HP brings data-center AI infrastructure closer to the workplace
Working with Red Hat and NVIDIA, HP presents a platform aimed at running AI inference closer to users and data, with up to 20 petaflops of FP4 performance for HP ZGX Fury.
On September 9, 2026, HP announced an approach with Red Hat and NVIDIA to bring artificial intelligence infrastructure closer to edge and enterprise environments. The aim is to run inference closer to users and where data is created.
The solution combines the HP ZGX Fury system with Red Hat AI Factory with NVIDIA. HP cites up to 20 petaflops of FP4 AI performance for the system and says the initial solution is aimed at enterprise inference scenarios.
This model can reduce the need to send data to a distant cloud for every task, improve response times, and give organizations more control in hybrid-cloud architectures. For users, the key consideration is not only the performance figure but also compatibility with existing data-center and AI software environments.
