OpenAI has unveiled its first custom-built artificial intelligence processor, marking a significant step in the company’s effort to gain greater control over the hardware powering ChatGPT and future AI products.
The chip, called Jalapeño, was developed in partnership with semiconductor giant Broadcom and is designed specifically for AI inference—the process through which AI models generate responses to user queries. OpenAI said the processor will support workloads across ChatGPT, Codex, its API services, and future agentic AI systems.
The announcement represents OpenAI’s latest move to build a more comprehensive AI infrastructure stack as demand for computing power continues to surge. The company has historically relied heavily on Nvidia’s graphics processing units (GPUs) to train and run its models, but the rapid growth of AI has increased competition for access to advanced chips.
READ: Nvidia unveils Cosmos 3 open foundation model for physical AI (June 1, 2026)
According to OpenAI and Broadcom, Jalapeño was designed from scratch around the specific requirements of large language models. The companies said the chip was developed from design to production in just nine months, with OpenAI’s own AI models helping accelerate portions of the engineering process.
Broadcom CEO Hock Tan said the processor is intended to support a broader multi-generation infrastructure roadmap aimed at scaling AI systems more efficiently. OpenAI plans to deploy the chip across its infrastructure beginning later this year. Manufacturing is being handled by Taiwan Semiconductor Manufacturing Company (TSMC), while Canadian technology company Celestica is assisting with system integration.
OpenAI said early testing indicates Jalapeño delivers significantly better performance per watt than current state-of-the-art alternatives, although the company has not yet released detailed benchmark data. Engineering samples are already running AI workloads in OpenAI laboratories, including models related to its GPT and Codex platforms.
READ: Nvidia-powered Windows PCs set for debut (May 31, 2026)
The chip is optimized for inference rather than AI training, meaning OpenAI will likely continue relying on Nvidia and other hardware suppliers for the most computationally intensive model development tasks. However, reducing inference costs could provide substantial savings as millions of users interact with ChatGPT and other AI services daily.
The move places OpenAI alongside major technology companies such as Google, Amazon, Meta, and Microsoft, all of which have pursued custom AI silicon to improve efficiency and reduce dependence on third-party hardware vendors.
OpenAI described Jalapeño as the first processor in a broader family of custom AI chips and part of its long-term strategy to make advanced AI systems faster, more reliable, and more affordable for users worldwide.