OpenAI Unveils Jalapeno Chip to Lower ChatGPT Operating Costs

• June 24, 2026

OpenAI has revealed its inaugural custom chip, named Jalapeno, aimed at lowering the operational costs associated with running AI models such as ChatGPT. This specialized hardware has been developed in collaboration with semiconductor giant Broadcom and is currently in the testing phase with initial samples already produced.

Targeted Efficiency Gains for AI Inference

The Jalapeno chip is specifically engineered for AI inference tasks, which involve executing pre-trained deep learning models. Unlike general-purpose graphics processing units (GPUs) commonly used for this workload, Jalapeno offers a tailored solution that optimizes computational resources for OpenAI’s suite of products.

According to Broadcom CEO Hock Tan, this chip can potentially reduce resource consumption by up to 50% compared to conventional GPUs. This improvement reflects a significant step toward enhancing efficiency in deploying large language models and other AI services.

By designing dedicated hardware closely aligned with its AI frameworks, OpenAI aims to gain a competitive edge through lowered infrastructure expenses and improved performance. This approach also underscores the broader industry trend of developing customized silicon to meet the growing demand for scalable and cost-effective AI solutions.

Details regarding the chip’s architecture, specifications, or broader availability were not disclosed. However, the announcement marks an important milestone for OpenAI as it expands beyond software innovations into hardware development to support its leading AI applications.

As AI models continue to increase in size and complexity, the need for efficient inference hardware becomes more critical. OpenAI’s collaboration with Broadcom reflects a strategic move to bridge this gap and potentially reshape the AI infrastructure landscape.

OpenAI introduces Jalapeno, its first specialized chip designed to reduce resource use for running AI models like ChatGPT by up to 50 percent.