OpenAI, Broadcom unveil LLM inference chip Jalapeño, target deployment by end-2026

Jalapeño has been architected around its roadmap for LLM inference and is designed to make advanced AI faster, more reliable and more accessible.

OpenAI and Broadcom have unveiled Jalapeño, OpenAI’s first Intelligence Processor, an inference accelerator designed for large language models (LLMs) and the first AI accelerator in a multi-generation compute platform being developed jointly by the two companies.According to the company, Jalapeño has been architected around its roadmap for LLM inference and is designed to make advanced AI faster, more reliable and more accessible.

The chip has been developed with Broadcom and Celestica, with Broadcom contributing chip implementation, networking technologies and scalable production systems, while Celestica is providing board, rack and system integration.The companies said engineering samples of Jalapeño are running machine learning workloads in the lab at production target frequency and power, including GPT-5.3-Codex-Spark.

While final performance is still being evaluated, OpenAI said early testing indicates the chip will deliver performance per watt substantially better than the current state of the art, with a detailed technical report to be released in the coming months.Jalapeño has been designed specifically for modern LLM inference rather than as a general-purpose AI accelerator.

According to OpenAI, the architecture is informed by the systems it operates across ChatGPT, Codex, its API and future agentic products, while remaining compatible with current and future LLMs across the industry.

The architecture is designed to reduce data movement and balance compute, memory and networking resources to improve utilisation.“Jalapeño is part of our long-term full-stack infrastructure strategy to make compute more abundant, resulting in AI which is faster, more reliable, more affordable for people and businesses, and can be used to solve more important problems.

By designing more of the stack ourselves, we can serve more intelligence with greater efficiency and keep pushing advanced AI toward broader access,” said Greg Brockman, President and Co-Founder of OpenAI.OpenAI said the chip was co-developed from initial design to manufacturing tape-out in nine months, adding that the accelerator programme represents what it believes is the fastest ASIC development cycle achieved in high-performance advanced semiconductors.

The company said OpenAI models were also used to accelerate parts of the chip design and optimisation process.Jalapeño is the first step in a multi-generation compute platform that OpenAI said is targeted for initial deployment by the end of 2026.

The platform will combine OpenAI-designed accelerators with Broadcom’s silicon implementation, networking and connectivity technologies, and Celestica’s board, rack and system expertise.The company said improvements in inference cost, speed and reliability are expected to support faster responses across ChatGPT, Codex and API products, while improving availability during periods of high demand.