Gigabyte unveils its Nvidia Spark-powered AI workstation - AI TOP AERO combines Blackwell architecture GPU and up to 128 GB memory, if you can get one

Three days after its stated launch, the one-liter AI TOP AERO still has no price, no regional configurations and no product page.

Gigabyte has announced the AI TOP AERO, a roughly one-liter desktop built on Nvidia's RTX Spark platform that pairs a 20-core Arm-based Grace CPU with a Blackwell architecture RTX GPU, up to 128GB of unified LPDDR5X memory and up to a 4TB PCIe Gen5 SSD.

The company is pitching the machine as a Personal AI Studio for local AI development, content creation, and gaming on Windows 11. However the announcement still lacks a price, a list of configurations, or anywhere to place an order, even as its closing line tells buyers to "refer to local GIGABYTE announcements for regional product configurations, pricing and availability".

As of November 13, Gigabyte's own AI TOP PC listing showed only the AI TOP 100 B850 and the AI TOP ATOM.

Two different kinds of Sparks on one shelf

Gigabyte now sells two kinds of Nvidia's Spark platform-based hardware. The AI TOP ATOM is a DGX Spark system built on Nvidia's GB10 chip. It runs the Linux-based DGX OS and, according to Gigabyte's October 2026 announcement of a 64GB version, can be clustered up to four units deep over built-in ConnectX-7 networking.

The AERO is an RTX Spark machine built on the N1X Superchip. It runs Windows 11 and lists USB Type-C, 10GbE, and Wi-Fi 7, with no mention of clustering hardware.

Under the lid, however, the two are close relatives: Nvidia's specification table lists the desktop N1X with 6,144 CUDA cores, 20 CPU cores, and a 140W TDP, and The Register describes RTX Spark systems as built on the same GB10 silicon as DGX Spark. Gigabyte's release, however, states none of those figures, nor the AERO's dimensions, weight, or port count.

The argument for the AERO is that Windows makes this hardware useful to more people. Jensen Huang, then Nvidia's chief executive, set that tone at Computex in May 2026, declaring, "This is the new PC. The personal AI computer," in a release naming Acer and Gigabyte as vendors that would follow the first six.

Gigabyte's version of the pitch leans on LM Studio, Ollama, ComfyUI and Nvidia's Nemotron models, plus Nvidia PAIR, a free, open-source router introduced in beta in September that spreads independent inference requests across the PCs on a local network.

A launch date with no availability and no price?

Finding a unit available online is just as hard as finding the price tag for an AI workhorse that already finds most of its competition sold out.

Gigabyte's silence on cost comes at a time when memory sets the price. Nvidia has raised its own 128GB DGX Spark to $6,950, almost three-quarters more than it cost at launch, and The Register judged at the time that the pricing "doesn't bode well" for RTX Spark notebooks and mini PCs. Gigabyte's 64GB AI TOP ATOM, the nearest reference point in its own range, starts at $4,999 with a 2TB SSD in tow.

The first RTX Spark desktops to reach checkout are no cheaper:

• MSI's 1.3-liter EdgeMesa N AI+ went on sale in the US this week at $4,699 with 64GB and 1TB, and $7,999 with 128GB and 4TB, the latter out of stock when Notebookcheck checked on November 11. That is a $3,300 step for the extra memory and storage for a platform that is already seemingly prohibitively expensive for end users.
• Microsoft lists its 128GB Surface RTX Spark Dev Box at $5,999 with shipping due this month, and the 128GB Surface Laptop Ultra has had lead times stretch to late December within days of its launch.
• Buyers who value capacity over CUDA have another route. GMKtec's EVO-X5 Pro, built on AMD's Gorgon Halo platform, launched from $6,399 with 192GB of memory, up to 160GB of which can be assigned to graphics.

None of the price gouging is likely to ease soon, with one forecast predicting the memory crisis will last at least another three years.

Until Gigabyte's regional offices publish prices and stock, the AI TOP AERO is a spec sheet with a date on it. The hardware is credible, largely because it already ships under other names. The open questions are how well a one-liter chassis cools 140W, how much of that 128GB a Windows workload can really use, and whether the top configuration can be bought at all before the year is out.