OpenAI and Broadcom debut Jalapeño chip, a custom LLM inference processor

OpenAI and Broadcom on Wednesday unveiled Jalapeño, a custom inference processor co-designed from the ground up for large language model workloads. The chip is OpenAI's first Intelligence Processor and represents a move by the AI company to own more of its compute stack.

The chip was developed from initial design to manufacturing tape-out in just nine months, which the companies claim is likely the fastest ASIC development cycle ever achieved in high-performance semiconductors. Engineering samples are already running ML workloads in the lab at production target frequency and power, including OpenAI's GPT-5.3-Codex-Spark model.

A blank-slate architecture for LLM inference

Unlike general-purpose accelerators adapted from earlier AI workloads, Jalapeño was designed from scratch around OpenAI's deep understanding of LLM fundamentals. The architecture reduces data movement and balances compute, memory, and networking resources to achieve utilization much closer to theoretical peak performance, according to the company. Broadcom's Tomahawk networking silicon is used in large-scale production configurations.

• Celestica, an electronics manufacturing firm, also contributed to bringing the silicon to production.
• Richard Ho, head of OpenAI's hardware program, said the architecture was optimized around kernels, memory movement, networking, and serving patterns that matter most for frontier AI models.
• OpenAI used its own AI models to accelerate parts of the design and optimization process, contributing to the rapid nine-month development cycle.
• Initial testing indicates Jalapeño will deliver performance per watt substantially better than current state-of-the-art, though final metrics have not been published.

Full-stack infrastructure strategy takes shape

Greg Brockman, president and co-founder of OpenAI, said Jalapeño is part of a long-term full-stack strategy to make compute more abundant, resulting in faster, more reliable, and more affordable AI. By designing more of the stack itself, OpenAI aims to serve more intelligence with greater efficiency and push advanced AI toward broader access.

The chip announcement signals OpenAI's intent to reduce reliance on merchant silicon for inference. The company has not disclosed when Jalapeño will be deployed at scale, but a detailed technical report on performance is expected in the coming months. The chip is the first in a planned multi-generation compute platform.