Asus Launched Grace Blackwell Pro Workstation

The ExpertCenter Pro ET900N G3 brings 72 ARM cores and 748GB of memory to edge AI workflows.

Updated on Sept. 27, 2026 in Semiconductors

Isometric editorial illustration of professional workstation internal hardware components, including cooling plates and optical modules, rendered in muted matte colors.
Asus released the ExpertCenter Pro ET900N G3, a workstation featuring the Nvidia GB300 Grace Blackwell Ultra Superchip for high-performance AI edge computing. AI Illustration. Upload story photo >

Live Poll

Would you pay over $100,000 for a desktop computer capable of running AI models at home?

Asus has released the ExpertCenter Pro ET900N G3, a professional workstation powered by the Nvidia GB300 Grace Blackwell Ultra Superchip. The system is designed to handle high-performance computing tasks with significant coherent memory capacity.

Why it matters

This workstation configuration targets compute-intensive AI workloads that require high-bandwidth memory access and massive core counts. It reflects the increasing demand for specialized hardware capable of sustaining complex model inference at the edge.

The system features 72 ARM cores paired with 748GB of coherent memory and two 400-gigabit network ports. It is powered by a 1,600-watt, 80 PLUS Titanium-rated power supply to handle the thermal and electrical demands of the Grace Blackwell Ultra architecture.

The players

Asus

A global hardware manufacturer known for its PC components, motherboards, and specialized enterprise workstations.

Nvidia

A semiconductor company that designs high-performance graphics and AI compute platforms, including the Grace Blackwell architecture.

The details

The workstation utilizes a sealed liquid cooling loop with copper cold plates to manage the thermal output of the processing architecture. To prevent throttling during high-speed data transfer, an internal fan directs dedicated airflow onto the network card's optical transceivers—the modules that convert electrical signals into light for fiber-optic networking. Data storage is handled by two 2TB NVMe drives configured in a mirrored array to provide redundancy.

Timeline

  1. September 27, 2026: Official publication of the workstation specifications.

The Tech Race

This development follows the rollout of the Nvidia GB300 Grace Blackwell Ultra Superchip, which has become the focal point for high-performance enterprise hardware. The workstation platform directly competes with other specialized inference servers designed for heavy AI model deployment.

The workstation is expected to ship with an estimated price range between $100,000 and $120,000. It is intended for enterprise environments requiring high-throughput processing and will likely be unavailable through standard consumer retail channels.

The takeaway

The ExpertCenter Pro ET900N G3 represents a significant push to bring data-center-level performance into a 27kg chassis. Readers should track the release of official pricing and availability windows to determine when this hardware enters commercial distribution.

Further reading

For more on the current state of high-performance compute architectures, visit Semiconductors.

Live Poll

Would you pay over $100,000 for a desktop computer capable of running AI models at home?