AI Cooling Equipment and Data Center Tax Valuation

Tax Services

November 16, 2026

How AI Cooling Equipment Is Reshaping Data Center Property Tax Valuation

Rapid growth of AI is changing more than the computing equipment housed inside data centers, altering the physical infrastructure required to support the equipment and, by extension, property tax valuation.

At the center of this evolution is heat. Successive generations of graphics processing units (GPUs) are delivering dramatically more computing power, but they are also concentrating more electrical load within individual racks, a phenomenon known as rack density.

At one point, traditional enterprise and cloud racks operated in the 5 to 20 kilowatt (kW) range. Current AI deployments can exceed 100 kW per rack, and industry road maps now contemplate 200 to 400 kW racks, 500 kW systems and, eventually, megawatt (MW)-class rack designs. For perspective, 1 MW would power an average of 800 U.S. households, all in a space the size of a refrigerator.

Rack density gives rise to heat density. The First Law of Thermodynamics tells us that each watt of power used by an AI server converts to heat, and that heat must be offset by an equal amount of cooling, or facilities risk melting down the very servers that generate that heat.

For property tax professionals, the important question is not simply how much these cooling systems cost, but what the technological transition says about the economic utility, remaining life, and contributory value of both new and existing data centers.

Cooling Capacity as a Measure of Functional Obsolescence

Historically, data center valuation has focused heavily on electrical capacity, redundancy, location and building quality. Increasingly, however, the ability to support high-density computing can make or break a facility's competitiveness.

Consider two otherwise similar data centers. One can economically support only conventional air-cooled racks, while another can accommodate high-density GPU clusters using direct-to-chip liquid cooling. Both facilities may remain physically functional, but their ability to serve emerging customer requirements could be materially different. If an older facility cannot economically accommodate the rack densities demanded by newer computing platforms, that loss of utility may be evidence of functional obsolescence.

This issue is particularly important when assessments rely heavily on the cost approach. Historical investment does not necessarily equal current market value if portions of that investment no longer provide competitive utility. Cost does not equal value.

The Next Cooling Step: 200 kW, 500 kW and Beyond

The industry is already planning for another step change in rack density. With organizations like ASHRAE and Vertiv anticipating a future of 200 kW racks and up, facilities will need more than simply larger chillers.

Direct-to-chip liquid cooling is becoming the baseline architecture for high-density AI. Cold plates remove heat directly from GPUs and central processing units (CPUs), while coolant distribution units (CDUs) separate the technology cooling loop from the facility water system.

At higher densities, facilities may also require larger piping, higher flow rates, redundant pumps, warm-water loops, dry coolers or adiabatic heat rejection, and increasingly sophisticated controls capable of reacting to rapid load changes. Rear-door heat exchangers can remain useful in hybrid designs. Eventually, full-blown immersion cooling may be appropriate for specialized deployments.

Power and cooling are also converging. Higher-voltage distribution, including emerging 800-volt direct current (VDC) architectures, can reduce current and copper requirements as rack power climbs. Structural design, leak detection, water chemistry, controls and the ability to service heavier liquid-cooled racks become part of the economic utility of the facility itself.

AI Cooling Equipment and Data Center Tax Valuation

Assessment Asymmetry: Capital Expenditure Arrives Faster Than Obsolescence

This transition creates an important property tax asymmetry. For legacy data center facilities, construction costs that include cooling infrastructure can be easy to capture, but the obsolescence of the infrastructure being replaced, or the diminished utility of a legacy facility that cannot support high-density racks, may be recognized much more slowly.

The reverse can also occur. A developer may spend heavily to future-proof a facility for 500 kW or even 1 MW racks before market demand for that capability is widespread. From a valuation perspective, the full cost of anticipatory capacity should not automatically be assumed to contribute dollar-for-dollar to current market value. The relevant question is whether a market participant would pay for that capability as of the valuation date.

Will Future Chips Reduce the Cooling Requirement?

There is an important counterargument to the assumption that rack heat density must rise indefinitely. AI chip and software efficiency are improving rapidly. NVIDIA, for example, reports substantial generational gains in inference throughput per MW (from Hopper to Blackwell) and projects further efficiency gains with Vera Rubin. If future chips generate materially more compute for each watt consumed, the heat produced per unit of useful computation can fall.

That does not necessarily mean rack cooling requirements will decline. Operators may use efficiency gains to place more accelerators in the same footprint, run larger models, increase utilization or generate more tokens within the same power envelope. In other words, better performance per watt can reduce the energy required for a given amount of compute while absolute rack power and heat density continue to rise.

For property tax valuation, this uncertainty matters. A forecast of ever-higher rack density should not be treated as inevitable, but neither should rapid efficiency gains be assumed to make advanced cooling unnecessary. The most supportable valuation approach is scenario-based: evaluate the density the facility can support today, the cost and feasibility of upgrading it, the market demand for higher-density capability and the extent to which future-proofing expenditures currently contribute to value.

A New Valuation Question

As AI reshapes data center design, assessors and taxpayers may increasingly need to look beyond square footage and total MW. A relevant question is what density of computing can this facility economically support.

For owners of legacy facilities, cooling limitations may provide measurable evidence of functional obsolescence. For new facilities, rapidly evolving cooling technologies may support shorter economic lives for certain specialized assets. And for facilities designed far ahead of current demand, excess or anticipatory capacity may raise a different question of contributory value.

The property tax challenge will be ensuring that assessments recognize all three effects: the cost of next-generation infrastructure, the obsolescence created by technological change and the possibility that not every dollar spent on future capability is reflected in present market value.

Stay Ahead with Kroll

Tax Services

Built upon the foundation of its renowned valuation business, Kroll's Tax Service practice follows a detailed and responsive approach to capturing value for clients.

Property Tax Services

Kroll engages with companies nationwide to provide independent, innovative and results-driven property tax services.