New Delhi, June 25(CXO Media): OpenAI has taken a major step toward building its own AI infrastructure.
The company has unveiled Jalapeño, a custom AI inference chip developed with Broadcom, marking OpenAI’s first dedicated processor designed specifically for running large language models (LLMs).
Unlike general-purpose AI accelerators, Jalapeño has been designed from the ground up for inference workloads, the process through which AI models generate responses for users. OpenAI said the chip has been optimized around memory movement, networking, and serving systems used by modern LLMs.
Engineering samples are already running machine learning workloads, including OpenAI’s GPT-5.3-Codex-Spark model. While final benchmarks have not yet been released, the company claims early tests indicate significantly better performance-per-watt compared with current high-end AI hardware.
The chip is the first product in a broader multi-generation computing platform being developed by OpenAI, Broadcom, and Celestica. OpenAI handled the architecture design, while Broadcom contributed silicon implementation and networking technologies, including its Tomahawk networking platform.
The companies aim to deploy the platform in large-scale AI data centers beginning in 2026. The initiative comes as demand for AI computing power continues to surge, pushing technology firms to seek more efficient and cost-effective hardware solutions.
OpenAI said Jalapeño moved from initial design to manufacturing tape-out in just nine months. The company also revealed that its own AI models were used during parts of the chip design and optimization process, highlighting how AI is increasingly being applied to semiconductor development.
The launch reflects a wider trend across the technology industry, where major AI companies are investing in custom silicon to gain greater control over performance, costs, and future infrastructure requirements.