Digital Infrastructure and Hydrological Stress - Colorado

Digital Infrastructure and Hydrological Stress: Projected Impacts of Data Centre Expansion on the Colorado River Basin (2025–2050)

The Colorado River Basin is experiencing a long-term structural water deficit driven by over-allocation, climate change, and rising demand. 

Concurrently, rapid expansion of water-intensive digital infrastructure—particularly hyperscale data centres supporting artificial intelligence (AI)—is introducing a novel and poorly quantified demand vector. 

This paper synthesised historical hydrological data, estimates the number of data centres dependent on the Colorado River system, and models basin decline over the next 25 years under multiple scenarios. 

Results suggest that while data centres remain a minority water user (less than 1 percent regionally), their rapid growth, spatial concentration, and consumptive use patterns amplify localised scarcity. 

Under high-demand scenarios, basin flows may decline by up to 35 percent by 2050, also intensifying allocation conflicts.

Introduction

The Colorado River supplies water to approximately 40 million people across the southwestern United States and Mexico. 

However, its allocation framework—established under the 1922 Compact—was based on overestimated flow regimes.

Modern hydrological evidence indicates:

• Long-term average flow: about 12 to 13 million acre-feet per year

• Allocated flow: about 16.4 million acre-feet per year

This creates an obvious structural deficit.

At the same time, hyperscale data centres are expanding rapidly in arid regions due to land, energy access, and policy incentives. These facilities require large volumes of water for cooling, raising concerns about cumulative impacts.

Hydrological decline

Observed declines since 2000 include:

- Reservoir storage falling from about 90 percent to about 30 percent

- Reduced snowpack and earlier melt cycles

- Increased evaporation due to higher temperatures

Groundwater depletion

Satellite observations show groundwater loss equivalent to roughly one Lake Mead over two decades, reducing long-term resilience.

Data centre water use

Data centres consume water primarily through evaporative cooling. This water is not returned to the basin, making it a consumptive use.

Estimating data centre dependence

We define a data centre as dependent on the Colorado River if at least 25 percent of its water supply originates from the river system.

Water consumption model

Total annual water use by data centres is calculated as:

• Wdc = sum of (Pi × lambda i × theta i)

Where:

• Pi = computing capacity in megawatts

• lambda i = water used per unit of electricity (gallons per megawatt-hour)

• theta i = cooling system factor (efficiency adjustment)

Hydrological projection model

Future river flow is estimated using:

Qt = Q0 × (1 - alpha x t) - (beta x Tt) + (gamma x Mt)

Where:

Qt = river flow at time t

Q0 = initial flow

alpha = baseline annual decline rate

beta = sensitivity to temperature

Tt = temperature increase at time t

gamma = mitigation effectiveness

Mt = mitigation effort at time t

Scenario design

Three modeled scenarios:

Scenario 1: Managed Adaptation

Scenario 2: High AI Growth

Scenario 3: Climate Shock

Results

Estimated data centre dependence

Estimated total facilities in basin: 300 to 500

Estimated dependent facilities: approximately 200 to 350

Water consumption projections

2025: about 5 to 10 billion gallons per year

2035: about 29 billion gallons per year

2050: about 40 to 60 billion gallons per year

Basin decline projections

Scenario outcomes:

Scenario 1: about 20 percent decline in flow

Scenario 2: 25 to 35 percent decline

Scenario 3: greater than 35 percent decline

Localised vs total impact

Data centres use a small share of total water, but:

- They are concentrated in high-stress regions

- Their water use is consumptive

- Their growth rate is extremely high

Energy-water linkage

Electricity production for data centres also consumes water, indirectly increasing total impact.

Key issues:

- Limited transparency in reporting

- Incentives encouraging desert siting

- Lack of integrated water-energy planning

Limitations

Limited public data on facility-level water use

Assumptions about cooling efficiency

Uncertainty in climate projections

The Colorado River Basin faces a convergence of:

• Declining supply

• Over-allocation

• Rapid demand growth

Data centres are not the dominant driver but represent a fast-growing and strategically important source of water demand that will increasingly influence policy and allocation decisions.

When the Colorado River Runs Short: The Infrastructure Crisis a 25% Decline Could Trigger

The Colorado River is not simply a river. It is the backbone of an enormous infrastructure network spanning the American Southwest — supplying cities, farms, power stations, reservoirs, canals and industrial systems across seven US states and parts of Mexico.

More than 40 million people depend on the river for municipal water, while approximately 5.5 million acres of farmland are supported by its water. The river also generates more than 8 billion kilowatt-hours of hydroelectricity each year.

That makes projections of a substantially smaller river more than an environmental warning. They are an infrastructure warning.

Recent modelling contains considerable uncertainty, but just a 15–25% reduction in Colorado River flows by around 2050 is within the range of mid-range scenarios cited in Bureau of Reclamation material. Higher-warming scenarios produce substantially larger declines.

If a decline of roughly 25% became the new normal, the consequences would extend far beyond having less water coming downstream.

A system already operating under pressure

The Colorado River infrastructure system was designed around a river that no longer reliably behaves as it once did.

Lake Powell and Lake Mead are the two largest reservoirs in the system, with a network of dams, canals and pumping facilities depending upon their elevations. The Bureau of Reclamation has warned that declining reservoir levels can threaten both water deliveries and hydropower production.

The problem is therefore cumulative.

Less river flow means less water entering the reservoirs. Less water in the reservoirs means less flexibility for operators.

And less flexibility means infrastructure that was designed to provide reliable supplies begins operating much closer to its limits.

A 25% reduction would not necessarily mean that every community suddenly receives 25% less water. Reservoirs, conservation programmes, groundwater, water recycling and changes to allocations can absorb some of the loss.

But they cannot make the missing water disappear.

Reservoirs would become the first major pressure point

The Colorado River's reservoir system functions as a giant strategic buffer.

During wet years, water can be stored. During dry years, that stored water can be released to maintain downstream supplies.

A persistent reduction in river flows would gradually erode that buffer.

The significance of this is already visible. In April 2026, Reclamation reported that Colorado River system storage had fallen to approximately 36% of capacity amid prolonged drought, record-low snowpack and extreme heat.

Lake Powell's elevation is particularly important because Glen Canyon Dam needs sufficient water elevation to maintain normal releases and generate electricity.

In February 2026, Reclamation warned that under its most probable scenario Lake Powell could fall to its minimum power pool during 2026, with still lower levels projected in 2027.

A long-term 25% reduction would make maintaining these reservoirs substantially more difficult.

Hydroelectric power becomes increasingly vulnerable

Water shortages do not only threaten taps and irrigation. They threaten electricity.

Hydroelectric turbines depend upon both water volume and hydraulic head — essentially the amount of water available and the pressure created by its elevation.

As reservoirs fall, generating capacity can decline.

That creates a particularly awkward problem for the Southwest because electricity demand is already heavily influenced by extreme summer temperatures and the enormous air-conditioning load of cities such as Phoenix and Las Vegas.

Reclamation says the Colorado River system produces more than 8 billion kilowatt-hours of electricity annually, enough to serve roughly 700,000 homes.

A prolonged reduction in reservoir elevations would therefore create another infrastructure challenge: replacing lost hydroelectric generation with other sources. That could mean additional solar, wind, gas, storage, transmission infrastructure or some combination of them.

Water-treatment plants would face a more complicated future

Water-treatment infrastructure does not simply require water to exist somewhere. It needs water at a particular location, at a particular elevation, with manageable quality.

Lower reservoir levels can change water chemistry and increase the concentration of some contaminants. Lower flows can also reduce the system's ability to dilute pollutants and salts.

Salinity is already recognised by Reclamation as one of the resources at risk within the Colorado River system. Treatment plants may consequently need additional processing capacity.

That means more infrastructure, more energy consumption and potentially higher operating costs.

Pumping water becomes increasingly difficult

Much of the Southwest's water infrastructure is effectively an enormous pumping network. Water has to be moved from reservoirs and rivers into canals, treatment plants, storage facilities and urban distribution systems.

As water levels fall, intake structures can become increasingly difficult to operate efficiently.

The infrastructure may still physically exist, but its operating conditions change.

This is one of the less visible consequences of water scarcity: infrastructure can remain intact while becoming progressively less useful.

Eventually, communities may need to modify intakes, deepen pumping systems, install new pumps or construct alternative connections.

Those projects can cost millions or billions of dollars.

Agriculture would put enormous pressure on the system

Agriculture is arguably where the consequences would become most economically visible.

The Colorado River supports approximately 5.5 million acres of farmland, producing a substantial share of important crops consumed in the United States.

A 25% reduction in river flows would not necessarily translate directly into a 25% reduction in agricultural production.

Farmers could respond through:

- drip irrigation

- improved canal efficiency

- crop switching

- fallowing

- groundwater pumping

- water trading

- changes in planting schedules

- greater use of drought-tolerant crops

But each solution has limits. More groundwater pumping, for example, transfers pressure from the river to underground aquifers.

Once again, the infrastructure problem moves rather than disappears.

Cities would have to build their way out of the shortage

The largest cities dependent upon the Colorado River have already invested heavily in conservation and alternative water supplies.

A deeper long-term shortage would accelerate that trend. Urban infrastructure could increasingly include:

Water recycling

Wastewater would be treated to increasingly high standards and returned to the municipal supply.

Desalination

Where economically and environmentally feasible, seawater or brackish water could provide another source — particularly in coastal California.

Aquifer storage

Excess water could be deliberately injected or allowed to infiltrate into underground formations for later recovery.

Stormwater capture

Cities could increasingly treat rainfall as a resource rather than simply something to drain away.

Leak reduction

Ageing pipes and distribution systems would come under greater scrutiny because every litre lost becomes more expensive.

These technologies can significantly improve resilience, but they require enormous capital investment.

The Colorado River's canals could become strategic infrastructure

The river is supported by an extraordinary network of canals and aqueducts.

Among the most important are systems transporting Colorado River water across large distances to agricultural regions and metropolitan areas.

A future of chronic scarcity would increase the strategic importance of these structures.

Canals may need:

- lining and rehabilitation

- improved measurement systems

- automated gates

- leak detection

- new pumping facilities

- greater storage capacity

- improved control systems

The irony is that a shrinking water supply could require more infrastructure rather than less.

Reclamation has already funded projects involving water-distribution structures, advanced metering, canal lining, agricultural efficiency, groundwater banking, recycling and water purification as part of efforts to improve Colorado River resilience.

Groundwater could become the hidden infrastructure casualty

When surface water becomes unreliable, communities and farmers naturally look underground.

But aquifers are not infinite reservoirs.

In some areas, excessive groundwater withdrawal can lower water tables, increase pumping costs and contribute to land subsidence.

That creates a potentially dangerous feedback loop:

Less river water → more groundwater pumping → falling water tables → deeper wells → higher energy costs → greater infrastructure requirements.

Groundwater can therefore act as an emergency reserve, but it cannot necessarily serve as a permanent replacement for a declining river.

Mexico would be part of the infrastructure equation

The Colorado River does not stop being an international resource at the US–Mexico border.

Water deliveries to Mexico are governed by the 1944 Water Treaty, while the river also supports Mexican communities and ecosystems.

Reclamation's current post-2026 planning explicitly treats Mexico's water deliveries as a separate binational process.

A permanently smaller river would therefore increase pressure not only between US states but between two countries.

Infrastructure decisions — reservoirs, canals, treatment plants, conservation projects and water transfers — would increasingly become part of an international negotiation.

The most expensive problem may be rebuilding the system itself

The greatest danger is not necessarily a dramatic day when the Colorado River suddenly stops supplying a city.

It is a slow degradation of reliability. A water system can continue functioning while becoming progressively more expensive.

Pumps work harder. Treatment becomes more complicated. Reservoirs provide less safety margin. Hydroelectric generation becomes less reliable. Farmers invest in more efficient irrigation. Cities build recycling plants. New pipelines are constructed. Old infrastructure is modified. Groundwater wells become deeper. Electricity demand rises.

Eventually, the cost of maintaining the old system under new hydrological conditions becomes enormous.

A river does not have to disappear to cause an infrastructure crisis

This is perhaps the most important point.

The Colorado River does not need to run dry for the Southwest to face a profound infrastructure challenge.

A 25% reduction in long-term flow could be enough to transform the economics and engineering of the entire system.

And the uncertainty cuts both ways.

Reclamation emphasises that long-term projections become increasingly uncertain as they look further into the future. Research published in Nature Communications has nevertheless found substantial projected long-term reductions in streamflow, with an intermediate estimate of approximately 20% by 2060 and 30.5% by 2100.

The precise number in 2050 therefore matters less than the direction of travel.

The infrastructure of the American Southwest was built around the assumption that water could be stored, transported and allocated with reasonable reliability.

Climate-driven reductions in river flow challenge that assumption.

And unlike building a new road or power station, there is no engineering project capable of manufacturing a second Colorado River at the scale required.

The future challenge may therefore be less about finding a technological replacement for the river and more about redesigning an entire civilisation's infrastructure around having less of it.

References

American Rivers. (2024). Data centers and water use in the Colorado River Basin.

Castle, S. L., et al. (2014). Groundwater depletion during drought threatens future water security. Geophysical Research Letters, 41(16), 5904–5911.

Consumer Reports. (2025). AI data centers’ impact on water and electricity.

Cook, B. I., et al. (2020). Climate change and megadrought in the American Southwest. Science Advances, 6(37).

Gleick, P. H. (2018). The Colorado River Basin: Water management challenges. Environmental Research Letters, 13(9).

NASA GRACE Mission Data (2023). Groundwater depletion in the Colorado River Basin.

U.S. Bureau of Reclamation. (2023). Colorado River Basin water supply and demand study.