HP introduces ZGX Fury GB300, an AI workstation for large-scale models

3 min readUpdated: Jun 6, 2026 01:43 PM IST

HP noted that a large majority of enterprise PCs currently run Windows, creating demand for AI systems that can integrate into existing workflows and IT infrastructure. (Image: HP)

HP has unveiled one of its most powerful AI-focused desktop systems to date – the ZGX Fury GB300 workstation powered by Nvidia’s Grace Blackwell Ultra platform.

Announced during Computex 2026, the new workstation is designed to bring data centre-class AI capabilities to enterprise desktops. HP says the system can support AI workloads involving models with up to one trillion parameters, a task traditionally reserved for large-scale server infrastructure.

At the heart of the ZGX Fury GB300 is Nvidia’s GB300 Grace Blackwell Ultra Desktop Superchip. The platform combines high-performance CPU and GPU resources with a large unified memory pool, enabling AI developers, researchers and enterprises to run and fine-tune large language models locally.

One of the system’s standout specifications is its memory capacity. The workstation can be configured with up to 784GB of unified memory, allowing AI models and datasets to be processed without relying heavily on external storage or cloud infrastructure.

According to Nvidia, systems based on the DGX Station architecture can deliver up to 20 petaflops of FP4 AI performance. This level of compute power is aimed at organisations developing AI agents, running inference on large models, and conducting advanced AI research.

HP’s workstation is based on Nvidia’s recently announced DGX Station platform for Windows. Nvidia positions the DGX Station as a desk-side AI supercomputer capable of delivering performance previously associated with dedicated AI data centres.

The workstation is expected to appeal primarily to enterprise customers who require local AI processing for security, compliance or performance reasons. Running AI models on-premises can help organisations keep sensitive data within their own infrastructure while reducing dependence on cloud-based services.

Story continues below this ad

HP says the system will support Windows environments which is crucial for enterprise customers. The company noted that a large majority of enterprise PCs currently run Windows, creating demand for AI systems that can integrate into existing workflows and IT infrastructure.

While HP has not disclosed official pricing, industry estimates suggest systems based on Nvidia’s DGX Station platform could cost anywhere from around $100,000 to nearly $200,000 depending on configuration. As a result, the workstation is expected to target large enterprises, AI developers, research institutions and specialised professional users rather than mainstream consumers.

HP has not confirmed a launch date, though the ZGX Fury GB300 is expected to become available later in 2026 alongside similar systems from manufacturers including Dell, ASUS, MSI and Supermicro.