OpenAI and Broadcom said on June 24 they jointly developed an AI inference server chip called Jalapeno.

It is part of OpenAI's strategy to reduce its reliance on Nvidia chips and secure its own hardware.

According to a report by SiliconANGLE, OpenAI plans to bring its first Jalapeno servers online by year-end and then expand its use of the chip.

Unlike Nvidia's flagship GPUs, which handle both training and inference, Jalapeno is designed only for inference. OpenAI said early tests showed Jalapeno delivered significantly higher performance per watt than "current state-of-the-art" products.

OpenAI did not disclose many details about the chip's design. It said the underlying architecture is designed to reduce data movement.

A Jalapeno-based inference cluster will also use Broadcom's Tomahawk series networking technology. The Tomahawk series is a chip for Ethernet switches and is used for data movement between servers in the same rack as well as between racks. The latest Tomahawk 6 handles up to 1.6 terabits per second of traffic and reduces network bottlenecks with a built-in congestion management engine.

OpenAI is also developing custom server racks equipped with Jalapeno and Broadcom networking gear. It is working with Toronto-based data centre equipment design services company Celestica for the effort.