Freyr AI was pleased to join the panel "100kW Racks and Beyond: Engineering AI Data Centers for Extreme Density," alongside representatives from Equinix and Empyrion Digital. Representing Freyr AI, James SOH, IDC Head, joined fellow industry leaders to share perspectives on the infrastructure and engineering challenges shaping the next generation of high-density AI data centers.
Power, Cooling and Design for Extreme Density
As AI compute density continues to rise, the discussion explored some of the most critical infrastructure considerations for next-generation AI deployments, including:
- Preparing data center infrastructure for 100kW+ rack environments
- Direct-to-chip liquid cooling vs. immersion cooling strategies
- Mechanical, power and floor design for extreme-density AI deployments
- The feasibility of retrofitting existing colocation facilities for AI workloads
Infrastructure Readiness for Next-Generation GPUs
One interesting takeaway from the discussion was the current pace of next-generation GPU infrastructure deployment in South Korea. Fellow panelists noted that NVIDIA GB300 infrastructure has yet to be deployed at scale within their respective environments in the market, highlighting both the early stage of the transition and the infrastructure readiness required for the next generation of AI compute.
Enabling AI Deployment at Scale
As rack densities move beyond 100kW toward increasingly dense architectures, power, cooling, and infrastructure design are becoming fundamental enablers of AI deployment at scale, rather than simply supporting systems.
At Freyr AI, we're excited to contribute to this important industry conversation and help shape infrastructure built for the next generation of AI.