Early results from modeled data center sites in Dallas and Paris suggest that shifting from air chillers to advanced liquid cooling can cut utility dependence while preserving power capacity. The findings, shared this week by project analysts, point to a practical path for operators facing rising energy costs, heat waves, and growing compute demand.
The two modeled sites, chosen for their different climates and grid conditions, were used to test how liquid architectures perform against standard air-based systems. Analysts said the goal was to see whether facilities could reduce their draw from the grid without giving up performance or uptime.
“Modelled sites in Dallas and Paris prove transitioning from air chillers to advanced liquid architectures reduces utility reliance without losing power.”
Why Cooling Is Under Pressure
Cooling has become a major constraint for high-density compute and AI workloads. Traditional air chillers move heat with large volumes of air, which can be energy-intensive, especially in hot or humid regions. Liquid systems bring coolant closer to the heat source, improving heat transfer and reducing the need for energy-hungry chillers and fans.
Dallas offers long, hot summers and frequent grid strain. Paris faces stricter efficiency rules and concerns over peak power draw in dense urban zones. Together, they provide a useful check on how cooling choices react to climate and policy.
Key Takeaways From the Modeling
The project team reported several consistent effects when replacing air chillers with advanced liquid approaches such as direct-to-chip and warm-water loops. While exact savings may vary by site, the modeled pattern supports a move to more targeted cooling.
- Lower utility reliance through reduced chiller and fan energy.
- Stable IT power capacity and performance at high rack densities.
- Improved resilience during peak temperature events.
Engineers familiar with the work said the most visible gains came during high ambient temperatures in Dallas, where air systems typically work hardest. In Paris, the analysis highlighted steadier operations under efficiency constraints.
Benefits, But Not Without Trade-Offs
Operators and vendors agree that liquid cooling can raise efficiency, but they warn about practical hurdles. Retrofitting older halls is complex. It can involve new piping, controls, and leak detection. Training staff to manage liquid systems is also essential.
Procurement and service availability remain issues for multi-country portfolios. Spare parts, compatible server blocks, and installer expertise can vary by region. That affects rollout timelines and risk management.
There are also questions about water use for certain liquid systems and the need for strict monitoring. Some operators prefer sealed or dielectric approaches to limit risk, though those choices can raise costs.
What It Means for AI and High-Density Builds
Rising AI workloads push rack power higher, stressing air-based designs. The modeling suggests liquid can support higher densities without extra utility draw. That could open space for new AI clusters in locations with limited grid headroom.
Colocation providers may see a competitive edge if they can offer liquid-ready space at scale. Cloud firms could standardize on liquid across regions to smooth performance during heat waves and cut operating variance.
How Operators Can Act Now
Experts recommend a phased approach instead of a full rip-and-replace. Target high-density aisles first, then expand as equipment refresh cycles allow. Standardizing connectors, manifolds, and monitoring tools can reduce complexity.
- Start with pilot rows to validate performance and staff processes.
- Align facility upgrades with server refresh windows.
- Track energy, temperature, and uptime metrics before and after changes.
Insurers and regulators will expect strong documentation of safety systems. That includes leak detection, containment plans, and emergency procedures tailored for liquid environments.
The Road Ahead
Demand for cooling efficiency is unlikely to ease. Heat waves and grid constraints put pressure on air systems, and high-density compute continues to climb. The cross-regional modeling for Dallas and Paris adds weight to liquid cooling as a near-term option rather than a distant goal.
Further real-world deployments will test supply chains, service skills, and total cost of ownership. If operators confirm similar results at scale, liquid cooling could become the default for new high-density halls and an important retrofit path for existing sites.
For now, the results offer a clear message: targeted liquid architectures can trim reliance on utilities without sacrificing power. The next test will be how quickly the sector can turn models into operating capacity—and at what cost.
