Overview
Federator.ai Smart Liquid Cooling (SLC) transforms the way modern AI data centers manage thermal efficiency by shifting from static, wasteful cooling to intelligent, workload-aware thermal optimization. Built to support high-density GPU infrastructure, this advanced solution dynamically adjusts liquid flow based on predictive heat modeling, not just simplistic GPU utilization metrics. Integrated seamlessly with Supermicro SCC and other standards-compliant infrastructure controllers, Federator.ai Smart Liquid Cooling enables continuous cooling optimization that conserves energy, preserves hardware lifespan, and accelerates AI performance.
Advanced Technologies Behind
Federator.ai Smart Liquid Cooling is powered by a patented Multi-Layer Correlation Engine and a workload-sensitive control system that tightly couples AI workload telemetry with real-time hardware actuation.
Smarter Cooling with Correlated Insights
The patented AI engine treats every AI workload as a chain of interlocking events across layers, enabling per-workload correlation with resource usage and infrastructure metrics to predict GPU load changes.
Workload-Aware Thermal Forecasting
Proactively predicts GPU rack heat load 30–60 seconds in advance by analyzing GPU board power, utilization trends, Kubernetes job metadata, and ambient conditions.
Intelligent Pump Speed Control for Energy Savings
AI-powered cooling adjusts pump speeds to maintain the temperature difference (ΔT) between inlet and outlet coolant within the ideal 8–15 °C range, cutting pump energy use by up to 70% during low-power periods.
Closed-Loop Integration with Supermicro SCC
Bi-directional observability and control, using Prometheus and SCC API to collect GPU power/temperature/flow metrics and actuate optimized pump/valve responses.
Benefits of Federator.ai Smart Liquid Cooling
Up to 30% Cooling Energy Savings
Adaptive throttling reduces pump energy during low-load phases by aligning flow to predicted heat output, not peak utilization.
Up to 45% Compute Acceleration
By reclaiming thermal headroom, Federator.ai allows higher GPU densities and extends rack capacity without breaching power or thermal limits.
Enhanced PUE and Sustainability
Live PUE improved from 1.20 to 1.18 or lower in large-scale deployments, with up to 7.9 GWh/year in energy savings at the facility level.
Extended Equipment Lifespan
Dynamically adjusts pump speeds instead of running at fixed maximum levels, reducing cavitation risk and extending pump-seal life by 35%.
Supports 250 kW-per-Rack Densities
Enables safe scaling to ultra-high-density racks by dynamically maintaining GPU junction temperatures below throttle limits, ensuring thermal stability as power demands rise.
Elastic to Workload Type
Whether training LLMs or running inference jobs, Federator.ai adapts cooling effort to each workload’s true thermal signature—not just how “busy” the GPU appears.
Figure: How Federator.ai SLC collects metrics and alerts to deliver timely alerts and optimal recommendations to CDU controllers