Data center operators face an escalating cooling crisis driven by the explosive growth of artificial intelligence, high-performance computing, and the physical limits of traditional cooling approaches. As rack power densities surge from 5 kW to over 50 kW—with AI workloads pushing beyond 100 kW per rack—the cooling infrastructure that served the industry for decades has reached its breaking point.
This isn't a future challenge. It's constraining data center deployment today.
Why Traditional Approaches Are Reaching Their Limits
Understanding the Two-Part Challenge
Data center cooling involves two distinct but interconnected systems, each with its own efficiency tradeoffs:
Cooling Delivery to IT Equipment
- Air cooling: Traditional approach using CRAC or CRAH units to cool the data center space, with servers drawing air across components
- Liquid cooling: Direct-to-chip cold plates, rear-door heat exchangers, or immersion cooling that deliver cooling directly to heat-generating components with 50-1,000x better heat transfer efficiency than air
Heat Rejection to the Environment
- Dry air cooling: Air-cooled chillers or dry coolers use fans to reject heat—no water consumption but higher electrical power consumption
- Evaporative cooling: Cooling towers use water evaporation to reject heat—significantly lower electrical consumption (70-90% reduction) but consume 1-5 liters of water per kWh of IT load
Regardless of whether facilities use air or liquid cooling delivery, all data centers require both a method to remove heat from IT equipment and a method to reject that heat to the environment. The challenge is that the most electrically efficient heat rejection method (evaporative cooling towers) consumes substantial water, while water-free dry cooling consumes significantly more electrical power—particularly problematic as ambient temperatures rise.
The Physics Problem
Air-based cooling delivery dominated data center design for good reason: it's simple, well-understood, and required minimal specialized expertise. But physics imposes hard limits. Water and other liquids transfer heat 50 to 1,000 times more efficiently than air, and this efficiency gap becomes critical as power density increases.
When rack power requirements remained below 20 kW, air-based cooling delivery maintained safe operating temperatures reliably. Today's high-performing racks routinely exceed 30 kW, and AI accelerators push single racks past 100 kW. While some air cooling systems can theoretically support these loads, they become inefficient, complex to maintain, and prohibitively expensive to operate at scale.
The Efficiency Gap
The numbers tell the story. Cooling infrastructure represents the largest energy consumer in typical data centers, accounting for approximately 50% of non-IT power consumption. Yet the average Power Usage Effectiveness (PUE) across data centers globally remains above 1.57, while efficient facilities target values below 1.2-1.3.
This efficiency gap isn't just an operational cost—it's a strategic liability. As electricity prices rise and sustainability commitments tighten, the delta between average and best-in-class performance translates directly to competitive disadvantage.
The Sustainability Mandate
Data center operators face a three-way tradeoff between water consumption, electrical efficiency, and capital cost that has no perfect solution:
Water Usage Effectiveness (WUE) has emerged as a critical metric as data centers expand into water-constrained regions. Evaporative cooling towers consume 1-5 liters of water per kWh of IT load, creating both regulatory risk and community opposition. Dry air cooling eliminates water consumption but increases electrical power consumption by 70-90% for the heat rejection system alone, directly worsening PUE and CUE metrics.
Carbon Usage Effectiveness (CUE) metrics expose the carbon intensity of cooling infrastructure. Facilities using dry air cooling to avoid water consumption face substantially higher electrical loads, particularly in hot climates. This creates a perverse dynamic: the most water-conscious choice increases carbon emissions through higher electricity consumption.
Increasingly, operators seek heat rejection solutions that eliminate both water consumption and high electrical power requirements—a combination that traditional cooling tower and dry cooler technologies cannot deliver simultaneously.
The Liquid Cooling Delivery Transition
Liquid cooling technologies for delivering cooling to IT equipment—including direct-to-chip cold plates, rear-door heat exchangers, and immersion cooling—offer up to 3,000 times the heat transfer efficiency of air. Early adopters report that even a 75% transition from air to liquid cooling delivery reduces facility power consumption by 27%, with liquid cooling systems consuming approximately 10% less energy overall.
However, liquid cooling delivery does not eliminate the need for heat rejection. Whether facilities deliver cooling via air handlers or liquid cold plates, the heat ultimately must be rejected to the environment through either dry air cooling or evaporative cooling towers.
Liquid cooling delivery introduces new operational complexities: higher capital costs, specialized maintenance requirements, and the need for purpose-built infrastructure create barriers to adoption, particularly for operators managing mixed workloads or retrofitting existing facilities.
The Heat Reuse Opportunity
A fundamental shift is underway in how the industry thinks about waste heat. Rather than viewing thermal output as a problem to be managed, forward-looking operators recognize it as an untapped resource.
Energy Reuse Effectiveness (ERE) metrics quantify this opportunity, providing credit for thermal energy exported from facilities for beneficial use—district heating, industrial processes, or on-site cooling production. Nordic countries are leading this transition, with Stockholm's Data Parks aiming to heat 10% of the city using data center waste heat by 2035.
Liquid cooling systems, with their higher fluid operating and return temperatures (104°F to 107°F), dramatically improve heat recovery potential compared to traditional air-cooled infrastructure. This creates a pathway not just to reduce cooling costs, but to transform data centers from pure energy consumers into energy suppliers.
ASHRAE Temperature Guidelines and Free Cooling Potential
The American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) recommends maintaining data center temperatures between 18°C and 27°C (64.4°F to 80.6°F), with updated thermal guidelines accommodating higher inlet temperatures aligned with W32-W45 classifications.
These expanded temperature ranges unlock significant free cooling potential—using ambient air or evaporative systems to reject heat without mechanical refrigeration. In moderate climates, this can reduce cooling energy consumption by 40-60% annually. However, free cooling effectiveness varies dramatically by geography and remains limited during peak demand periods when cooling loads are highest.
The Path Forward
Modern data center cooling strategies must address two distinct challenges simultaneously:
Cooling Delivery Strategy: Match IT equipment density to appropriate delivery method—air-based cooling for low-to-moderate density (sub-20 kW racks), liquid cooling for high-density AI/HPC workloads (30+ kW racks). This choice primarily affects capital costs and operational complexity.
Heat Rejection Strategy: Select between water-intensive evaporative cooling (lower electrical consumption, 1-5 L/kWh water use), water-free dry air cooling (70-90% higher electrical consumption, zero water use), or alternative approaches that eliminate both water consumption and excessive electrical loads.
The critical insight is that these are separate decisions. A facility can deploy liquid cooling delivery to chips while still using evaporative cooling towers for heat rejection—gaining the efficiency of liquid delivery but maintaining water consumption.
Operators must balance three competing demands: thermal performance adequate for high-density computing, energy efficiency that delivers competitive PUE, and sustainability metrics (both WUE and CUE) that satisfy regulatory requirements and corporate commitments.
The operators who will thrive aren't those who adopt a single technology, but those who architect integrated strategies addressing both cooling delivery and heat rejection while matching their specific density profiles, geographic constraints, water availability, and thermal recovery opportunities.
Key Takeaways
- Data center cooling involves two distinct systems: cooling delivery (air vs. liquid to IT equipment) and heat rejection (dry air cooling vs. evaporative cooling towers)
- All facilities need both systems—liquid cooling delivery still requires heat rejection to the environment via chillers and either dry coolers or cooling towers
- Heat rejection presents a fundamental tradeoff: evaporative cooling towers are 70-90% more electrically efficient but consume 1-5 L/kWh of water; dry air cooling eliminates water use but dramatically increases electrical consumption
- Rack power density has increased 10x in the past decade, driving AI and HPC facilities beyond the practical limits of air-based cooling delivery
- Cooling consumes ~50% of non-IT power in typical data centers, with heat rejection representing a significant portion
- Liquid cooling delivery offers 50-1,000x better heat transfer than air, with 75% adoption reducing facility power by 27%—but doesn't eliminate the heat rejection challenge
- Operators increasingly seek solutions that eliminate both water consumption AND excessive electrical loads—a combination traditional cooling towers and dry coolers cannot deliver simultaneously
- Waste heat recovery via ERE metrics transforms data centers from pure energy consumers to potential suppliers while potentially addressing heat rejection without water or high electrical consumption
- ASHRAE thermal guidelines expansion enables free cooling strategies, though effectiveness varies by climate
Powerdrive Thermal develops waste heat recovery solutions that convert low-grade thermal energy from on-site power generation into mechanical cooling. By using waste heat as the energy source and air-cooled heat rejection, TCCS eliminates both water consumption and the electrical power requirements of traditional heat rejection systems, enabling data centers to improve PUE while achieving zero WUE—addressing the fundamental heat rejection tradeoff that constrains conventional approaches.
— Powerdrive Thermal