Broadcom Inc (NASDAQ:AVGO) has said OpenAI and Broadcom have unveiled Jalapef1o, OpenAI2019s first Intelligence Processor and the first AI accelerator in a multi-generation compute platform they are building together.
The companies said the accelerator is designed around OpenAI2019s vision for the future of large language model inference, with the aim of making advanced AI faster, more reliable and more accessible.
OpenAI said Jalapef1o was delivered to OpenAI chief executive Sam Altman and president Greg Brockman by Broadcom president and chief executive Hock Tan and Semiconductor Solutions president Charlie Kawwas. The announcement marks an important step in OpenAI2019s strategy to build more of the stack behind its models and products.
The chip was designed from scratch around OpenAI2019s understanding of LLM fundamentals and informed by its roadmap of models, kernels, serving systems and product needs. Broadcom and Celestica helped industrialise the platform through chip implementation, board and rack system integration, high-performance networking and scalable production systems.
OpenAI said engineering samples of Jalapef1o are already 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 measured, early testing indicates the chip will deliver performance per watt substantially better than current state-of-the-art.
The companies said Jalapef1o is intended to reduce data movement and balance compute, memory and networking resources more efficiently, with Broadcom2019s silicon implementation and networking technologies, including Tomahawk networking silicon, supporting large-scale production.
Broadcom said the collaboration is the first step in a multi-generation roadmap, with initial deployment planned by the end of 2026 and expansion expected in the years ahead. It added that the platform is being developed for gigawatt-scale data centres with Microsoft and other partners beginning in 2026.