● Covered by 1 source · 1 reportMedium impact1 neutral
HP ZGX Fury AI Station with NVIDIA GB300 Superchip Now Available for Order
Key points
- HP ZGX Fury AI station is now orderable.
- Features NVIDIA GB300 Superchip with 748GB unified memory.
- Provides up to 20 petaFLOPS of FP4 compute.
- Designed for shared inference tasks outside data centers.
Availability and Collaboration
The HP ZGX Fury AI station is now available for order. HP announced this availability alongside a collaboration with Red Hat and NVIDIA to integrate Red Hat AI Factory with NVIDIA onto the system.
Hardware Specifications
The ZGX Fury is built around one NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip, which includes a Blackwell Ultra GPU with 252GB of HBM3e and a 72-core Grace CPU with 496GB of LPDDR5X. This configuration provides 748GB of coherent unified memory, allowing the GPU to directly address CPU memory. The system is capable of up to 20 petaFLOPS of FP4 compute.
The Grace CPU's memory consists of four 128GB SOCAMM modules, delivering 396GB/s. The Grace CPU is soldered to the host processor module. This memory architecture enables trillion-parameter inference and fine-tuning of models in the 100 billion parameter class, with quantization at FP4.
Storage and Networking
For storage, the system includes two embedded M.2 slots connected to the Grace CPU via PCIe 5.0 for the operating system in a software RAID 1 mirror. Two additional M.2 slots, coming off the PCIe switch inside the ConnectX-8 SuperNIC, form a RAID 0 data volume with either 2TB or 4TB of self-encrypting NVMe storage.
Networking is handled by a ConnectX-8 with two QSFP112 ports, each offering 400Gbps, allowing two ZGX Fury systems to link. It also includes a 10GbE RJ-45 for the host and a separate 1GbE RJ-45, Mini-DP, and micro-USB for the BMC. Standard workstation I/O includes multiple USB-A and USB-C ports, audio jacks, and a Kensington slot.
Form Factor and Cooling
The ZGX Fury is offered in a tower chassis that also includes rails for a 5U rack slot. The system utilizes liquid cooling with optimized airflow. An optional NVIDIA RTX PRO GPU can be added to drive monitors, dedicating the Blackwell Ultra GPU to inference tasks.
Target Use Case
HP positions the ZGX Fury as a shared inference box rather than a personal workstation. It is intended for use by departments, on factory floors, or in branch offices that require AI capabilities without needing a full data center infrastructure.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
Reporting from
HP's ZGX Fury AI station, featuring the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip and 748GB of unified memory, is now available for order. This system is positioned as a shared inference box for departments or branch offices, capable of handling trillion-parameter inference and fine-tuning 100 billion parameter models.