ASUS rolls out Ascent GX10, a compact AI supercomputer for education
ASUS launched the Ascent GX10, a mini-PC AI supercomputer designed specifically for educational and research institutions. It packs the NVIDIA GB10 Grace Blackwell Superchip—the same chip that powers the NVIDIA DGX Spark—combining a 20-core Arm CPU and a Blackwell GPU linked by NVIDIA NVLink-C2C.
The GX10 delivers up to 1,000 AI TOPS and packs 128GB of unified system memory. That’s big enough to handle massive AI models with up to 200 billion parameters. Its large, coherent memory pool removes bottlenecks caused by slow data transfers, offering smooth training and inference on huge datasets. 
Memory size is critical here. Consumer GPUs lack enough VRAM for big AI workloads. The GX10’s unified memory and local compute power let researchers and educators run large AI models fully on-site—no cloud lag or privacy risk.
Multiple GX10 units can link via NVIDIA ConnectX-7 for even more processing muscle. And with its small, quiet chassis, the system fits comfortably on any desk or lab, making high-performance AI accessible beyond top-tier universities.
The GX10 gives schools control over sensitive data, avoiding costly and risky cloud-based AI platforms. Its ready-to-deploy design suits classrooms, labs, and offices. It’s aimed at faculty and students alike, tightening the link between AI theory and hands-on practice.
Institutions handling interdisciplinary projects—from linguistics to bioinformatics—can now share high-end AI hardware easily on a local level.
For even heavier AI needs, ASUS also introduced the ExpertCenter Pro ET900N G3. This desktop workstation uses the NVIDIA GB300 Grace Blackwell Ultra Superchip, featuring a Blackwell Ultra GPU and up to 784GB of shared memory. It targets larger-scale model training in a workstation form factor.
The Ascent GX10 is set to arrive soon. ASUS will demo their AI hardware at the ISTE Conference and Expo in San Antonio, Texas, from June 30 to July 2, 2025—one of the biggest events in educational technology.
This marks a major step for bringing powerful AI within reach of more researchers, educators, and students without the space or budget for traditional servers or datacenters.
