Enabling AI Supercomputing in Space-Constrained Urban Edge Sites

Scenario

A Tier-1 telecom provider deploying 5.5G networks planned to integrate AI edge inference services—a critical move for their future competitiveness. They needed to deploy supercomputers in multiple urban Edge MEC (Multi-Access Edge Computing) sites to deliver real-time, low-latency AI services. However, these urban locations came with unique spatial and environmental constraints.

Pain points

1. Extremely limited space, with often only 3 to 5 rack units available per site, making it difficult to accommodate traditional large-scale cooling equipment

2. Strict noise regulations, as these sites are often located in commercial or residential areas where operational noise must be low enough not to disturb neighbors

3. Deploying a high-power supercomputer in a minimal footprint while meeting stringent acoustic standards

ASUS solution

In the Liquid Cooling Lab, ASUS presented a breakthrough solution. We demonstrated how a single CDU could efficiently power racks of NVIDIA HGX B300 systems, proving the feasibility of high-density AI computing in a confined space.

Outcome

This demonstration convinced the telecom provider that space and noise were no longer barriers to deploying AI at the edge. ASUS turned the lab results into a "get-it-right-the-first-time" deployment guide, ensuring that a low-noise 5.5G + AI edge inference service could be rolled out quickly and efficiently across urban areas, delivering superior edge AI services to their customers.

FAQ