The cooling shift does not reduce the campus’s approved water allocation, it changes the electricity load profile, raising the grid draw required to operate the facility

Decision Focus

Google confirmed on July 23, 2026, that at least the first of three planned data center buildings on its 312-acre campus in Botetourt County, Virginia, will use air-cooling technology rather than water. The announcement, communicated through the county’s website, attributed the decision to “business demand and infrastructure requirements.” For Global Heads of Data Center Energy, the operative signal is not the community relations narrative. It is the direct substitution of grid electricity for water as the primary cooling medium, and what that substitution implies at scale for power infrastructure planning and interconnection capacity sizing.

90-Second Brief

Now, google will deploy air cooling in at least one building at Botetourt; cooling technology for the remaining two buildings is still under evaluation. The performance agreement with the Western Virginia Water Authority allowing up to 8 million gallons per day at full build-out remains in force, and no updated water consumption figures have been submitted to the authority since the announcement. The cooling shift does not reduce the campus’s approved water allocation, it changes the electricity load profile, raising the grid draw required to operate the facility.

What Is Really Happening?

Drought conditions and civic opposition in Southwest Virginia appear to have accelerated a decision Google frames as infrastructure-driven, but the physics are straightforward: air cooling is less energy-efficient than liquid cooling. A 2022 study by the American Society of Mechanical Engineers quantified the gap — moving from 100% air cooling to a 75% liquid, 25% air mix reduces facility power use by 27%. Running that logic in reverse, a facility defaulting to full air cooling carries a materially higher electricity load than a comparable liquid-cooled installation at identical IT density.

Cooling already accounts for at least 40% of power use across U.S. data centers, which consumed roughly 176 terawatt hours in 2023 — equal to about 4.4% of total national electricity use, according to a congressional report. At a campus scale involving three buildings and three substations, cooling method is not a design detail. It is a load-shaping variable that flows directly into interconnection capacity requirements, PPA sizing, and utility draw forecasts.

What distinguishes this moment from a standard technology announcement is the divergence between operational direction and contractual baseline. The water authority has received no updated consumption figures, and the 8 MGD performance agreement remains the governing document. Whether Google will seek to renegotiate that ceiling — or simply reduce actual draw while preserving contractual optionality — has not been stated publicly.

Why It Matters for Global Heads of Data Center Energy

A campus choosing air cooling is, by definition, requesting more grid capacity than the same campus running liquid cooling at equivalent workload. For operators managing interconnection queue strategy across multi-building campuses, the implied load difference compounds existing constraints. If the 27% efficiency gap from the ASME study holds as an order-of-magnitude reference, interconnection application sizing and substation capacity planning for a large campus shift materially based on cooling method alone — independent of IT load growth.

There is a second-order consequence worth tracking at the portfolio level. Community resistance to water consumption is not unique to Botetourt. Municipal supply constraints and civic opposition have emerged in multiple data center markets across drought-stressed geographies. If that pressure becomes a recurring variable in cooling method selection, the industry’s aggregate electricity demand rises not only from AI-driven workload growth but from infrastructure choices responding to local resource limits. These two demand drivers compound rather than substitute for each other, and they do not always appear in the same forecasting model.

The contractual sequence at Botetourt also sets a precedent. The water agreement ceiling was established before the cooling method was determined, and the operational announcement has not yet triggered a contract revision. That gap — announced change preceding formal agreement update — creates ambiguity about actual resource commitments and may prompt water authorities and utilities in future negotiations to require cooling technology commitments earlier in the development process.

Forward View

Three fronts are worth active monitoring. First, whether Google submits revised consumption estimates to the Western Virginia Water Authority and whether that triggers formal renegotiation. The authority’s executive director has noted that providing water beyond the initial allocation would be at the authority’s discretion — meaning the ceiling is not guaranteed, but the floor has not been revised downward either.

Second, how the cooling technology decision for buildings two and three resolves. If all three buildings ultimately deploy air cooling, the campus’s aggregate grid draw will be substantially higher than a liquid-cooled equivalent, with direct consequences for the three planned on-site substations and regional capacity planning across the broader Virginia grid.

Third, whether comparable community-pressure-driven cooling pivots emerge in other water-stressed markets. Operators with large-footprint campuses in drought-sensitive regions may face structurally similar pressure. The Botetourt sequence — community opposition, technology pivot, unchanged contractual water ceiling — could become a pattern that shapes both utility engagement strategy and site-level resource planning.

What Is Still Uncertain

The net power load increase from this cooling shift has not been quantified publicly for the Botetourt campus. The water authority has not received updated consumption estimates, so the actual reduction in water draw remains unspecified. Cooling technology for buildings two and three is under active evaluation, leaving the full campus energy profile genuinely open as of this date. Whether Google will renegotiate the 8 MGD performance agreement — or whether the authority will require it before approving additional supply — is unresolved. The specific infrastructure constraints that made air cooling preferable for building one have not been disclosed.

One Question for Your Team

If community resistance to water use accelerates air-cooling adoption across your portfolio markets, how does your current interconnection capacity planning account for the higher grid draw that replaces the water load — and have your PPA structures been sized against that revised baseline?


Sources

  • Cardinalnews — Google to use electric cooling instead of water in at least 1 of 3 data centers on Botetourt campus (Link)