In my earlier article on the Colorado River, I asked readers to imagine a pension fund that had promised $16.5 billion in future payouts but held only $12.4 billion, and that had been drawing down its balance for 25 years to cover the annual shortfall. That is the Colorado River Compact: 16.5 million acre-feet (MAF) of water promised each year against roughly 12.5 MAF of actual flow, with the difference made up by draining Lake Mead and Lake Powell.

Now imagine that the same pension fund had a line of credit. Whenever the fund came up short, the managers drew on it, and everyone (the beneficiaries, the auditors, the sponsors) took comfort in knowing it was there. Nobody looked very closely at the balance. It was simply assumed to be available when needed.

For the Colorado River Basin, that line of credit is groundwater. And a study published this week in Water Resources Research suggests that it is being drawn down far faster than anyone had planned for, and fastest of all in exactly the conditions when it is needed most.

Researchers at Virginia Tech, using NASA satellite data from 2004 through 2023, estimate that the Upper Colorado River Basin lost about 4.6 trillion gallons of groundwater during the 2020–2022 drought. That is roughly 14 MAF, enough to fill Lake Powell about halfway, lost in three years. Losses have accelerated since 2019. And in a reversal of the historical pattern, the typically wetter Upper Basin lost more groundwater during the drought than the drier Lower Basin. As the study’s lead author, Esther Oyedele, put it: “groundwater cannot simply just be treated as an unlimited backup anymore.”

This creates some problems for corporate balance sheets. First, groundwater is a hedge that fails in exactly the scenario it is meant to cover. Second, the river and the aquifer are not two separate accounts, so drawing on one depletes the other. And third, the same groundwater is being counted as a backup by almost every user in the basin at once.

What the New Science Shows

The Virginia Tech study is not the first warning. In 2025, a team at Arizona State University led by James Famiglietti estimated that the basin as a whole lost about 13 trillion gallons of groundwater between 2002 and 2024, roughly 41 MAF, or about 72% of the storage capacity of the basin’s federal reservoirs. In other words, more than a Lake Mead’s worth of water has disappeared underground while public attention has been fixed on the bathtub rings at Lake Mead and Lake Powell. The ASU team also found that depletion over the most recent decade ran about three times faster than in the decade before.

Historically, most of that loss has been in the Lower Basin: about 29.2 MAF, compared with about 11.8 MAF in the Upper Basin, according to the ASU study. Parts of Arizona, including the Prescott area, the Hualapai Valley and the southeast near the Mexican border, rely on groundwater for 80% to 100% of supply, mostly for agriculture. Only about 18% of Arizona’s land area is subject to active groundwater management.

What the new Virginia Tech study adds is a shift in where the stress is appearing. During the most recent drought, the mountainous headwaters of the Upper Basin, the part of the system that is supposed to feed everyone downstream, drew down their underground reserves faster than the desert Lower Basin. Warmer temperatures, drier soils, lower snowpack and rising demand are all contributing.

Wrong-Way Risk

In finance, there is a concept called wrong-way risk. It describes a situation in which your exposure to a counterparty grows at the same time that the counterparty’s ability to pay deteriorates. The textbook example is 2008. Banks bought credit default swaps from AIG to protect themselves against a collapse in mortgage-backed securities. But the same collapse that triggered the payouts also undermined AIG’s ability to make them. The protection failed because it was correlated with the risk it was meant to cover.

Groundwater is a hedge with wrong-way risk. When surface water is plentiful, nobody needs to pump much groundwater. When drought cuts surface deliveries, farms, cities and industries turn to their wells all at once, and pumping surges. At the same time, the drought reduces the snowmelt and rainfall that recharge the aquifer. The backup is drained fastest in the very years when the primary supply has failed. The Virginia Tech data show this pattern playing out in real time: the sharpest groundwater losses coincided with the 2020–2022 drought.

The River and the Aquifer Are Not Separate Accounts

The second problem is that the line of credit and the pension fund are not actually separate. In accounting terms, they are related parties. Draining one directly depletes the other.

This has been understood for some time. A 2016 U.S. Geological Survey study estimated that, on average, about 56% of the streamflow in the Upper Colorado River Basin originates as groundwater that seeps into streams and rivers, known as baseflow. Snowmelt does not simply run off the mountains into the river. Much of it first soaks into the ground and reaches the river months or years later. When groundwater levels in the Upper Basin fall, less of that water makes its way into the Colorado, and less arrives at Lake Powell.

Susanna Werth, a Virginia Tech co-author, made the point directly. Losses underground could affect rivers and reservoirs, with “lots of consequences for farmers and water users, even if they don’t pump groundwater.”

This has an uncomfortable implication for the argument I made in my earlier article. The structural shortfall of roughly 4 MAF per year between Compact allocations and actual flows was calculated from historical streamflow. If the groundwater that sustains a large share of that streamflow is itself declining, then future flows are likely to be lower than the historical average suggests. The insolvency of the river may be understated.

It also complicates the current political settlement. Under the post-2026 operating guidelines finalized by the Interior Department in August, the Lower Basin states must cut about 1.25 MAF per year in 2027 and 2028, with Arizona absorbing the largest share at roughly 760,000 acre-feet. The Upper Basin faces no mandatory reductions. That allocation implicitly assumes the Upper Basin is the stable source of supply. The groundwater data suggest that the source itself is weakening.

Everyone Is Counting the Same Backup

The third problem is the one most directly relevant to corporate risk managers. Groundwater is not just a backup for one company. It is the backup for almost everyone in the basin, at the same time.

Think of it as a bank run. In normal times, each bank in a system holds a modest liquidity buffer and relies on access to a shared emergency credit facility in a crisis. Each bank’s individual contingency plan looks perfectly reasonable. But if every bank plans to draw on the same facility at the same moment, the plans are only as good as the facility’s total capacity. Each plan is individually rational and collectively impossible.

This is the situation in much of the Colorado River Basin. Farms facing surface water cuts plan to pump more. Cities with Colorado River contracts hold groundwater as a drought reserve. Data centers and semiconductor fabs sited in Arizona point to groundwater access in their water risk assessments. Developers rely on groundwater to demonstrate long-term supply. Each of these contingencies may look sound on its own terms. Together, they are drawing on the same declining aquifers.

Arizona has already seen what happens when the backup is examined closely. In June 2023, the Arizona Department of Water Resources released a groundwater model for the Phoenix Active Management Area projecting about 4.86 MAF of unmet demand over 100 years. As a result, the state stopped issuing new Certificates of Assured Water Supply for subdivisions in those sub-basins that proposed to rely on groundwater. Some legal analysts argued the headlines overstated the problem, and existing water providers with assured supply designations were largely unaffected. But the episode demonstrated how quickly an assumed asset (”there is groundwater available”) can become a constrained one once someone actually does the accounting.

Where This Shows Up on the Balance Sheet

In my earlier article, I noted that water rights are typically carried on corporate books as permanent assets with no discount for junior priority, and that operating costs are modeled at current water prices rather than post-2026 prices. Groundwater adds several further blind spots:

• Contingency plans that assume groundwater availability. Many corporate water risk assessments treat groundwater as the fallback for surface water curtailment. If that fallback is depleting fastest during drought, the plan’s resilience is overstated, and so is the value of any asset whose continued operation depends on it.

• Asset valuations premised on unregulated pumping. In the roughly 82% of Arizona outside active groundwater management, pumping has historically faced few limits. Agricultural land, industrial sites and data center campuses in those areas may be valued as if that access is permanent. As depletion accelerates and political pressure grows, that assumption is increasingly fragile.

• Development and real estate contingent on assured supply. Housing and commercial development in Arizona must demonstrate a 100-year water supply. Where that demonstration relies on groundwater, the Phoenix model shows how quickly the regulatory answer can change.

• Rising cost of pumping and land subsidence. As water tables fall, wells must be deepened and pumping costs rise. Research published in 2025 found that parts of the Phoenix area are slowly sinking because of groundwater depletion, creating risks for buildings, canals and other infrastructure. None of these costs typically appear in forward operating models.

• Correlated exposure across the supply chain. A company may have secure water rights itself but rely on agricultural suppliers, utilities or component manufacturers that depend on the same aquifers. In a severe drought, those dependencies fail together.

In each case, the problem is the same one this series keeps returning to. The information is available: in peer-reviewed studies, in USGS data, in state groundwater models. But it has not made its way into the numbers. Groundwater access is being carried at something close to full value when it should be carried with a substantial conditionality discount that reflects correlation with drought.

What Has (and Has Not) Changed

There is some movement. Arizona’s 2023 decision to halt new groundwater-based subdivisions in parts of the Phoenix area was a meaningful acknowledgement of the limits of the aquifer. Some water providers are investing in recharge and storage programs. And the scientific picture is now much clearer than it was even five years ago, thanks to satellite-based measurement of groundwater that was not available to earlier generations of water managers.

But the broader governance picture is moving in the opposite direction from certainty. Lake Powell fell to a new record low in August, at about 22% of capacity. Nevada has sued over the federal operating plan, and on October 16, Arizona moved to join that lawsuit. The surface water fight is now in the courts, which raises the likelihood that more users will lean on groundwater while it plays out. Meanwhile, the vast majority of the basin’s groundwater remains lightly managed or unmanaged.

The result is that the one source of supply that companies have counted on to cover surface water shortfalls is the least protected and, according to the latest data, among the fastest declining.

Three Questions for Executives with Colorado River Basin Exposure

1) Does our drought contingency plan assume groundwater will be available when surface water is curtailed, and have we stress-tested that assumption against a scenario where both supplies decline at the same time?

2) Who else is drawing on the same aquifer as our operations and key suppliers, and what happens to our access if farms, cities and other industrial users all turn to it in the same drought year?

3) Are we carrying groundwater-dependent assets, including land, facilities and development rights, at values that reflect the possibility of tighter pumping regulation, rising pumping costs and declining water tables? And are we disclosing that exposure?

Sources

New and recent groundwater research

• KJZZ / Mountain West News Bureau: Colorado River Basin lost trillions of gallons of groundwater from 2020 to 2022 (Oct. 20, 2026)

• Oyedele et al., Water Resources Research (2026)

• Virginia Tech News: Colorado River drought drives unprecedented pattern of groundwater loss

• Colorado Sun: Colorado River’s hidden, below-ground reservoir is quickly shrinking (ASU study, 2025)

• USGS: More than half of streamflow in the Upper Colorado River Basin originates as groundwater

Policy and legal context

• Las Vegas Sun: Feds finalize 10-year Colorado River management framework (Aug. 22, 2026)

• Navajo Times: Interior sets 1.25M-acre-foot Colorado River cut for Lower Basin

• Bureau of Reclamation: Post-2026 Record of Decision

• States Newsroom: Arizona joins Nevada lawsuit over Colorado River cuts (Oct. 16, 2026)

• Newsweek: Lake Mead water levels as Nevada sues over Colorado River cuts

Arizona groundwater and assured supply

• Kyl Center, ASU Morrison Institute: Phoenix AMA groundwater model explainer (2023)

• Greenberg Traurig: The Phoenix AMA groundwater model — what it means and what’s next

Snell & Wilmer: Phoenix AMA has abundant groundwater (counterpoint)