEÑO AI INFERENCE CHIP
In a significant advancement for artificial intelligence hardware, OpenAI and Broadcom have officially unveiled their first custom AI inference chip, named "Jalapeño." This chip is specifically designed as a purpose-built processor for large language model (LLM) inference, distinguishing itself from the more general-purpose GPUs offered by competitors like Nvidia and AMD. The introduction of Jalapeño marks a pivotal moment in the evolution of AI infrastructure, as it aims to optimize performance for applications such as ChatGPT, Codex, and other future AI products.
Jalapeño is not only a technological innovation but also a strategic move for OpenAI and Broadcom, as they position the chip to potentially serve external AI firms as well. This capability is emphasized in the companies' announcements, which highlight that Jalapeño has been "built from the ground up for current and future LLMs across the industry." This suggests a broader vision for Jalapeño beyond just OpenAI's internal applications, potentially opening new avenues for collaboration within the AI ecosystem.
SPEEDING UP DEVELOPMENT: HOW OPENAI'S MODELS ACCELERATED JALAPEÑO
The development timeline of Jalapeño has been remarkably swift, with the chip moving from early schematics to fabrication readiness in just nine months. This rapid pace is particularly noteworthy in an industry where new processor development cycles are typically measured in years. The accelerated timeline can be attributed to a deep software-hardware co-development process that leveraged OpenAI's own models to enhance various aspects of the chip design.
By integrating their advanced AI models into the design process, OpenAI was able to streamline workflows and optimize performance parameters for the Jalapeño chip. This innovative approach not only reduced the time required for development but also ensured that the chip would meet the specific demands of LLM inference, ultimately resulting in a product that is finely tuned for high-performance AI applications.
JALAPEÑO: A CUSTOM CHIP DESIGNED FOR OPENAI'S LLM INFERENCE
The Jalapeño chip is engineered specifically for the complex workloads associated with large language models. Unlike conventional processors that are designed for a wide range of tasks, Jalapeño is tailored to excel in the unique requirements of AI inference, making it a game-changer for applications that rely on real-time processing of vast amounts of data.
OpenAI's focus on creating a custom chip underscores its commitment to advancing AI technology and enhancing the efficiency of its products. Jalapeño is expected to provide significant improvements in performance, particularly for applications like ChatGPT and Codex, which demand rapid and accurate processing capabilities. As AI continues to evolve, the introduction of specialized hardware like Jalapeño is likely to play a critical role in shaping the future of AI applications.
THE PARTNERSHIP BETWEEN OPENAI AND BROADCOM IN AI HARDWARE INNOVATION
The collaboration between OpenAI and Broadcom represents a strategic alliance aimed at pushing the boundaries of AI hardware innovation. This partnership, which was publicly announced in October 2025, has facilitated a unique synergy between software and hardware development, enabling both companies to leverage their respective strengths in the AI landscape.
By combining OpenAI's expertise in AI model development with Broadcom's capabilities in semiconductor manufacturing, the partnership has created an environment conducive to rapid innovation. The Jalapeño chip is a testament to this collaboration, showcasing how joint efforts can lead to breakthroughs in technology that benefit not only the companies involved but also the broader AI community.
DEPLOYMENT PLANS FOR OPENAI'S JALAPEÑO CHIP IN DATA CENTERS
Following the successful development of the Jalapeño chip, OpenAI has outlined its plans for deployment across active data centers by the end of this year. The company has already commenced testing the chip with one of its prior generation models, GPT‑5.3‑Codex‑Spark, in a production workload environment, albeit within a test setting.
This early testing phase is critical for ensuring that Jalapeño meets the performance expectations set by OpenAI and its partners. As the deployment progresses, the company aims to integrate the chip into its existing infrastructure, enhancing the capabilities of its AI products and services. The successful rollout of Jalapeño could potentially set a new standard for AI inference hardware, paving the way for further advancements in the field.