The HP ZGX Nano AI Station G1n is part of a new category of AI workstations, released in late 2025. The ZGX Nano AI Station goes beyond traditional desktops, serving as a personal AI supercomputer optimised for AI agents, reasoning AI and edge development workloads that previously required cloud-scale infrastructure.
Before the ZGX Nano and other personal AI supercomputers, professional users and data scientists typically relied on more general-purpose HP Z Workstations, such as the Z2 Mini or larger Z8 models, or used cloud-based GPUs for large-scale AI model training and inference.
Enter the AI era
AI is transforming our workflows and HP is responding, moving to the ‘edge’ and reducing our reliance on the cloud for faster, more private processing with the HP ZGX Nano AI Station G1n. We are demanding smarter, personalised experiences and automated tasks, and tasking our workstations with advanced graphics and data analysis, and this shift requires an advanced AI workstation to deliver. The power harnessed on the device enables faster, real-time analysis, which is crucial for processing AI and for streaming needs.
Accelerated by NVIDIA
The HP ZGX Nano AI Station G1n is powered by an NVIDIA GB10 Grace Blackwell Superchip combining a 20-core ARM processor with a Blackwell GPU, capable of delivering up to 1,000 TOPS of AI performance for FP4 workloads. HP likes to preface the NVIDIA chip as giving the ZGX Nano “Powerful AI Compute” capabilities.
The NVIDIA GB10 Grace Blackwell Superchip is extremely powerful for its size, delivering up to 1 Petaflop of AI performance (at FP4 precision) in a compact, power-efficient system, enabling users to run large language models (LLMs) up to 200 billion parameters locally, making high-end AI development accessible on a desk. AI models with such a size are normally found in larger servers or cloud instances, so HP has come up with something special in a compact size.
FP4 comes up quite a bit now, so it is worth stating it stands for FP4 – 4-bit Floating-Point) – which is a data format and processing methodology used when AI is generated, and what deploying this methodology to process data does is it achieves a speed boost for the workstation and drastically cuts the power and memory needed, and so perhaps a long way of saying that the HP ZGX Nano AI Station G1n has powerful AI compute capabilities.
128 GB of unified LPDDR5x memory
The HP ZGX Nano has a 128GB memory, making it an excellent choice for LLM inference and image generation. Image generation is frequently required in a variety of professions, spanning creative industries, technology development and data analysis – think of all those instant graphics we need to explain the data. These roles now utilise AI to create new content, enhance existing visuals, and develop underlying technologies. The beauty is that users can now keep sensitive data local instead of sending it to cloud servers.
Ultra-compact AI workstation
The HP ZGX Nano AI Station measures 150x150x51mm, which is slightly larger than a typical mini-PC. Despite its size, it runs NVIDIA DGX™ OS, an AI-oriented software stack, enabling simplified prototype and development workflows. Unlike a general-purpose OS, NVIDIA’s DGX OS is optimised for GPU-accelerated AI compute, integrating system management, container orchestration, and AI software frameworks in a unified environment. NVIDIA’s DGX OS deserves its own article, but for those serious about AI in terms of training and deploying models, then it is an essential component based on its compatibility, managing compute resources, scale optimisations and the fact that it is enterprise-ready, and thus it integrates with orchestration tools well.
The HP ZGX Nano AI Station, which includes the HP ZGX Toolkit, equips developers with open-source frameworks, automated model evaluation through Ollama, and streamlined local model serving. It’s essentially the developer toolbox that sits on top of NVIDIA DGX™ OS, giving the user workflow tools that are tailored to HP’s on-device AI hardware.
Last thing to mention about the HP ZGX Nano is that it offers up to 4TB NVMe storage, is built for Wi-Fi 7, SSD capacity of 1TB, has Bluetooth version 5.4, a 10GbE Ethernet, HDMI 2.1, multiple USB-C 20Gbps ports, and two 200Gb/s QSFP+ ports.
Reasons to buy
1 You need serious AI power without relying on the cloud
Cloud GPUs are expensive with usage fees, sometimes unpredictable and often slow to access with low latency and poor data transfer speeds, so if cost, control, privacy, or availability make cloud GPU use difficult then this HP station may be the answer you are looking for.
2 You need to run large AI models locally
Thanks to its massive 128GB unified memory, the HP ZGX Nano offers far beyond what a typical workstation GPU can handle.
3 You work in AI development, robotics, CV, ML ops or agent pipelines
The ZGX Nano is a developer-first device — you get the ZGX AI toolkit, DGX OS stack, and optimised model workflows out of the box.
4 You need secure, private, on-prem AI for sensitive workloads
Sectors that have sensitive data processing requirements include Healthcare, Legal, Finance and Defence. The ZGX Nano keeps all data inside the users’ network.
It’s essentially a mini DGX, built for individuals and small teams who need private, server-grade AI compute on their desks.
