OpenAI announces first in-house AI chip “Jalapeno”

OpenAI today announced a new chip developed in collaboration with Broadcom. The chip, called Jalapeno, is the company’s first “intelligence processor.” This is an important step in OpenAI’s plan to build a full stack for services.

ChatGPT’s parent company says the chip builds on its vision for the future of LLM and is the first step in a “multi-generational computing platform” that the two companies are jointly building. The chip was designed by OpenAI, so it takes into account the company’s future plans for its products and models. Jalapeno is already running machine learning tasks in OpenAI’s lab.

The chip is intended to perform LLM inference, and OpenAI’s goal is to create something that can provide performance similar to leading AI accelerators while reducing latency. In fact, the company’s early tests suggest that the new chip offers better performance per watt than current cutting-edge options.

Building the infrastructure to power its products could prove very advantageous for OpenAI. Such granular control allows the company to optimize models such as the upcoming GPT-5.6 to be faster and more affordable. It also helps improve training efficiency, allowing OpenAI to develop and power more advanced AI models. Additionally, it can reduce the company’s dependence on chips supplied by third parties such as NVIDIA.

The entire process from initial design to manufacturing tape-out took just nine months, and the company reportedly used its own AI model to handle parts of the process. OpenAI said it plans to introduce Jalapeno by the end of 2026 and will continue to develop more such chips for AI needs in the coming years.

We’ll likely learn more about the new processor when OpenAI releases a technical report in the coming months. If these initial claims prove accurate, OpenAI could lead to even faster AI development.