TL;DR

  • AI Workloads Overwhelming Traditional Cooling: AI’s demand for concentrated compute power is driving rack densities from 5-10 kW to beyond 50-100 kW, creating extreme heat that traditional, water-intensive air cooling systems can no longer handle.
  • Infrastructure Must Evolve: Thermal management, power delivery, and resource efficiency are now co-dependent, requiring developers to treat data center infrastructure as an integrated system rather than sacrificing sustainability for performance.
  • The Immersion Cooling Advantage: Fully submerging IT components in thermally conductive fluid provides superior, uniform heat removal, enabling higher rack densities while significantly lowering both water and energy consumption compared to traditional methods.
  • Fluid Selection is the Linchpin: The ultimate success and sustainability of immersion cooling rely on selecting the right engineered fluids, which dictate heat transfer efficiency, hardware lifespan, and overall environmental impact.

# # #

The Current State of Data Center Cooling

In 2026, data centers are caught between surging demand and tightening constraints. AI workloads are driving the need for higher compute density, while developers face limits related to power availability, water usage, and permitting timelines. Policymakers and communities are scrutinizing new projects, slowing approvals due to various concerns ranging from grid strain to land and water use. Meanwhile, hyperscalers and enterprise operators must meet aggressive sustainability targets even as their infrastructure footprints expand.

These pressures are exposing fundamental structural challenges in how data centers are built. Traditional air cooling with water-intensive systems cannot scale to meet AI’s mounting demands. Developers must now tackle multiple variables at once: more compute per square foot, lower energy and water intensity, and alignment with evolving regulatory frameworks. Meeting these challenges requires rethinking infrastructure as an integrated system where performance, efficiency, and sustainability are engineered together, not sacrificed for each other.

AI as the Catalyst for Convergence

Artificial intelligence has fundamentally changed infrastructure development strategies. Unlike traditional workloads, AI training and inference require highly concentrated compute power, delivered through GPUs and accelerators operating at significantly higher thermal design power (TDP). Rack densities that once averaged 5–10 kW are now pushing beyond 50–100 kW, with next-generation deployments expected to drive densities even higher.

The implications for power infrastructure are already appearing. According to Dean Nelson, founder and chairman of Infrastructure Masons, the industry is building as much data center capacity in the next three years as was built in the past thirty. The infrastructure required to support that pace of growth must align with these operational realities. After all, the cascading effects are significant. Higher compute density drives more heat, which demands more cooling. Cooling systems consume more power and, in many cases, more water, putting compounding pressure on resources already under strain. Power availability is becoming a gating factor, with utilities struggling to keep pace with the speed and scale of data center growth.

As a result, AI performance is determined by the infrastructure that supports it, not just by model architecture or silicon. Thermal management, power delivery, and resource efficiency are now co-dependent variables. As such, decisions from the chip to the facility cooling plant must be evaluated through a single, system-wide lens that takes these factors into account.

Immersion Cooling as an Infrastructure Strategy

Cooling is now a critical lever for enhancing performance and efficiency. As TDPs rise across the technology stack, liquid cooling is becoming a requirement at leading AI facilities. Immersion cooling, in particular, is uniquely positioned to address the three constraints underpinning modern data center development: heat management, water use, and power consumption.

Compared to traditional air cooling and direct-to-chip liquid cooling, immersion cooling delivers significant improvements in thermal management efficiency. By fully submerging IT components in a thermally conductive fluid, heat is removed directly and uniformly at the source, eliminating the limitations of air cooling and the complexity of localized cold plates and plumbing. Direct-to-chip cooling targets high-heat components like GPUs and CPUs but still relies on air or secondary systems to cool the rest of the hardware. Immersion cooling, by contrast, manages the entire system within a single thermal environment.

The Operational Benefits of Immersion Cooling

The superior heat capacity and thermal conductivity of immersion fluids enable consistent, high-efficiency heat transfer across all components, even at extreme power densities. This helps data centers support significantly higher rack densities without complex airflow management, extensive piping networks, or oversized mechanical cooling systems. Benefits such as better performance, longer hardware lifespan, and greater operational reliability are proving critical for high-density AI workloads.

Immersion cooling also reduces reliance on air handling and evaporative systems, significantly lowering water consumption. It mitigates overall energy demand for cooling as well, easing pressure on constrained power infrastructure. The net effect is not only thermal efficiency, but a more balanced and sustainable resource profile that regulators, investors, and communities increasingly prioritize.

Fluid Selection: The Linchpin of Data Center Infrastructure Sustainability

However, realizing these benefits depends on fluid selection. Fluid qualities like thermal stability, heat transfer efficiency, material compatibility, and environmental profile all determine system performance and lifecycle outcomes. The right fluid enhances heat transfer, extends system longevity, and reduces maintenance requirements. It also shapes sustainability metrics, from lifecycle emissions to end-of-life handling.

As the demands of AI workloads intensify, these decisions carry increasing weight. Now, when selecting a cooling technology, data center developers are making long-term infrastructure commitments that will determine performance, scalability, and environmental impact for years to come. Immersion cooling, when paired with thoughtfully engineered fluids and integrated system architectures, can help developers align all three qualities to improve operational efficiency while addressing common challenges.

# # #

About Robert Schuetzle

Robert Schuetzle is President and CEO of Infinium, a technology company transforming how the world powers, moves, and computes. Backed by strategic investors including Amazon, MHI, NextEra Energy, and AP Ventures, Infinium is best known for its leadership in ultra-low carbon electrofuels and is now expanding that expertise into digital infrastructure through Infinium Edge™, its platform focused on energy efficiency and thermal management for power- and heat-intensive systems such as data centres.