OpenAI on Wednesday unveiled its first custom-designed artificial intelligence chip, marking a major milestone in the company's effort to gain greater control over the hardware powering ChatGPT and its growing portfolio of AI products.
The chip, named Jalapeño, was developed in partnership with semiconductor giant Broadcom and is designed specifically for AI inference, the process of generating responses to user prompts after a model has already been trained.
OpenAI executives say the move is part of a broader strategy to reduce dependence on third-party suppliers while building what the company describes as a complete AI infrastructure stack.
"The world is moving to a compute-powered economy," OpenAI President Greg Brockman said in a statement accompanying the announcement. He added that designing more of the technology stack internally would help the company deliver AI services more efficiently and expand access to advanced AI systems.
The launch places OpenAI alongside technology giants such as Google, Amazon, Microsoft, and Meta, all of which have invested heavily in custom AI silicon to reduce costs and optimize performance.
OpenAI and Broadcom said Jalapeño was built from the ground up for large language model inference rather than adapted from a general-purpose processor. OpenAI Head of Hardware Richard Ho said the chip was specifically engineered "around the kernels, memory movement, networking, and serving patterns that matter most for frontier AI models."
Early testing suggests the product delivers substantially improved performance per watt compared with existing AI hardware, though detailed benchmarks have not yet been released.
Broadcom Chief Executive Officer Hock Tan told Reuters that the new chip performs at a level comparable to Nvidia's Blackwell processors and Google's Tensor Processing Units, two of the most advanced AI chips currently available.
The announcement comes as AI developers face mounting challenges securing enough computing power to train and run increasingly sophisticated models. OpenAI, like rivals including Anthropic, has spent billions of dollars on infrastructure while competing for access to Nvidia's highly sought-after graphics processing units, which dominate the AI hardware market.
OpenAI's push into chip design has been years in the making. In October 2025, the company announced a strategic collaboration with Broadcom to develop and deploy 10 gigawatts of custom AI accelerators across its data center network. The agreement envisioned OpenAI designing the chips while Broadcom handled development and deployment.
The manufacturing of Jalapeño is being handled by Taiwan Semiconductor Manufacturing Company, while Canada-based Celestica is responsible for system integration. The chip is already operating in OpenAI laboratories and has been tested with GPT-5.3-Codex-Spark, one of the company's latest coding-focused models. OpenAI said artificial intelligence tools were used to accelerate the chip's development, helping engineers complete the design in just nine months.
Despite the introduction of Jalapeño, OpenAI is expected to continue using chips from Nvidia and AMD for many workloads. However, the company says the new processor represents the first step in a multi-generation hardware roadmap aimed at creating a more self-sufficient AI infrastructure platform.
