OpenAI and Broadcom Introduce ‘Jalapeño’: A Groundbreaking Custom AI Chip
In a significant development for the artificial intelligence industry, OpenAI and semiconductor giant Broadcom have officially launched Jalapeño, OpenAI’s inaugural custom-designed AI inference processor. Unveiled on June 24, 2026, this purpose-built application-specific integrated circuit (ASIC) is specifically engineered to manage OpenAI’s extensive language model (LLM) inference workloads. This involves running pre-trained AI models in real-time in response to live user requests.
What sets Jalapeño apart is its rapid development timeline. The chip was designed from scratch in a mere nine months, with OpenAI’s own AI models playing a crucial role in the design process. This achievement is being celebrated as potentially the fastest ASIC development cycle ever for high-performance semiconductors. Early testing results indicate a significantly superior performance-per-watt ratio compared to current leading alternatives, which likely includes Nvidia’s flagship GPUs.
The division of labor sees OpenAI overseeing the underlying chip architecture, Broadcom supplying silicon implementation and networking hardware (including its Tomahawk networking silicon), and Canadian firm Celestica managing board and rack system integration. Current engineering samples are already running workloads including GPT-5.3, Codex, and Spark in the lab.
Broadcom CEO Hock Tan confirmed plans for initial deployment of Jalapeño chips by the end of 2026, with broader scale-up in 2027 and full production ramp in early 2028. This strategic move places OpenAI alongside tech giants Google and Amazon, both of which have developed proprietary AI accelerators to lessen their dependence on Nvidia.
Source: TechCrunch — OpenAI unveils its first custom chip, built by Broadcom