OpenAI has introduced its first in-house AI chip — Jalapeño — to speed up ChatGPT
In partnership with Broadcom, OpenAI has unveiled its first specialized artificial intelligence processor called Jalapeño. This move allows the developer of ChatGPT to reduce dependence on external hardware suppliers and build its own full technology stack.
The main purpose of the new chip is to accelerate inference, the process of handling everyday user queries in AI services. Jalapeño belongs to the class of ASIC chips. Unlike Nvidia’s general-purpose GPUs, this new chip is optimized for OpenAI’s specific workloads. This makes production more cost-efficient and performance more effective.
It took the partners nine months to develop the processor, during which OpenAI engineers designed a significant part of the computing system integration themselves. The company’s leadership refers to the device as an “intelligence processor” designed to improve the speed, reliability, and accessibility of neural networks for a wider audience.
On Wednesday, OpenAI received the first physical sample of Jalapeño. The official deployment of the new processors in the company’s data centers is planned for the end of 2026.
ORIENT
