HP Introduces ZGX Nano G1n AI Workstation with NVIDIA Blackwell GB10 Superchip

HP has officially lifted the curtain on the ZGX Nano G1n, a workstation that redefines what a compact AI-focused machine can achieve. While at first glance its tiny 15 x 15 x 5 cm chassis recalls the earlier HP Z2 Mini G1a, the new model packs an entirely different level of performance thanks to NVIDIA’s Blackwell-based GB10 superchip.

At the heart of the ZGX Nano G1n lies a fusion of NVIDIA’s Grace CPU and Blackwell GPU, linked together via NVLink-C2C.

This connection ensures seamless communication between the two components, enabling a peak of 1000 TOPS of AI compute. The GPU alone houses 6144 CUDA cores and 5th-generation Tensor Cores, pushing forward new levels of efficiency and speed for AI workloads.

Memory architecture is another standout. Instead of separating resources, HP opted for 128 GB of LPDDR5X unified memory, accessible by both CPU and GPU. This unified approach eliminates bottlenecks that have historically slowed AI workstations, allowing developers to push larger models and handle higher throughput without the typical stalls.

What makes the ZGX Nano G1n even more interesting is its scalability. HP has integrated a 200 Gbps QSFP/ConnectX-7 interconnect, meaning two units can be paired together for effectively double the performance. For labs or startups experimenting with rapidly growing AI models, this interconnect offers flexibility without jumping to rack-scale systems.

Connectivity also reflects a forward-thinking design. The workstation provides USB Type-C ports at 20 Gbps, a 10 Gbps Ethernet port, WiFi 7, and Bluetooth 5.4, covering both wired and wireless needs. While display output is limited to a single HDMI port, the specific HDMI version has not yet been clarified by HP.

From a software standpoint, HP is bundling the ZGX Toolkit, alongside full support for NVIDIA’s AI Stack (DGX OS). This pairing is intended to streamline the development process, giving engineers and researchers pre-optimized tools for rapid prototyping and deployment.

However, reactions from early observers have been mixed. Some argue that, despite the impressive performance, the pricing could position it as less attractive than competing mini PCs such as AMD’s Strix Halo or HP’s own Ryzen-based Z2 Mini G1a. Enthusiasts point out that the G1n’s AI horsepower, while strong, might not compare favorably to the raw compute of an RTX 5090 at a much lower cost. Others counter that its ARM-based architecture and unmatched efficiency make it a compelling choice for AI research, particularly when space and power constraints matter more than raw gaming GPU numbers.

HP has confirmed that the ZGX Nano G1n AI Workstation will ship this Autumn, though an exact release date and pricing remain under wraps. For professionals in AI development, this launch underscores the accelerating trend toward smaller yet immensely capable AI hardware that blends portability, scalability, and serious compute muscle.