The cooling technology choice carries direct energy and water-use implications that differ materially from liquid-cooled, high-density AI compute facilities
Decision Focus
On July 23, 2026, Google confirmed it will deploy air-cooling technology at its planned data center in Botetourt County, Virginia. The facility sits outside the established Northern Virginia data center corridor — a geographic choice worth noting in its own right — and the cooling technology selection adds a second layer of signal. For Global Heads of Data Center Energy, the operative question is not what Google built, but what this technology choice reveals about energy load profile, water exposure, and workload intensity planning in a market where every hyperscaler infrastructure move is closely watched.
90-Second Brief
Today, google announced on July 23 that its Botetourt County data center will use air-cooling rather than water-based or liquid-cooling systems. The facility is located in Virginia’s Roanoke Valley region, not in the dense Ashburn corridor where grid and water constraints are acute. The cooling technology choice carries direct energy and water-use implications that differ materially from liquid-cooled, high-density AI compute facilities. No capacity figure, commissioning timeline, or workload type was confirmed in the announcement.
What Is Really Happening?
. Air-cooled data centers operate efficiently across a range of compute densities but face physical limits when rack power densities climb into ranges now associated with GPU-heavy AI inference and training clusters. Selecting air-cooling for a new build in 2026 — when most hyperscaler infrastructure investment is explicitly oriented toward high-density AI compute — suggests one of several possibilities: the Botetourt facility is designed for workloads at lower rack densities, Google is deliberately diversifying its cooling architecture across its portfolio, or site-specific constraints around water availability or permitting influenced the decision. None of these explanations has been confirmed.
Botetourt County draws on regional water infrastructure distinct from the Loudoun County and Prince William County systems that have faced growing scrutiny over data center water consumption. Whether water availability influenced this decision is not confirmed, but the geography is not incidental. Operators evaluating sites in secondary Virginia markets face the same water access questions, and a hyperscaler’s publicly confirmed technology choice in the same state creates a precedent point in regulatory and community engagement conversations.
Why It Matters for Global Heads of Data Center Energy
Air-cooling changes a facility’s energy consumption profile in ways that affect procurement, PUE targets, and sustainability reporting. At high densities, air-cooled systems consume more electricity per unit of compute output than liquid or direct chip cooling alternatives; at moderate densities, they can deliver predictable, lower-capital energy profiles with reduced water withdrawal. For operators managing Scope 2 reporting and 24/7 carbon-free energy matching, the cooling architecture directly shapes the energy volume that must be covered by clean procurement.
The location signal matters equally. If a hyperscaler is building outside established corridors, it implies either that interconnection capacity or land costs in those corridors have constrained expansion, or that secondary markets have reached sufficient infrastructure maturity to support a new build. Either implication affects how portfolio managers should evaluate Virginia’s secondary markets in their own site and interconnection pipeline reviews. A Google commitment in Botetourt County will attract developer attention, utility engagement, and may accelerate grid investment in that sub-region — reshaping the energy procurement landscape for anyone else considering the same geography.
Forward View
Three fronts are worth tracking. First, if Google expands Botetourt County capacity or announces additional Virginia facilities outside the Ashburn corridor, it will signal a deliberate geographic diversification strategy that other hyperscalers and colo providers are likely to follow, compressing site availability and interconnection timelines in secondary markets. Second, continued air-cooling deployments by major operators in new builds would indicate the industry is not converging uniformly on liquid cooling — with implications for transformer sizing, electrical infrastructure planning, and long-term energy procurement volumes. Third, Virginia regulators and utilities have shown increasing scrutiny of data center water use; if air-cooling adoption accelerates in response, it may partially rebalance that regulatory pressure while shifting energy load calculations upward.
What Is Still Uncertain
The confirmed information is limited to a single announcement: air-cooling at Botetourt County. The facility’s capacity in megawatts, expected commissioning date, workload type, and the specific reason air-cooling was selected over alternatives are not established in the available record. Whether this reflects a site-specific decision or a deliberate shift in Google’s infrastructure cooling strategy across its portfolio is unknown. The interconnection status of the facility — whether it is already in queue, at what voltage level, and with which utility — has not been reported. Without those details, the energy procurement and grid impact implications remain directional rather than precisely actionable. Operators should treat this as a signal to investigate, not a confirmed trend.
One Question for Your Team
Given that a hyperscaler has committed to an air-cooled facility in a secondary Virginia market, does your current site evaluation process explicitly assess how cooling technology requirements interact with water access, grid interconnection capacity, and long-term energy procurement volume at the subregional level — and if not, what would it take to add that layer to the next pipeline review?
Sources
- Wdbj7 — Google to Use Air-Cooling at Future Botetourt County Data Center (Link)
