Data Centers Risk Fines at 100 CFU/mL Cap [2026]

Data center operators scouting sites near oil and gas basins are running into a question that has nothing to do with chips, racks, or power purchase agreements: what do you do with produced water? A report from TechRadar published this week lays out the operational and regulatory hurdles facing any facility that wants to turn oilfield wastewater into cooling-tower feedstock, and the list is longer than most hyperscalers would like. Produced water is high-salinity wastewater pulled up alongside oil and gas, and it can carry naturally occurring radioactive material, known in the industry as NORM. That combination makes it attractive on paper (there is a lot of it, and it is not drawn from municipal supply) and genuinely difficult to manage in practice.

The timing is not an accident. AI data center construction has pushed water demand into the same conversation as electricity demand, and operators in gas-rich regions are now evaluating non-potable sources, including produced water and brackish aquifers, as a way to avoid straining municipal systems. The storage and treatment question sits right in the middle of that shift, and it is the part of the story that gets the least attention next to gigawatt announcements and GPU allocations.

Google · Preferred Sources

Don't miss new tech stories on Google

Add TrendinTech once in the Google app and our stories appear in your news suggestions.

Add Now

What Produced Water Actually Is, and Why It Shows Up Near Data Centers

Produced water is the wastewater that comes up out of the ground alongside oil and gas during extraction. It is typically high in salinity, far saltier than seawater in many basins, and it can contain naturally occurring radioactive material. That is the baseline description used in the reporting this week, and it is the reason produced water has historically been treated as a liability rather than a resource: operators inject it back underground, truck it to disposal wells, or pay to have it hauled away.

What has changed is the economics around data centers themselves. Facilities built to support AI workloads consume large volumes of water both directly, for cooling, and indirectly, through the water used at power plants that generate their electricity, according to a Lawrence Berkeley National Laboratory analysis of the environmental footprint of U.S. data centers. In gas-rich basins such as the Permian, operators are increasingly looking at on-site or near-site natural gas for power and non-potable water, including produced water, for cooling, so the facility is not competing with towns and farms for the same aquifer. That is the pitch. The storage and treatment reality is a separate, much harder problem, and it is the one regulators, water utilities, and engineers are now trying to write rules for.

The NORM Problem: Radioactivity Doesn’t Disappear on Its Own

Naturally occurring radioactive material is the detail that separates produced water from ordinary reclaimed wastewater. NORM concentrations vary by basin and by well, and they do not get diluted away just because the water is being repurposed for industrial cooling instead of being injected into a disposal well. Any treatment process that removes dissolved solids from produced water, including advanced thermal or membrane-based treatment, tends to concentrate whatever NORM is present into the leftover brine or solid residue, rather than destroying it.

That is why the proposal described in this week’s reporting is explicit about location. The treatment stages where NORM becomes concentrated should happen off-site, at an oilfield gathering station or a dedicated industrial midstream facility, not inside the data center’s fence line. Only water that has already been through that treatment and concentration step should ever be piped onto the data center campus. It is a sequencing argument as much as a technical one: keep the radioactive handling where oilfield operators already have permits, trained staff, and disposal pathways, and let the data center receive a finished product rather than a hazard.

That division of labor also changes who holds the liability. An oilfield midstream operator handling NORM-bearing residue is working within an existing regulatory framework built for exactly that material. A hyperscale data center operator taking on the same residue on campus would be stepping into radiological handling requirements that have nothing to do with servers or racks, which is a large part of why the off-site model is being floated as the default rather than an edge case.

Zero Liquid Discharge: Using Waste Heat to Close the Loop

The technical centerpiece of the proposal is thermal desalination powered by the data center’s own waste heat, aimed at zero liquid discharge, or ZLD. In a ZLD system, the facility does not treat produced water once and then release a treated effluent back into a river or municipal sewer. Instead, the system is designed so that essentially all of the water is either recovered for reuse in cooling or concentrated down into a solid or near-solid residue that gets hauled away, with nothing liquid leaving the site as wastewater.

Pairing that with data-center waste heat is the part that makes it more than a theoretical exercise. Modern AI data centers reject enormous quantities of low-grade heat from chip cooling loops, and thermal desalination processes, such as multi-effect distillation, need exactly that kind of heat input to evaporate and recondense water at scale. Using heat the facility would otherwise be dumping into the atmosphere to drive the desalination step is the efficiency argument behind the whole design: it turns a cooling byproduct into a treatment input, rather than running a separate, fully powered desalination plant alongside the data center.

Whether that pairing is economical at scale is still an open question. ZLD systems are capital-intensive and have historically been reserved for situations where regulators leave no other disposal option, such as certain power plant and refinery wastewater streams. Applying the same approach to produced water at data center scale, continuously, for the life of a facility, is a different cost equation, and it is one the industry has not yet tested in a public, operating project.

The Water Quality Numbers a Reuse System Has to Hit

Regardless of where the source water comes from, once it is inside a data center’s cooling loop it has to meet the same microbiological and chemical thresholds that water reuse specifications already set for reclaimed water in direct evaporative cooling. A water-reuse specification cited in the reporting lays out concrete numbers rather than vague guidance, and they are strict enough that they function as a real engineering constraint on any produced-water project, not just a regulatory formality.

Parameter	Required Threshold	Where It Applies
Disinfectant residual (pipeline)	0.5 mg/L minimum	Distribution pipeline carrying treated water
Free-chlorine residual	0.2 mg/L minimum	Inside direct evaporative-cooling (DEC) units
Turbidity	Below 3 NTU (max 5 NTU in 95% of samples)	Treated water supply
Total suspended solids	Below 5 mg/L	Treated water supply
E. coli	Nondetectable in 75% of samples; no sample above 25 CFU/100 mL	Treated water supply
L. pneumophila	Below 100 CFU/mL	Storage tanks and cooling units
Disinfectant monitoring	Continuous, with alarm below 0.2 mg/L	Entire reuse system

Those thresholds come from a water-reuse specification referenced in the Association of California Water Agencies and EPA webinar materials on data center water reuse. They were written with reclaimed municipal wastewater in mind, but they are the same bar any produced-water treatment train has to clear before the output can be called safe for cooling-tower use. The continuous-monitoring requirement is particularly relevant for produced water specifically, because salinity swings and intermittent well-stream chemistry mean influent quality can change faster than it would with a steady municipal reclaimed-water feed.

Why Evaporative Cooling Makes the Problem Worse, Not Better

It would be easy to assume that evaporative cooling, which is the dominant water-cooling method at large data centers, is a neutral step in this process. It is not. Evaporative cooling works by letting water evaporate off, which pulls heat out of the system, but it also concentrates whatever salts and contaminants were in the water to begin with, because only the pure water molecules leave as vapor. Everything dissolved in that water, including the salts that make produced water saline in the first place, stays behind and builds up in the remaining liquid.

That concentration effect is exactly why treating produced water and then returning any of it to a watershed becomes harder and more expensive the longer it cycles through evaporative cooling. Each pass through the cooling tower raises the dissolved-solids concentration in the blowdown water that eventually has to be disposed of or further treated. For a municipal reclaimed-water feed that starts out relatively low in salinity, that is a manageable cost. For a produced-water feed that starts out high in salinity, it compounds a problem the facility already had before the water ever touched a server.

This is part of the reason some researchers studying produced water reuse in basins such as the Permian have recommended dry cooling, meaning air-based cooling that does not consume water at all, as the more defensible default for industrial operations in water-stressed regions, based on discussion from a University of Texas at Austin Bureau of Economic Geology water resources podcast on beneficial use of produced water. Dry cooling sidesteps the entire storage, treatment, and NORM-handling chain by never putting water into the loop in the first place, though it typically comes with efficiency and siting trade-offs that liquid cooling does not have.

Storage: The Part of the Problem Nobody Wants to Own

Treatment gets most of the engineering attention, but storage is where liability tends to concentrate. Storage tanks holding produced water, or water that originated as produced water, have to be monitored on an ongoing basis for both E. coli and Legionella, because standing water in tanks and cooling units is exactly the environment where bacterial growth accelerates if disinfectant residuals drop. That is why the continuous monitoring requirement with an alarm threshold below 0.2 mg/L of free chlorine is written into the reuse specification rather than left as a periodic sampling exercise.

There is also a siting question layered on top of the water-quality question. If the NORM-concentrating treatment stages stay off-site at a gathering station or midstream facility, as the proposal in this week’s reporting describes, then the data center’s own storage tanks only ever hold treated water. That keeps the facility’s on-site radiological exposure low, but it also means the data center is fully dependent on the midstream operator’s treatment consistency. Any upset at the off-site treatment facility, whether a NORM excursion or a disinfection failure, becomes the data center’s problem the moment the water crosses onto its property, which is why contractual water-quality guarantees between data center operators and midstream treatment providers are likely to become as important as the power purchase agreements that already govern electricity supply.

Who Ends Up Holding the Compliance Burden

In practice, that split creates three separate compliance owners: the upstream oil and gas operator producing the water, the midstream facility treating and concentrating it, and the data center receiving and storing the finished product. Each of those parties is currently governed by different regulatory bodies, oilfield waste rules for the first two, and state or local water-reuse and public-health rules for the third, which is a large part of why no state has yet published a single, unified produced-water-to-data-center standard. The specification numbers cited above are the closest thing available today, and they were written for reclaimed municipal water, not produced water specifically.

Cooling Method Comparison: Where Produced Water Fits

Produced water reuse is not competing against a blank slate. It is one option among several that data center operators already weigh against each other depending on site location, climate, and water availability. The table below lays out how produced-water reuse with zero liquid discharge compares against the two cooling approaches it is most often weighed against.

Cooling Approach	Water Source Dependency	Primary Technical Risk	Regulatory Complexity
Municipal reclaimed water + evaporative cooling	Depends on local wastewater treatment capacity	Salt concentration in blowdown over repeated cycles	Established state water-reuse frameworks
Dry (air) cooling	None; no process water required	Lower cooling efficiency in high-ambient-temperature climates	Minimal; no water-reuse permitting needed
Produced water + thermal desalination + ZLD	Depends on access to oilfield gathering infrastructure	NORM handling, salinity swings, treatment consistency	Fragmented across oilfield waste and water-reuse rules

The comparison helps explain why produced water is being discussed now rather than five years ago. It is not that the water chemistry changed. It is that AI-driven data center construction has made water availability a siting constraint in regions, like parts of the Permian Basin, where produced water happens to be abundant and freshwater is not. Reports on Tetra Tech’s modeling of water and power demand for data center cooling point to a similar conclusion: resilient water strategies for large facilities increasingly combine new supply sources with storage and reuse rather than relying on a single municipal connection.

Historical Context: From Freshwater Fights to Oilfield Wastewater

Data centers and water have been a recurring friction point for years, well before produced water entered the conversation. Communities from Georgia to the Netherlands have pushed back on new data center campuses over concerns about groundwater withdrawals, and a January 2026 fact sheet from the Environmental Law Institute describes how facilities have proposed or used groundwater withdrawals directly, or as backup, to meet cooling needs. That history is the backdrop against which produced-water reuse is being pitched: not as a novelty, but as an alternative to the freshwater and groundwater fights that have already slowed or blocked data center projects elsewhere.

The Congressional Research Service has also tracked this shift at the federal level, noting that how local water resources are affected by data center withdrawals depends heavily on local water conditions and competing demands. That framing, water impact as a local and regional variable rather than a fixed cost, is exactly why gas-rich basins with large produced-water volumes look different to a site-selection team than a data center campus proposed next to a municipal reservoir already serving a growing metro area. It is the same reasoning that has pushed companies like Taiwan’s ProLogium toward AI data center supply contracts as traditional end markets soften, and it mirrors how cloud infrastructure teams managing compute costs across ECS and EKS deployments are increasingly forced to treat physical-site constraints, not just software architecture, as a cost driver.

Market Impact: What This Means for Site Selection and Capex

For hyperscalers and colocation operators, the produced-water pathway adds a new line item to site-selection models that have historically weighed power availability, fiber connectivity, and tax incentives far more heavily than water sourcing. A facility that can credibly point to an on-site or near-site non-potable water supply, rather than a request to draw from a stressed municipal system, has an easier permitting conversation with local governments, and permitting speed has become one of the biggest variables in how fast a gigawatt-scale AI campus can actually come online.

That said, the capital cost of building a thermal desalination and ZLD system, sized for continuous operation rather than periodic use, is not trivial, and it has to be weighed against simply paying for municipal reclaimed water or switching to dry cooling. Operators managing multi-cloud infrastructure with tools like Terraform alongside Kubernetes 1.37 are used to optimizing for marginal compute cost; physical water infrastructure decisions operate on a completely different timeline, with treatment plants and pipeline easements that take years to permit and build, not sprints. The near-term market impact is likely to be concentrated in a handful of gas-rich regions, rather than becoming a general industry pattern, at least until a pilot project demonstrates the approach works continuously at scale.

Regulatory Risk: No Single Agency Owns This Yet

One of the clearest takeaways from this week’s reporting is that there is no single regulatory home for produced-water-to-data-center projects. Oilfield waste handling falls under state oil and gas commissions. Radiological material handling intersects with separate NORM-specific rules that vary significantly by state. Water reuse for industrial cooling falls under state environmental or water-resources agencies, using specifications like the one detailed in the table above. A project that spans all three has to satisfy all three regulators independently, and none of them was written with this specific use case in mind.

That regulatory fragmentation is itself a market signal. It means the first operators to successfully deploy produced-water cooling at scale will have effectively written the compliance playbook that everyone else follows, similar to how early cloud cost-governance tools shaped how teams using Kubecost and OpenCost standardized Kubernetes spend monitoring before it became an industry norm. Whoever gets there first on produced water will likely have outsized influence on whatever formal standard eventually emerges.

Competitive Landscape: Who Is Positioned to Move First

The operators best positioned to pilot produced-water cooling are the ones already building in gas-rich basins for power reasons, not water reasons. Behind-the-meter natural gas generation and produced-water access tend to show up in the same geography, since both come from the same oil and gas infrastructure. That overlap means the decision to co-locate a data center near gas production for electricity supply, a trend already reshaping where gigawatt-scale campuses get built, naturally puts produced water on the table as a secondary resource, even if it was not the original reason for the site choice.

Midstream oilfield service companies are the other side of this equation, and they have an incentive to move quickly. A produced-water stream that used to be a pure disposal cost, trucked to injection wells, becomes a revenue-generating feedstock if a nearby data center is willing to pay for treated output. That economic reversal, waste becoming a sellable input, is the same dynamic that has reshaped parts of the hardware supply chain, including how companies pivoting away from softening end markets, such as battery makers redirecting output toward AI data center demand, have found new revenue by serving the data center buildout rather than their original customer base.

Predictions: Where Produced-Water Cooling Goes From Here

Expect the first publicly disclosed pilot projects to appear in the Permian Basin and other Texas gas-rich regions, where produced-water volumes and data center construction are both already concentrated.
State environmental agencies will likely issue produced-water-specific reuse guidance within the next regulatory cycle, rather than continuing to apply reclaimed-municipal-water specifications by default, once a pilot project forces the question.
Off-site NORM treatment, at oilfield gathering stations rather than on data center campuses, will probably become the default architecture other proposals copy, since it keeps radiological handling inside an existing regulatory lane.
Dry cooling will remain the more common choice for new data centers outside gas-rich basins, since it avoids the storage, treatment, and monitoring burden entirely, even where it costs some efficiency.
Expect water-quality guarantees between data center operators and midstream treatment providers to become a standard contract term, mirroring how power purchase agreements became standard once behind-the-meter generation took off.

The Engineering Checklist: What a Compliant System Has to Include

Pulling the technical requirements described in this week’s reporting together, a produced-water cooling system that satisfies current reuse specifications needs each of the following elements in place before it can legally and safely feed a data center cooling loop.

1. Off-site treatment and NORM concentration at a gathering station or midstream facility
2. Only treated water piped onto the data center campus
3. Pipeline disinfectant residual maintained at 0.5 mg/L minimum
4. In-unit free-chlorine residual maintained at 0.2 mg/L minimum inside DEC units
5. Turbidity held below 3 NTU (max 5 NTU in 95% of samples)
6. Total suspended solids held below 5 mg/L
7. E. coli nondetectable in 75% of samples; no sample above 25 CFU/100 mL
8. L. pneumophila held below 100 CFU/mL in storage tanks and cooling units
9. Continuous disinfectant-residual monitoring with an alarm below 0.2 mg/L
10. Thermal desalination stage sized to data-center waste-heat output, targeting zero liquid discharge

That is a long list for a resource that was, until recently, treated purely as a disposal problem for oil and gas operators. It also explains why no operator has yet claimed to be running this system at full scale: every item on that list requires coordination between parties that have not historically had to work together, from oilfield midstream engineers to data center facilities teams managing systems similar to how cloud infrastructure teams coordinate across AWS Lambda, Azure Functions, and Cloud Run deployments, except here the coordination problem is physical pipes and tanks rather than API calls.

What to Watch Next

The next concrete signal will be whether any named data center operator or midstream company announces a produced-water pilot with public specifications, rather than the general proposal-level discussion that defines the current reporting. Until that happens, treat produced-water cooling as a credible, technically grounded option that is still unproven at commercial data center scale, not as a solved problem. The regulatory fragmentation described above, combined with the NORM handling question, means the timeline for broad adoption is likely measured in years, not quarters, even in the gas-rich regions where the economics look most favorable today.

Frequently Asked Questions

What is produced water, exactly?
Produced water is high-salinity wastewater that comes up out of the ground alongside oil and gas during extraction. It can also contain naturally occurring radioactive material, or NORM, depending on the basin and well.

Why would a data center want to use produced water for cooling?
Data centers in gas-rich regions are looking for non-potable water sources so they do not compete with municipal and agricultural users for the same freshwater supply. Produced water is abundant in basins like the Permian, where many new AI data centers are also being sited for power reasons.

Is it safe to store radioactive produced water on a data center campus?
The proposal described in current reporting explicitly avoids that scenario. It calls for NORM-concentrating treatment to happen off-site, at an oilfield gathering station or dedicated midstream facility, so only already-treated water is ever piped onto the data center property.

What is zero liquid discharge, and why does it matter here?
Zero liquid discharge, or ZLD, is a treatment approach where essentially all water is either recovered for reuse or concentrated into a solid or near-solid residue, with no liquid wastewater leaving the site. It matters for produced water because it avoids the cost and permitting complexity of discharging treated effluent back into a watershed.

What water quality standards does treated produced water have to meet?
Current water-reuse specifications require a 0.5 mg/L minimum disinfectant residual in the pipeline, a 0.2 mg/L minimum free-chlorine residual inside direct evaporative-cooling units, turbidity below 3 NTU, total suspended solids below 5 mg/L, nondetectable E. coli in 75% of samples, and L. pneumophila below 100 CFU/mL in storage tanks and cooling units.

Does evaporative cooling make produced water harder to manage?
Yes. Evaporative cooling concentrates salts and other contaminants as pure water evaporates off, which makes any eventual treatment and return of water to a watershed more difficult and more costly, especially starting from an already-saline produced-water feed.

Is any data center currently storing produced water today?
Current reporting does not identify a named facility already doing this at commercial scale. The proposal and the associated treatment architecture are being discussed as a forward-looking approach, not a documented, operating deployment.

Could dry cooling avoid this problem entirely?
Yes. Dry, or air-based, cooling does not require process water at all, which sidesteps the storage, treatment, and NORM-handling chain altogether. Researchers studying produced water reuse in the Permian Basin have pointed to dry cooling as the more straightforward option for water-stressed industrial sites, though it typically trades away some cooling efficiency, particularly in hot climates.

Data Centers’ Produced Water Caps Risk at 100 CFU/mL [2026]