Efficiency gains are reducing power consumption per unit of compute, but aggregate electricity demand continues to rise as adoption accelerates
The System Pressure
The growth thesis for digital infrastructure is intact. AI, cloud adoption, and enterprise digitization are sustaining a demand curve that has consistently outpaced projections, with capital potentially reaching $10 trillion by 2030.
But the bottleneck has moved. Power availability — not capital, not land — is the primary constraint on data center expansion today. Grid interconnection timelines measured in years, not months, are a standard feature of major markets. That structural lag means sites funded today are competing for queue positions that may not clear for years.
What is changing is that power availability is no longer the only upstream constraint. Water access and workforce capability are now converging with power in ways that transform the risk profile from a single-path problem into a systems problem. A project that clears one bottleneck can stall at the next. That interaction is not widely modeled in capital allocation decisions — and it is where execution risk is accumulating.
The Drivers, Dependencies, and Constraints
Power demand growth is the primary driver. AI compute workloads are pushing rack densities higher than infrastructure planning cycles anticipated, compressing the gap between substation design assumptions and actual load requirements. Efficiency gains are reducing power consumption per unit of compute, but aggregate electricity demand continues to rise as adoption accelerates. Overall data center electricity demand is growing, not flattening.
Water is a second-order dependency that is becoming a first-order constraint. Global data center water consumption currently stands at approximately 560 billion liters per year. On current trajectories, that figure could more than double to 1.2 trillion liters by 2030. In fast-growth markets, cooling water demand alone could rise by 870% — a figure that reflects not just aggregate consumption, but local basin stress and peak demand concentration. A typical facility draws around 300,000 gallons per day; large campus sites can reach 5 million gallons, the equivalent daily consumption of a town of 50,000 people.
The second-order effect matters specifically for energy strategy: cooling architecture shapes power demand. Higher-density racks that rely on water-intensive cooling in water-stressed basins create a feedback loop where energy and water constraints bind simultaneously. Operators selecting cooling architecture without stress-testing local water availability are accepting a compounding risk, not two independent risks.
Workforce is the third constraint — and the one most consistently underweighted in capital planning. Fifty-eight percent of data center operators currently report difficulty attracting and retaining qualified staff. That gap concentrates in the specialist roles closest to power and cooling systems: commissioning engineers, substation technicians, and O&M specialists experienced in high-density environments. As these systems grow more complex, those vacancies carry operational consequences beyond HR: longer mean time to recovery, higher SLA exposure, and elevated probability of disruptive incidents linked to human error, even absent major equipment failure.
Open Dependencies
Several dependencies in this system remain unresolved at the operating level, and the ambiguity carries material consequences.
The cyber-physical boundary of power infrastructure is one. Cyber risk in data center operations is now a strategic business issue, not a narrow IT function. As control systems, substations, transformers, and site energy assets become interconnected and third-party managed, the attack surface expands. Large-scale systemic cyber events affecting energy infrastructure are already influencing insurability assessments. What is not yet confirmed is how rapidly insurance and financing markets will formalize those exposures into pricing or exclusions. Operators building power strategies around third-party O&M and distributed energy assets should treat this as an active variable, not a background concern.
The relationship between efficiency improvement and risk concentration is a second unresolved dependency. Advanced compute and cooling configurations can concentrate more value and interdependence inside a smaller physical footprint. The conventional assumption — that efficiency automatically reduces exposure — does not hold when failure in a denser environment carries a larger operational and financial consequence. Whether existing recovery, redundancy, and risk-transfer frameworks are calibrated for that concentration is an open question at most operators.
The water-energy co-constraint in specific geographies also lacks sufficient granular public modeling. Operators siting in contested basins should not rely on aggregate projections; local hydrology assessments and regulatory pipeline visibility carry more decision weight than headline figures.
The Operating Exposure for Global Heads of Data Center Energy
For Global Heads of Data Center Energy, the practical exposure concentrates at transition — the period when a project moves from construction to live operation, systems are energized, and commissioning responsibility shifts between owners, operators, utilities, OEMs, and O&M providers. That is precisely when grid connection timing, cooling performance, contractor availability, and SLA exposure converge, at the moment when tolerance for failure is lowest.
If ownership, escalation rights, and recovery pathways across power, water, and workforce are not clarified before transition begins, a contained issue — a delayed energization, a cooling shortfall, a vacant specialist role — can quickly become a delivery, continuity, or financial event. The risk is not that any single constraint proves insurmountable; it is that the interaction between constraints under time pressure amplifies the cost of each individual gap.
Bankability is increasingly downstream of this. Permitting, investor confidence, and insurance coverage are all now more sensitive to how operators demonstrate credible management across all three constraint domains, not power alone. Boards and lenders are beginning to apply the same scrutiny to human capital risk that they have applied to power infrastructure — the engineering and construction skills gap is a present execution issue, not a future planning consideration.
Signals the System Is Shifting
The indicators worth tracking are those that reveal whether the three constraints are being treated as a system rather than separate workstreams. Watch for interconnection reform that shortens queue timelines while adding water discharge conditions — that would confirm regulatory recognition of the co-constraint. Watch for insurance market repricing of cyber exposure for distributed energy assets, which would accelerate the formalization of that risk into deal economics. And watch for workforce aging statistics in commissioning and substation engineering disciplines, which are leading indicators for execution gaps in the current build wave.
Capital formation is not the leading signal here. The $10 trillion trajectory tells you where demand is going. The three constraints tell you which fraction of that investment will actually deliver at plan — and under what conditions.
Sources
- Aon — People, Power and Water are Defining Digital Infrastructure Operational Risks (Link)
