AUSTIN, Texas (Diya TV) — OpenAI has unveiled its first custom-built artificial intelligence chip, Jalapeño, marking a major step in the company’s effort to expand its technology infrastructure and support growing demand for AI services.
The company announced that it had developed the chip in partnership with Broadcom. OpenAI said Jalapeño is designed specifically for inference workloads that power ChatGPT, Codex, its API platform, and future AI agents. As a result, the company aims to improve efficiency while reducing its dependence on third-party hardware providers.
In a post on X, OpenAI said it designed the chip from the ground up to meet the demands of large language models. The company added that building its own processor strengthens its full-stack AI platform, which now spans products, models, and infrastructure.
The announcement comes as competition in the AI industry continues to intensify. Demand for computing power has surged over the past two years, prompting major technology companies to invest heavily in custom silicon. Therefore, OpenAI’s move places it among a growing list of firms seeking greater control over their AI hardware strategy.
Broadcom played a key role in the project. The semiconductor company worked with OpenAI throughout the development process, while Celestica helped industrialize the platform. Together, the companies moved the chip from concept to production-ready hardware in a remarkably short period.
OpenAI said Broadcom President and CEO Hock Tan and President Charlie Kawwas personally delivered a physical sample of Jalapeño to OpenAI CEO Sam Altman and President Greg Brockman. The delivery marked an important milestone for the project as the companies prepare for large-scale deployment.
According to OpenAI, engineers designed Jalapeño using insights gained from years of work on language models, software systems, and AI products. The company said its teams built the chip around the specific needs of modern AI workloads rather than adapting an existing design.
Moreover, OpenAI revealed that engineering samples are already running machine learning tasks in the lab at target production frequency and power levels. The company said the chip is also being tested on workloads associated with GPT-5.3-Codex-Spark. However, engineers are still evaluating final performance results.
Early testing has produced encouraging signs. OpenAI said Jalapeño appears to deliver significantly better performance per watt than current state-of-the-art alternatives. Better energy efficiency could help reduce operating costs while supporting larger AI deployments. Even so, the company plans to release a detailed technical report in the coming months before making broader performance claims.
The development timeline also attracted attention across the semiconductor industry. OpenAI said it completed the journey from initial design to manufacturing tape-out in just nine months. The company described the achievement as the fastest ASIC development cycle ever recorded in advanced high-performance semiconductor manufacturing.
Notably, OpenAI used its own AI models during parts of the design and optimization process. The company said those tools helped accelerate development and improve efficiency throughout the project.
Jalapeño represents only the first step in a broader roadmap. OpenAI and Broadcom are already working on a multi-generation compute platform that will support future AI systems. Initial deployment is expected by the end of 2026, and both companies plan to expand the platform in the years ahead.
Industry observers quickly viewed the announcement as a strategic effort to lessen OpenAI’s reliance on Nvidia, which currently dominates the AI chip market. By developing its own hardware, OpenAI could gain greater flexibility over performance, costs, and future product development.
Investors responded positively to the news. Following the announcement, Broadcom shares rose more than 3%, reflecting confidence in the company’s growing role within the rapidly expanding AI ecosystem.
