OpenAI and Broadcom have unveiled Jalapeño, OpenAI's first custom AI chip, built from scratch to run large language models more efficiently. The inference accelerator is the first product in a multi-generation compute platform the two companies are building together, with first deployment in gigawatt-scale data centres planned by the end of 2026.

OpenAI announced the chip on 10 November 2026. It says engineering samples are already running machine learning workloads in its lab at production target frequency and power, including GPT-5.3-Codex-Spark, and that early testing shows performance per watt substantially better than current state-of-the-art accelerators. A detailed technical report on performance is promised in the coming months.

Designed for LLM inference from scratch

Unlike general-purpose accelerators adapted from earlier AI workloads, Jalapeño was designed from the ground up around large language model inference, informed by the systems OpenAI runs every day across ChatGPT, Codex and its API. OpenAI says the architecture reduces data movement and balances compute, memory and networking to achieve utilisation much closer to theoretical peak performance.

Broadcom handled the silicon implementation and networking, including its Tomahawk networking silicon, while Celestica contributed board, rack and system integration. The chip is designed to work with all large language models, not only OpenAI's own, and to suit interactive products at scale, from faster ChatGPT answers to Codex tasks that complete with less waiting.

Nine months from design to tape-out

The companies say Jalapeño went from initial design to manufacturing tape-out in just nine months, helped by OpenAI's own models accelerating parts of the design and optimisation process. OpenAI describes the programme as what it believes to be the fastest ASIC development cycle ever achieved in high-performance advanced semiconductors.

"The world is moving to a compute-powered economy," said Greg Brockman, OpenAI's president and co-founder. "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."

Broadcom chief executive Hock Tan said the collaboration would enable "the deployment of gigawatt scale data centers with Microsoft and other partners beginning in 2026". The move puts OpenAI among the AI companies designing their own silicon to cut inference costs, following similar efforts by Google, Amazon and Meta.

Source links: OpenAI and W.Media.