HP Japan announced the "HP ZGX Fury G1n AI Station," a tower workstation specialized for AI development, on October 26. Pricing starts from 28,490,000 yen, with sales beginning on October 27. It is positioned as a high-performance machine capable of delivering data center-class AI processing right at the desk side, targeting R&D sites and office deployments.

Its most significant feature is the inclusion of the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, equipped with 748GB of coherent memory. This configuration supports model inference on the scale of one trillion parameters and makes fine-tuning large models exceeding 100 billion parameters a viable option. Another strength is its ability to facilitate the entire AI development, validation, and deployment process on a single machine, allowing in-house management of data and workflows while minimizing cloud migration costs and latency fluctuations.

In terms of applications, it is designed to support simultaneous inference by multiple users and parallel work by multiple developers. Furthermore, it supports scale-out, allowing performance expansion by linking multiple units, making it easy to operate flexibly as AI workloads increase. On the software front, it offers the "HP ZGX Toolkit," a suite of pre-configured free AI tools, enabling the utilization of open-source-based environments. It also allows for the secure deployment of AI agents in isolated sandbox environments provided by NVIDIA OpenShell.

The chassis supports both desk-side use and rack mounting, making it easy to deploy in offices, labs, or data centers. Windows compatibility is scheduled for December 2026 or later.