The cycle runs in both directions: heat stored underground in summer can offset winter heating loads, and cold stored in winter can assist summer cooling
Decision Focus
On July 9, 2026, researchers at the Prairie Research Institute at the University of Illinois Urbana-Champaign published findings in the journal Groundwater proposing aquifer thermal energy storage (ATES) as a cooling mechanism for AI data centers. The study identifies cooling as consuming between 10 and 40 percent of total data center electricity depending on facility design, and positions ATES as a way to reduce that load by substituting underground temperature stability for mechanical refrigeration. The operating signal for energy heads is not about supply addition — it is about demand reduction achieved without a new interconnection position, a PPA commitment, or a transformer procurement cycle.
90-Second Brief
As the week closes, illinois researchers propose routing cool aquifer water through data center heat exchangers, then returning warm water underground for later seasonal use. The cycle runs in both directions: heat stored underground in summer can offset winter heating loads, and cold stored in winter can assist summer cooling. The study frames geology and climate as the primary feasibility filters, not the technology itself. The practical barrier the researchers identify is financial, ATES delivers benefits over a 20 to 40-year asset life, while most data center capital projects are evaluated on 5 to 10-year horizons.
What Is Really Happening?
Data centers sit at what the Illinois team describes as the water-energy nexus: reducing cooling energy typically increases water consumption through evaporative loss, while reducing water use often demands more mechanical cooling electricity. ATES is framed as an unusual case where both sides of the tradeoff improve simultaneously, because groundwater functions as a thermal carrier that is cycled rather than consumed.
The mechanism depends on three site conditions: seasonal temperature swings sufficient to create meaningful underground thermal differentials, aquifer geology that responds well to thermal exchange, and groundwater availability that does not compete with potable supply. The Illinois team is specific on water sourcing: saline aquifers — some saltier than seawater — and contaminated groundwater or water-filled abandoned mines are identified as viable inputs. That sourcing flexibility materially changes the availability calculus for sites in regions where municipal water access is constrained.
What elevates this beyond a regional geology story is the implied energy arithmetic. Cooling load is not a marginal line item. A technology that reliably reduces it by substituting geothermal exchange for compressor cycles addresses the portfolio directly — without adding capacity to the procurement stack.
Why It Matters for Global Heads of Data Center Energy
The primary operating implication is that demand reduction, not supply expansion, becomes a lever for closing the power gap at constrained sites. Reducing cooling load shrinks the interconnection capacity required per facility, softens budget pressure on new builds in queue-constrained markets, and lowers Scope 2 exposure without requiring offsetting renewable certificate purchases.
The bidirectional seasonal storage function adds a compound efficiency that single-direction cooling systems do not offer. Warm water returned to the aquifer after summer operations can be extracted in winter to supplement heating, reducing heating energy draw in continental climates. For large campuses with 20-plus year operational horizons — exactly the profile of hyperscaler and large colocation builds — this bilateral efficiency accumulates in ways that standard 5-year NPV screens structurally cannot capture.
The Illinois team explicitly dismisses workforce availability as a deployment barrier: drilling expertise already exists across the oil, gas, and water-well industries. That removes a common objection to emerging infrastructure technologies. The constraint that remains is financial framing: a 30-year system generating returns from year seven onward fails an evaluation screen that would not reject equivalent grid infrastructure investment on identical terms.
Forward View
Three fronts are worth watching as this research moves toward potential commercialization. First, whether operators with long-lived campuses — those already executing 15-year PPAs — begin commissioning site-specific ATES feasibility studies. Their capital authorization horizons already extend beyond the mismatch the researchers identify, making them the most natural early movers.
Second, how state-level water regulation evolves around saline and contaminated aquifer use for thermal exchange. In most U.S. jurisdictions, this application sits in a permitting gray zone. Regulatory clarity — or its absence — will determine whether deployment scales through repeatable approvals or remains locked in case-by-case review.
Third, whether build-out pressure in water-stressed regions shifts the cost calculus by making evaporative cooling alternatives effectively unavailable, not merely expensive. If water access becomes the binding constraint rather than power access, ATES’s water sourcing flexibility becomes a strategic differentiator independent of its energy efficiency case.
What Is Still Uncertain
The Illinois findings are a research proposal published in an academic journal, not a validated commercial deployment. Efficiency gains are described as conditional on ideal subsurface conditions; no quantified electricity savings at data center scale are provided. The specific magnitude of cooling load reduction attributable to ATES versus a baseline mechanical system is not stated. Geographic applicability to aquifer geology outside Illinois’s specific glacial deposits is asserted but not modeled for other U.S. markets. Regulatory pathways for saline or contaminated groundwater use in thermal exchange systems remain unresolved across most jurisdictions. Capital cost estimates for ATES integration with planned or existing cooling infrastructure are absent from the published research. Until a funded commercial pilot produces operating data, the energy savings case remains analytically promising but empirically unconfirmed.
One Question for Your Team
For your next site selection review: does your evaluation framework currently include an aquifer thermal feasibility screen alongside standard power availability and interconnection assessments — and if not, what threshold of cooling load reduction potential would justify adding one?
Sources
- Timestribunenews — Team looking to tap underground ‘thermal batteries’ to cool AI data centers, save water (Link)
