In a game-changing move for AI development, ASUS has unveiled a new class of desktop AI supercomputers, bringing the power of data center computing to your desk. These next-generation systems—led by the ASUS ExpertCenter Pro ET900N G3 and the ASUS Ascent GX10 mini-PC—are powered by NVIDIA's cutting-edge Grace Blackwell superchips, designed to handle massive AI workloads with ease.

---

Why This Matters: The "Last Mile" of AI

Just like logistics struggle with “last-mile” delivery, AI computing has long faced a similar bottleneck—how to deliver cloud-level performance locally to developers, researchers, and engineers who need it on-site.

ASUS addresses that challenge by putting supercomputer performance into compact, deployable form factors that don’t require an entire server rack or data center setup.

---

Meet the Powerhouses

ASUS ExpertCenter Pro ET900N G3

This high-performance desktop is built around the NVIDIA GB300 Grace Blackwell Ultra Superchip, combining:

• A Blackwell Ultra GPU

• A Grace CPU with 72 Arm cores

• The NVIDIA NVLink-C2C interconnect for ultra-fast data transfer

With up to 784 GB of coherent memory, the ET900N G3 is purpose-built for training large language models (LLMs) and executing AI inference at enterprise-grade speed. It’s the desktop equivalent of the NVIDIA DGX Station, and runs the NVIDIA AI software stack, including DGX OS, optimized for AI, machine learning, and analytics workloads.

---

ASUS Ascent GX10: Compact AI Supercomputing

For environments where desktop-class systems are still too large, the Ascent GX10 offers petaflop-scale performance in a mini-PC form factor.

Powered by the NVIDIA GB10 Grace Blackwell Superchip, it delivers:

• Up to 1,000 AI TOPS of processing power

• 128 GB of unified memory to run models with up to 200 billion parameters

• Expansion via dual-GX10 configurations using NVIDIA ConnectX-7 networking to support larger models like Llama 3.1 (405 billion parameters)

---

Purpose-Built for AI, Not Just Performance

These systems go far beyond high-end gaming PCs or consumer-level workstations. They are designed specifically for AI workflows, offering:

• Massive unified memory pools to reduce model-loading delays

• Grace CPUs engineered for efficient, high-throughput AI computing

• Compact, energy-conscious hardware that makes local AI deployment feasible and scalable

---

A Complete AI Portfolio from ASUS

Whether you’re an AI researcher, developer, or enterprise team, ASUS offers a full range of solutions:

• DIY AI PC builds

• AI-optimized laptops

• Desktop AI supercomputers (ET900N & GX10)

• Rack-mount AI servers

• End-to-end AI data center infrastructure

The ET900N and GX10 mark a significant leap in making AI development more accessible, scalable, and high-performing—all without the need for large physical infrastructure.