OpenAI Unveils Jalapeno Chip With Broadcom

OpenAI and Broadcom unveiled "Jalapeño," a custom inference chip designed to run large language models, marking the company's entry into silicon hardware [1]. The move signals a shift toward vertical integration to control the cost and speed of AI compute as demand for inference capacity grows [1].

Chip architecture and performance

The chip was developed from design to production in nine months, a timeline the companies claim is the fastest for high-performance semiconductors [1]. Early testing indicates Jalapeño delivers performance per watt "substantially better than current state-of-the-art" accelerators, though final metrics are pending [1]. Engineering samples are currently running workloads including GPT-5.3-Codex-Spark at target frequency and power in the lab [1]. Unlike general-purpose accelerators adapted for AI, Jalapeño is a blank-slate design built specifically for LLM inference to reduce data movement and balance compute resources [1].

Deployment and strategy

This launch expands OpenAI's platform from software and models down to the physical chip layer [1]. By designing its own infrastructure, OpenAI aims to create a "flywheel" where better efficiency lowers costs, enabling more capable models and cheaper products [1]. The chip is the first in a multi-generation roadmap planned for gigawatt-scale deployment with data center partners, including Microsoft, beginning in 2026 [2]. Broadcom CEO Hock Tan described the collaboration as a commitment to scaling physical infrastructure for the next decade of AI [2].

What to watch

• A detailed technical report on performance is expected in the coming months [1].
• The initial deployment of gigawatt-scale data centers with Microsoft and other partners is slated for 2026 [2].
• Future generations of the compute platform as part of the multi-year roadmap [1].

The 2026 deployment date will test whether OpenAI can actually translate custom silicon into cost savings and performance gains at the massive scale required for its global products.