HP Unveils ZGX Nano G1n Mini AI Station With NVIDIA GB10 Blackwell Superchip
Rather resembling the HP Z2 Mini G1a that we just reviewed, this system sports strikingly similar hardware inside: a pile of CPU cores, a big fat integrated GPU, an extra-wide memory bus for suitable bandwidth, and 128GB of LPDDR5X memory. The key difference is that rather than x86-64 CPUs and Radeon WGPs, the HP ZGX Nano G1n boasts twenty Arm v8 cores and a whopping 48 Blackwell Shader Modules (SMs), giving it considerably more theoretical compute throughput than the AMD alternative.

The form factor isn't exactly the same, either. Where the Z2 Mini G1a is roughly square in the front and longer than it is tall, the ZGX Nano AI Station is slightly smaller, being wider and flatter than the AMD machine. Connectivity was a minor sore spot for us on the Z2 Mini G1a, but it would be even more of a struggle on the ZGX Nano, as it only has four 20-Gigabit USB Type-C ports, an HDMI connection, an RJ-45 jack for Ethernet, and then a pair of QSFP+ plugs for 200-Gigabit links via NVIDIA ConnectX-7.
That 200-Gigabit connection can be used to link the ZGX Nano G1n to a second ZGX Nano to support larger AI models, seemingly in a similar fashion to the way NVLink enables users to combine two NVIDIA GPUs into what is functionally a single larger unit. However, due to the relatively limited connectivity, HP actually recommends that the ZGX Nano be connected to a network and accessed remotely, as a sort of personal cloud compute server.

We had a chance to talk to Brian Allen at HP's Customer Welcome Center in New York City, who informed us that the HP ZGX Nano G1n will be coming this Fall, so pretty shortly. Pricing will be announced closer to the release, but we would expect it to fall in the same $3,000 range as the other GB10-based systems.