ASUS Brings Data-Center AI to Deskside with NVIDIA-Powered Supercomputer

Event summary

  • ASUS unveiled the ExpertCenter Pro ET900N G3 at CES 2026, featuring NVIDIA’s Grace Blackwell Ultra Superchip.
  • The system boasts 775GB of coherent CPU-GPU memory and up to 20 PFLOPS of AI performance.
  • The ET900N G3 is based on NVIDIA DGX Station architecture and is expected to be available in late Q1 2026.
  • The system aims to address the 'last-mile' challenge for AI development by providing data-center-class compute in a deskside form factor.

The big picture

ASUS's move signals a shift towards decentralized AI compute, recognizing the limitations of cloud-only solutions for many workflows. The ET900N G3 addresses a growing need for local, private AI processing, particularly for tasks like LLM fine-tuning and complex simulations where latency and data security are paramount. This trend could reshape the AI infrastructure landscape, potentially reducing reliance on hyperscale cloud providers for certain use cases.

What we're watching

Market Adoption
The success of the ET900N G3 hinges on whether developers and researchers will embrace a deskside solution over cloud-based AI infrastructure, particularly given the potential for security and data privacy benefits.
Competitive Response
Other hardware vendors will likely respond to ASUS’s offering, potentially leading to a price war or a race to integrate similar high-performance AI compute capabilities into smaller form factors.
Software Integration
The effectiveness of the NVIDIA AI software stack integration will be critical; a cumbersome or inefficient workflow could limit adoption despite the hardware’s capabilities.

Related topics

Semiconductors