Competing Uses Data Centers and Agriculture

The third discussion of the 2026 Marshall & Stevens Water Forum examined how agriculture, data centers and other users fit within water systems that are already under pressure. The panel addressed the difference between water use and water stress, the importance of location in land-use decisions, whether agriculture and data centers directly compete for the same water, opportunities to use recycled water, and the regulatory and permitting approaches emerging around data-center development.

The Forum was moderated by Jennifer Simmonson, MAI, Director in the Real Estate practice at Marshall & Stevens. She was joined by Megan Nicholas Harper, Director of Water Resources at Nuveen Natural Capital; Rosemary Knight, Professor in Stanford University’s Geophysics Department and Founding Director of the Center for Groundwater Evaluation and Management; Nati Louzon, VP of Global O&M at IDE Technologies; and Bobby Majumder, Partner at FBT Gibbons and Co-Head of the firm’s Energy Group. Together, the panel brought perspectives spanning groundwater science and management, institutional agricultural investment, water technology and operations, and energy and industrial development.

Water Use Should Be Considered Alongside What It Produces

Megan Nicholas Harper began by identifying what she sees as a disconnect in discussions of agricultural water use: the focus often remains on the volume consumed without considering the value created by that water. She acknowledged that agriculture is the largest user of water in California and globally, accounting for approximately 70 to 75 percent of consumptive water use, while emphasizing that the water supports food and fiber production, jobs, export revenue and broader economic activity.

Nicholas Harper did not suggest that agricultural practices are beyond improvement. Her point was that a comparison based only on volume leaves out both the purpose of the use and the productivity achieved. She therefore reframed the question from which sector uses the most water to how efficiently and productively the water is being used.

“Agriculture also produces food, fiber, jobs, export revenue, and economic activity.” – Megan Nicholas Harper

Statewide Totals Can Conceal Local Effects

Nicholas Harper applied that distinction to growing concern about the water used by hyperscale and artificial-intelligence data centers. She said California data centers were estimated to consume roughly 10,000 acre-feet of water annually. She compared that figure with an average of approximately 200 million acre-feet of annual precipitation in California and described the statewide data-center footprint as less than 0.1 percent of water use.

She also translated the estimate into agricultural equivalents: approximately 5,000 acres of wine grapes, 4,000 acres of figs, 3,000 acres of nectarines or 3,500 acres of pistachios. Those comparisons were presented against approximately 24 million acres of California farmland, about 10 million of which are irrigated.

For Nicholas Harper, however, the statewide percentage was not the deciding measure. A data center requiring 1,000 to 3,000 additional acre-feet per year could have a meaningful effect if it were placed in an overdrafted subbasin. She compared that local demand with adding hundreds or thousands of acres of permanent plantings to an already constrained area. Her distinction was between total water use across a large geography and the stress created when additional demand enters a particular water system.

“The challenge is adding demand to systems that are already totally allocated.” – Megan Nicholas Harper

Nicholas Harper connected that concern to SGMA implementation in the San Joaquin Valley. Citing an estimate from the Public Policy Institute of California, she said more than approximately one million acres may need to come out of production to balance groundwater demand. In that setting, she said, adding hyperscale or AI facilities could further constrain the system. She described successful management as requiring all users—farms, cities and emerging industries—to operate within the resource’s long-term limits rather than selecting one sector over another.

Land Use Decisions Begin With the Natural System

Rosemary Knight approached competition among uses as a land-use question: what activity is best suited to which location? She acknowledged that development and existing land uses prevent decision-makers from starting with a blank slate. Even so, she said the underlying principle should be to consider the best use of the land based on measurements, data and an understanding of how the natural system functions.

That analysis includes the natural limits of a location, the presence of surface water and groundwater, and the connections between the two. Knight offered deep groundwater recharge as an example. Shallow recharge can increase water supply, but deeper recharge may be necessary to address subsidence associated with pumping from lower aquifers. She described satellite data showing where natural recharge moves from the surface to a deep aquifer as information that can help identify where an intervention should occur.

“It’s really a question of what should we do where? Is this the best use for this land?” – Rosemary Knight

Knight’s comments did not prescribe one use for a particular parcel. They described the information she said should guide the decision: how water moves through that location, what the natural system can support and whether the proposed use works with those conditions.

Watershed Level Analysis Can Reveal What Broad Comparisons Miss

Bobby Majumder reinforced the emphasis on location by describing watershed-level analysis. He pointed to Minnesota as an example of a state moving decisions about permits for industrial users toward local watershed authorities. In his description, those authorities can assess whether a use is complementary at a more localized level.

Majumder cautioned that national comparisons—and, by implication, other broad geographic comparisons—do not necessarily describe the conditions that exist in a specific watershed. The availability, quality and allocation of water can differ from one location to another, making localized facts important to permitting and development decisions.

“The national comparisons are not terribly helpful. They can be misleading, actually.” – Bobby Majumder

The Panel Presented Different Views of Direct Competition

When Simmonson asked whether competition between agriculture and data centers is a real concern, Majumder said he did not view agriculture and data centers as directly competing for water. He based that position on his distinction between agricultural users’ use of non-potable water and data centers’ use of potable water. He also identified a seasonal overlap: water demand rises during hot months because agriculture needs water for crop growth and data centers need additional cooling.

Majumder discussed efficiency in agricultural irrigation and cited evaporation and transpiration associated with agricultural water use. He also referred to 2023 data indicating that U.S. data centers represented approximately 0.3 to 0.4 percent of total national water withdrawals at that time. He qualified the comparison by noting that the figure preceded much of the growth in hyperscale centers.

His conclusion was that the question should not be treated as a binary decision between agricultural and industrial users. That view differed in emphasis from Nicholas Harper’s warning that a new data center can have a significant local impact when added to an overdrafted basin. The two responses addressed different aspects of the issue: Majumder focused on water type and national usage, while Nicholas Harper focused on the effect of additional demand within a constrained local system.

Reuse Can Reduce Dependence on Drinking Water

Nati Louzon extended the location discussion to the relationship between water users and water-treatment facilities. He proposed locating a data center near a wastewater-treatment plant so treated wastewater could provide a separate water source for the facility rather than requiring tap or drinking water.

Louzon also described a water-treatment plant serving multiple users with water treated to different levels according to each user’s requirements. In his account, that operating model can reduce prices and operating costs while increasing production efficiency. He framed the approach as area-wide water management rather than water management limited to an individual factory or facility.

“It will be an independent source for the data center without using tap water or drinking water.” – Nati Louzon

The example linked siting, reuse and operations. Proximity to a wastewater-treatment plant creates the physical opportunity; treatment tailored to different users creates a usable product; and shared infrastructure can serve the data center, other customers and the treatment plant itself.

Data Center Regulation Is Expanding

Majumder said regulation of data centers was accelerating. He referred to more than 260 data-center bills considered across more than 40 states in 2025, covering matters such as siting and grid connections. He also described state-level restrictions or moratoriums in New York, Texas, Maine, Vermont and Oregon.

According to Majumder, New York had paused new environmental permits for data centers over 50 megawatts for one year. He said Texas had paused approval of new grid connections pending state review of proposed projects’ power, water and ownership details. He explained that much of the initial regulatory attention was driven by concern about electricity costs, but water had also moved higher on legislators’ agendas as they considered the total demands of development.

Majumder again favored a localized approach over broad statewide directives. He referred to Minnesota legislation requiring watershed-level review and said this allows decisions to reflect facts that are more readily known at the local level.

“Rather than issuing broad dictates, we’re doing it at the more localized level, where the facts are more readily known.” – Bobby Majumder

Water Efficiency Is Becoming a Development Metric

Majumder identified water usage effectiveness as an increasingly important measure for data centers, alongside power usage effectiveness. He described it as a measure of water used per unit of information-technology energy produced by a facility.

He also said California and Arizona already require power-plant developers to evaluate water-efficiency options, including dry cooling, as part of the siting process. Within the scope of the Forum discussion, these examples illustrated how water use is entering both the performance measures applied to facilities and the regulatory reviews that determine whether and where projects move forward.

Competition Is a Question of Place, Sources and Limits

The third Forum discussion did not reduce water allocation to a statewide comparison between agriculture and data centers. Nicholas Harper distinguished total water use from local water stress and emphasized that new demand has different consequences in an overdrafted basin. Knight focused on using measurements of surface water, groundwater and their connections to determine the most appropriate land use for a location. Majumder questioned the premise of direct competition, emphasized water type and seasonality, and supported watershed-level review. Louzon described how recycled water and coordinated siting could reduce a data center’s dependence on drinking water.

The panelists did not express an identical view of the degree to which agriculture and data centers compete. Their comments instead identified the specific considerations they believed should shape decisions: what the water is used for, how efficiently it is used, the condition and allocation of the local system, the type of water required, the availability of recycled sources, and the regulatory framework governing the location.

Explore the other articles in this four-part series for additional perspectives and insights from the 2026 Water Forum.