The digitalization of the European economy and the rapid expansion of AI infrastructure are confronting data center owners with a fundamental technical challenge: conventional air cooling has reached its physical limits for modern AI workloads. Liquid cooling – particularly Direct Liquid Cooling (DLC) and Immersion Cooling – has evolved from a niche technology to the industry standard for hyperscale data centers in 2026. For ArcTeron, as an owner and financier of hyperscale infrastructure across Europe, the strategic integration of liquid cooling technologies is a central element of portfolio planning.
Why Air Cooling Is No Longer Sufficient for AI Infrastructure
Just a few years ago, a rack density of 10–15 kW was considered "high density." AI training clusters based on modern GPU architectures (NVIDIA H100, H200, Blackwell) now require 40 to over 100 kW per rack – an increase of 7 to 10 times within just a few years. Air cooling can no longer efficiently dissipate these heat densities: the volume of air required to cool 100 kW per rack would drive airflow velocities and pressure drops in conventional data centers to uneconomical levels.
The conclusion is clear: liquid cooling for AI infrastructure is not an option – it is a necessity. Market analyses show that the share of liquid-cooled capacity in European hyperscale data centers will rise to over 60% by 2030, compared to less than 10% in 2022.
Technology Overview: DLC, Immersion, and Rear-Door Cooling
Direct Liquid Cooling (DLC)
Direct Liquid Cooling is the most widely deployed liquid cooling technology for hyperscale environments. Cooling water is brought directly to the heat sources – CPU, GPU, memory – typically via cold plates mounted directly on the processors. Heat is transferred to the building cooling system via a heat exchanger.
Advantages of DLC:
- PUE values of 1.05 to 1.15 achievable (vs. 1.3–1.5 for air cooling)
- Compatible with existing server architectures (retrofit possible)
- Scalable to rack densities of 40–150 kW
- Waste heat available at temperature levels of 40–60°C
Immersion Cooling
In immersion cooling, servers are fully submerged in a dielectric fluid (single-phase or two-phase). The technology enables the highest cooling capacities and is particularly suited for extreme rack densities (>100 kW).
Advantages of Immersion Cooling:
- PUE values close to 1.03 achievable
- No fans required (noise reduction, energy savings)
- Highest rack densities realizable
- Waste heat available at higher temperature levels (up to 70°C)
Challenges: Higher capital costs, specialized maintenance personnel, limited server compatibility.
Rear-Door Cooling
As a transitional technology, rear-door cooling enables retrofitting of existing racks with water-cooled doors that dissipate waste heat directly at the rack. Suitable for rack densities up to approximately 30 kW.
Regulatory Framework: EnEfG and EU Requirements
German Energy Efficiency Act (EnEfG) 2026
Germany's Energy Efficiency Act places concrete compliance requirements on data center owners that effectively make liquid cooling mandatory:
- PUE ≤ 1.2 for new data centers from July 2026 (mandatory)
- PUE ≤ 1.6 for existing facilities by 2027, PUE ≤ 1.4 by 2030
- Waste heat utilization: At least 10% of energy input as waste heat from July 2026, 15% from 2027, 20% from 2028
These requirements are barely achievable with conventional air cooling for modern AI workloads. Liquid cooling is the technological key to EnEfG compliance.
EU Regulation: Common Rating Scheme and Minimum Performance Standards
The European Commission proposed a common rating scheme for data centers on September 21, 2026, mandating transparent reporting of energy and water consumption. Minimum performance standards are expected by Q2 2027. Data center owners investing in liquid cooling today are positioning themselves regulatorily for the next decade.
Waste Heat Utilization as a Value Creation Factor
A strategically underestimated advantage of liquid cooling is the quality of usable waste heat. While air cooling generates waste heat at low temperature levels (25–35°C), DLC delivers waste heat at 40–60°C – sufficient for district heating networks, industrial processes, and greenhouse heating. Immersion cooling achieves even 60–70°C.
Economic perspective: In Germany, district heating networks pay 5–15 EUR/MWh for waste heat from data centers, depending on temperature level and contract duration. For a 100 MW data center with a 20% waste heat utilization obligation (from 2028) at a temperature level of 50°C, this represents an annual revenue potential of EUR 8–24 million – a significant contribution to the overall return on the infrastructure asset.
Investment Strategy: Liquid Cooling as Asset Differentiation
For ArcTeron as an owner and financier of hyperscale data centers, liquid cooling is a central differentiating feature in the competition for hyperscaler tenants:
1. Tenant Requirements
Hyperscalers such as Microsoft, Google, and Amazon, as well as AI companies like xAI and Anthropic, are increasingly explicitly requiring liquid cooling capability in their RFPs (Requests for Proposals). Data centers without DLC infrastructure are excluded from the selection process.
2. Capital Structure and Financing
Liquid cooling infrastructure increases the capital expenditure (CapEx) of a hyperscale data center by typically 15–25% compared to pure air cooling. This additional investment is offset by higher rental rates (NNN lease) and longer contract terms (15–20 years). Institutional investors and banks are increasingly valuing liquid-cooling-capable assets with lower risk premiums, as regulatory compliance is secured for the long term.
3. Site Planning and Water Supply
Liquid cooling significantly reduces water consumption for evaporative cooling – a critical factor in water-scarce regions. At the same time, DLC requires a reliable water supply for the primary circuit. ArcTeron integrates water supply concepts (well water, rainwater harvesting, closed-loop systems) into site planning from the outset.
4. Modular Construction and Time-to-Market
The combination of modular construction (prefab modules) and integrated liquid cooling infrastructure enables ArcTeron to deliver hyperscale capacity in 18–24 months – compared to 36–48 months with conventional construction. This time advantage is a decisive competitive edge in the current market environment, where hyperscalers book capacity with 2–3 years of lead time.
Market Development and Outlook
The European market for liquid cooling infrastructure in data centers is growing at a CAGR of over 25% (2024–2030). Key drivers include:
- AI industrialization: Generative AI is no longer an experimental workload but business-critical infrastructure
- Regulatory pressure: EnEfG, EU minimum standards, and ESG requirements from institutional investors
- Energy costs: Liquid cooling reduces energy consumption for cooling by 30–50% compared to air cooling
- Hyperscaler demand: All major cloud providers have defined liquid cooling as the standard for new capacity
For data center owners investing today in liquid-cooling-capable infrastructure, this creates a strategic positioning for the next decade: regulatorily compliant, technologically future-proof, and optimized for high-growth AI workloads.
ArcTeron: Liquid Cooling as Infrastructure Standard
ArcTeron plans and finances hyperscale data centers with integrated liquid cooling infrastructure as standard. With a target portfolio of 5 GW across six European countries (Germany, Spain, Portugal, Poland, Czech Republic, UK) and direct access to 380/400 kV grids as well as gas and hydrogen solutions for energy supply, ArcTeron positions itself as the preferred infrastructure partner for hyperscalers and AI companies in Europe.
The integration of waste heat utilization concepts, EnEfG-compliant planning, and modular construction makes ArcTeron assets long-term, value-stable infrastructure investments – for institutional investors, banks, and hyperscaler tenants alike.
Learn more about ArcTeron as a European hyperscale infrastructure owner and energy partner: arcteron.ai