← Retour au blog
tech 24 June 2026

OpenAI Unveils Its First Custom Chip, Built by Broadcom

OpenAI takes a significant leap by unveiling Jalapeño, its first custom chip designed with Broadcom to optimize inference systems.

Article inspired by the original source
OpenAI unveils its first custom chip, built by Broadcom ↗ techcrunch.com

Introduction

OpenAI has taken a bold step in optimizing its artificial intelligence systems by introducing Jalapeño, its first custom chip developed in collaboration with Broadcom. This advancement marks a strategic shift for the company, aiming to reduce its reliance on Nvidia's GPUs and enhance its inference capabilities.

Jalapeño: A Tailor-Made Processor

The Jalapeño chip is specifically designed for the unique needs of OpenAI's inference systems. In an announcement on its in-house podcast, OpenAI President Greg Brockman explained that the company sought to optimize specific workloads by building a processor capable of accelerating them efficiently.

Enhanced Performance

According to OpenAI, Jalapeño offers significantly better performance per watt than current alternatives. This energy efficiency is crucial, especially for real-time coding models, where operational cost can be a decisive factor in profitability. By optimizing inference, OpenAI hopes not only to enhance performance but also to cut costs.

The Trend of Custom Chips

OpenAI's initiative is part of a broader trend where giants like Google and Amazon are also developing their own custom chips, often referred to as "AI accelerators." These chips are designed to speed up machine learning workloads, highlighting the growing importance of hardware optimization in the AI economy.

Comparison with Existing Solutions

While Google has its TPU and Amazon its Inferentia, Jalapeño stands out with its focus on inference. Although Nvidia remains an option for more intensive tasks like pre-training, the adoption of Jalapeño could transform OpenAI's infrastructure.

The Future of AI at OpenAI

With Jalapeño, OpenAI is not just enhancing its hardware infrastructure. This innovation could also influence the future development of its agentic products like Codex. Improving inference at every level of the stack could make these products more economically viable and performant.

Economic Implications

By reducing inference costs, OpenAI could reallocate resources towards innovation and expanding its AI capabilities. This could facilitate access to advanced technologies for more businesses and develop more sophisticated AI applications.

Conclusion

OpenAI's introduction of Jalapeño, designed in collaboration with Broadcom, is a milestone in the history of AI system optimization. This advancement could redefine the AI landscape by offering a more economical and performant solution for inference workloads. Let's discuss your project in 15 minutes.

OpenAI Custom Chip Inference AI Accelerator Broadcom
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call