High temperatures, humidity, and water-minimization constraints are reshaping datacenter cooling and power demand

Overview

This report studies how heat, humidity, drought, and water-use minimization requirements affect data center cooling technology choices and then influence PUE, power demand, water consumption, capital expenditure, and related stock performance. The core view is that AI-driven compute-density increases combined with datacenter site environmental constraints shift cooling from standardized infrastructure to site-specific optimization decisions.

Core views

The report presents four core investment impacts: first, cooling technology choices will affect industrial stock performance depending on exposure to environmentally adaptive cooling, controls, and components; second, reducing direct water use is likely to raise overall power consumption, benefiting the power generation, transmission and distribution, and engineering construction supply chain; third, prioritizing water for human and agricultural use creates new opportunities in water efficiency, recycling, desalination, and brine/water-sourcing diversification; fourth, the importance of innovations that lower both water and electricity consumption is rising, benefiting advanced liquid cooling and higher-efficiency server solutions.

Analysis framework

The report breaks data center cooling into two layers: white space and gray space. White space is rack-power-density driven, ranging from CRAC/CRAH air cooling to RDHx/Sidecar and then direct-to-chip or immersion liquid cooling. Gray space is driven by ambient temperature, humidity, water resources, and power constraints, balancing among evaporative systems, dry coolers, air-cooled chillers, economizer systems, and mechanical cooling backup.

Methodology notes

• Data center cooling frameworkWhite space and gray space layering

  White space is responsible for absorbing heat from IT equipment, while gray space is responsible for transferring heat to the external environment and regulating non-IT infrastructure temperatures.

  This framework separates internal cooling choices driven by rack power density from external heat-dissipation choices driven by environmental and resource constraints.

• Resource constraint assessmentPower-water trade-off

  Evaporative systems with lower PUE typically consume more water, while near-zero direct-water-air cooling systems generally have higher PUE.

  The report uses PUE and WUE to compare technology pathways and notes that air cooling can shift direct water pressure into higher electricity demand.

• Investment theme screeningFive beneficiary verticals

  Environmentally adaptive cooling, advanced liquid cooling, water efficiency and supply diversification, thermal equipment and waste-heat utilization, and power generation and transmission/distribution supply chains.

  The report selects related Buy-rated stocks around differentiated cooling pathways and resource constraints.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Environmentally adaptive cooling, controls, and components stocks

  Benefit from differentiated cooling pathways under heat, humidity, and environmental constraints.

  Strengths

  Vertiv (VRT.N), nVent (NVT.N), Parker-Hannifin (PH.N), Comfort Systems (FIX.N), Schneider Electric (SCHN.PA), Envicool (002837.SZ), Carel (CRLI.MI), Jabil (JBL.N), Celestica (CLS.TO), Delta Electronics (2308.TW), Hitachi (6501.T), and similar companies have cooling, control, component, or engineering exposure.

  Weaknesses

  Demand realization depends on datacenter buildout pace, regional exposure, and customer adoption paths.

  Comparison

  Compared with standardized air-cooling suppliers, companies with high ambient-temperature tolerance, precise controls, and liquid-cooling component capabilities benefit more.

  Risks

  The speed of liquid-cooling adoption, capex cycles, supply-chain competition, and customer concentration could affect the benefits.

• Advanced liquid-cooling solutions and components

  Higher rack density and AI training server demand are driving direct-to-chip and immersion liquid-cooling demand.

  Strengths

  Liquid cooling can support higher rack density and partially decouple the power-water trade-off; the report cites AVC (3017.TW), Kstar (002518.SZ), Auras (3324.TW), Celestica (CLS.TO), Jabil (JBL.N), nVent (NVT.N), Vertiv (VRT.N), and others.

  Weaknesses

  Compared with CRAC/CRAH baselines, RDHx capex is about 2x-3x and liquid cooling about 5x-20x.

  Comparison

  In AI datacenters with >50kW/rack, liquid cooling is more necessary than air cooling.

  Risks

  Cost, standardization level, maintenance complexity, and customer deployment cadence remain key uncertainties.

• Water efficiency, reuse, desalination, and supply diversification

  Water scarcity and community constraints increase the value of water-saving technologies.

  Strengths

  Xylem (XYL.N), Veralto (VLTO.N), Legence (LGN.OQ), WaterBridge Infrastructure (WBI.N), Organo (6368.T), and similar companies benefit from water-efficiency, recycling, desalination, or brine/brackish-water supply solutions.

  Weaknesses

  Project-based demand can be influenced by local permitting, infrastructure build cycles, and water pricing mechanisms.

  Comparison

  In water-pressure regions, water solutions are more directly aligned with community and regulatory constraints than cooling equipment alone.

  Risks

  Regulatory approvals, project financing, customer budgets, and techno-economic viability can impact implementation.

• Thermal equipment, transfer, and waste-heat utilization consulting

  Where local regional heating, industrial users, or agricultural demand exists, datacenter waste heat can be captured and reused.

  Strengths

  AECOM (ACM.N), Legence (LGN.OQ), Siemens Energy (ENR1n.DE), Schneider Electric (SCHN.PA), Mitsubishi Electric (6503.T), and others have exposure to thermal equipment, engineering, or consulting.

  Weaknesses

  Waste-heat utilization depends on local heat demand, heat-network infrastructure, and economic fit.

  Comparison

  Compared with simply improving cooling efficiency, waste-heat utilization requires stronger local infrastructure coordination.

  Risks

  Project complexity, local planning constraints, and insufficient heat demand can limit commercialization.

• Power generation, distribution, transmission, and engineering supply chain

  Prioritizing water savings leads to higher power consumption, increasing power infrastructure demand.

  Strengths

  Xcel Energy (XEL.N), Vistra (VST.N), NRG (NRG.N), Quanta Services (PWR.N), Prysmian (PRY.MI), Fujikura (5803.T), Sempra (SRE.N), and similar companies benefit from U.S. datacenter power-demand growth.

  Weaknesses

  Benefits depend on grid interconnection, transmission buildout progress, and regulatory return frameworks.

  Comparison

  In the U.S. market, the report suggests water-priority and datacenter growth make power-demand tailwinds more pronounced.

  Risks

  Grid congestion, permitting delays, interest rates, regulatory limits, and electricity price volatility could weaken returns.

Key data

• Environmental constraint coverage56%/55%The share of incremental global/U.S. datacenter capacity in 2026-2035 expected to be located in regions with high heat, humidity, or drought risk.
• White-space liquid-cooling penetrationabout 43%New incremental datacenter capacity is expected to adopt direct-to-chip or immersion liquid cooling by 2035.
• Gray-space high-constraint cooling demandabout 56%Planned capacity may require air-cooled chillers, economizers, or mechanical cooling backup to handle drought or heat stress.
• Global/U.S. PUE impact+5 percentage points/+11 percentage pointsPotential rise in PUE by 2035 from cooling-technology choices driven by environmental factors.
• Incremental U.S. electricity demand60-70 TWhEstimated increase in power demand by 2035 from U.S. datacenter cooling-driven PUE changes.
• U.S. power demand CAGR impact+0.3 percentage pointsEstimated incremental contribution to U.S. power demand CAGR for 2026-2030.
• Global baseline PUEaround 1.41Under a baseline scenario, global datacenter PUE may rise from about 1.36 currently to about 1.41 by 2035.
• PUE sensitivity rangeabout 1.30-1.47About 1.30 in a high white-space liquid-cooling adoption scenario and about 1.47 in a low-adoption scenario.
• Planned capacity heat-humidity risk50%/46%/87%Share of planned datacenter capacity in hot-humid regions globally in 2035 by region: Global/U.S./APAC.
• Planned capacity water-pressure risk20%/24%/15%Share of planned datacenter capacity in water-pressure regions by 2035: Global/U.S./APAC.

Impact & implications

The investment implication centers on two main themes: first, datacenters adopting air cooling, dry coolers, mechanical cooling, and liquid cooling to reduce water use may increase electricity demand and spur power infrastructure investment; second, high rack density and environmental constraints raise cooling system capex intensity, favoring suppliers with liquid cooling, controls and component, heat-exchange, water treatment, and engineering capabilities.

Risks

• Future cooling-efficiency innovation could lower the PUE increase and electricity-demand increment estimated in the report.
• China's new datacenter PUE target of <=1.25 could suppress APAC average PUE increases, but could also tilt toward more water-intensive cooling pathways.
• Heat, humidity, water pressure, and power constraints vary significantly by region, and no single technology pathway fits all datacenters.
• Liquid cooling and RDHx capex are significantly higher than traditional CRAC/CRAH, and customer adoption pace may be slower than expected.
• Water-use minimization can shift pressure onto the power system, and if grid capacity is insufficient, datacenter commissioning timelines may be constrained.
• Community opposition, permitting constraints, and water-priority policies may affect datacenter siting and construction.

What to watch

• The actual share of incremental global and U.S. datacenter capacity from 2026-2035 located in high heat, humidity, or drought-risk regions.
• White-space direct-to-chip and immersion liquid-cooling penetration, especially in AI training servers and high-density rack deployments.
• Whether U.S. datacenter electricity demand, PUE, and power-demand CAGR align with report estimates.
• Adoption of air-cooled chillers, economizers, dry coolers, and mechanical cooling backup in gray space.
• Order flow and capex for water-efficiency, recycling, desalination, and brackish-water supply projects.
• Grid interconnection, transmission/distribution buildout, and transformer and cable supply-chain delivery capability.
• Policy changes regarding datacenter PUE, water consumption, and siting from regulators.