OpenAI unveils Jalapeño, Its first AI chip to accelerate inference

Thursday 25 June 2026 - 10:27

OpenAI takes a significant step forward in the development of its technological infrastructure with the introduction of Jalapeño, its first proprietary chip dedicated to artificial intelligence. Developed in collaboration with Broadcom, this innovation marks a strategic evolution for the American company, which now aims to strengthen its mastery of the technology chain, from software to hardware.

Unlike the versatile graphics processors widely used for training and running AI models, Jalapeño relies on an ASIC (Application-Specific Integrated Circuit) architecture, specifically designed for inference operations. This phase corresponds to the moment when a pre-trained model analyzes a request and generates a response, a process that has become central with the democratization of conversational assistants and generative AI applications.

OpenAI's stated goal is twofold: to improve performance while reducing energy consumption and costs associated with intensive computing. Initial internal tests indicate better energy efficiency per watt consumed compared to currently used solutions. The company plans to publish a detailed technical report to document the performance of this new generation of processors.

The development of Jalapeño has been characterized by its speed. OpenAI states that it finalized the design in just nine months, a particularly short timeframe in the semiconductor industry. The company notes that its own AI models were utilized to accelerate certain design and optimization stages of the chip.

Designed to meet the growing needs of large language models, the new chip is not limited to technologies developed by OpenAI. Its architecture has been designed to work with various AI models, paving the way for broader use within modern computing infrastructures. Initial deployments are planned in data centers operated with Microsoft and other technology partners.

For Greg Brockman, president and co-founder of OpenAI, this initiative fits into a long-term strategy aimed at making computing power more accessible. By improving the efficiency of infrastructures, the company aims to provide faster, more reliable, and more affordable AI services for individuals, students, researchers, and businesses.

Beyond technological innovation, Jalapeño signifies a transformation in OpenAI's positioning. Previously primarily recognized for its AI models, the group is now investing in the hardware infrastructure sector, a strategic field dominated by a few major semiconductor players. This evolution reflects the company's desire to strengthen its technological independence while supporting the global growth of AI usage.