OpenAI and Broadcom Launch ‘Jalapeño’ — A Game-Changing Custom AI Chip
In a significant development that marks the end of the AI industry’s near-total reliance on Nvidia GPUs, OpenAI and chipmaker Broadcom (NASDAQ: AVGO) have introduced Jalapeño — OpenAI’s first ever custom-built AI inference processor. This “Intelligence Processor” was uniquely designed from the ground up in an impressive nine-month development cycle, expedited in part by OpenAI’s proprietary AI models.
Jalapeño is a dedicated Application-Specific Integrated Circuit (ASIC) that is exclusively optimized for large language model (LLM) inference — the process of executing AI models in real time to generate user responses. Preliminary engineering samples are already processing machine-learning workloads in the lab at the targeted production frequency and power, including GPT-5.3-Codex-Spark. The companies assert that the performance per watt will significantly surpass the current state-of-the-art hardware.
This chip is the cornerstone of a wider 10-gigawatt strategic collaboration between the two companies. Deployment at gigawatt-scale data centers is set to begin in late 2026, with Microsoft expected to purchase approximately 40% of the initial production. “This is just the beginning of a multi-generation roadmap,” stated Hock Tan, President and CEO of Broadcom. This announcement led to a roughly 1% drop in Nvidia’s shares, while positioning Broadcom as a formidable custom silicon partner for the AI era.
Source: OpenAI – OpenAI and Broadcom unveil LLM-optimized inference chip