What a nitrate settlement teaches about industrial water risk

On March 31, 2026, Amazon agreed to pay $20.5 million to settle its part of Pearson v. Port of Morrow, a federal class action over nitrate contamination in Oregon's Lower Umatilla Basin, and in the same breath denied contributing to the problem at all. The company said it settled to "focus our time and resources on supporting the community rather than on litigation." Lead plaintiff attorney Steve Berman took the money and kept the knife out: "We appreciate Amazon taking the first step toward solving the nitrate pollution problem, but the work is far from over."

Sixteen defendants remain in the case.

Here is what makes the Morrow County record required reading for anyone siting water-hungry industry: nothing in it happened at a zoning hearing. Amazon opened its first Morrow County data center in Boardman in 2011, expanded into Hermiston in Umatilla County from 2017 onward, and built up to 13 facilities in and around the basin, per OPB's reporting. The approvals came. The buildings went up. The entitlement risk that finally cost $20.5 million never attended a single public hearing, because it was sitting underground the entire time, in an aquifer that is the sole water source for roughly 45,000 people.

How does a fully approved corridor end up in federal court?

Slowly, and then in one filing. Oregon state agencies had acknowledged the Lower Umatilla Basin needed cleanup more than 30 years before this decade, per OPB and the Food and Environment Reporting Network. The contamination predates data-center construction and reflects a mix of agricultural and industrial contributors, which is precisely why the case swept in food processors and farms like Lamb Weston, Madison Ranches, and Threemile Canyon Farms alongside the Port and Amazon. Then in 2024 the Oregon Health Authority formally identified the crisis: at least 634 domestic wells testing above federal nitrate limits, some at nearly ten times the federal limit. The class complaint followed: Michael Pearson and his co-plaintiffs filed in U.S. District Court, seeking recovery for tens of thousands of basin residents.

Read that sequence as a screening problem and the uncomfortable part is how legible it was. A 30-year paper trail of state acknowledgment is a pre-built plaintiff exhibit list. A defined geography where every household drinks from the same aquifer is a textbook class-certification pattern. When data-center cooling water returns to a regional aquifer that agriculture also loads with nitrogen, any nitrate signal gets attributed across every industrial user in the basin, and the deepest pockets get named first. None of that required a crystal ball. It required someone to read the water record with the same care the market reads the power record.

And the local politics kept running the other direction the whole time. The Hermiston City Council voted in 2025 to annex 800 acres specifically to attract more data-center investment, during a state-identified contamination crisis, while the litigation was live. That divergence between a city's economic-development posture and its state regulator's public-health posture is not a curiosity. It is the seam plaintiff firms work, because it produces exactly the record of "everyone knew and expansion continued" that settlements are made of.

What did the riverkeeper actually change in Georgia?

Swap coasts and the mechanism changes but the lesson does not. Hyundai's Metaplant in Bryan County, Georgia, a $5.54 billion EV and battery complex on 2,906 assembled acres, never had a meaningful rezoning fight; the Joint Development Authority megasite came pre-entitled, with governor-level sponsorship from the May 20, 2022 announcement onward. The contest arrived one permit later. Georgia EPD issued a groundwater withdrawal permit authorizing roughly 6.625 million gallons per day from the Floridan Aquifer, against an expected plant draw around 4 million gallons per day, and the Ogeechee Riverkeeper, a Statesboro-based nonprofit with a real legal docket, appealed it.

The target was not the volume. It was the clock. The permit gave the project 25 years to bring an alternate, non-aquifer water source online, and the Bryan County record shows the appeal centering on that timeline and on the sufficiency of aquifer monitoring. The January 2025 settlement, reported by Georgia Public Broadcasting, compressed the replacement-water deadline from 25 years to 15 and added structured water-quality monitoring; the consent agreement was executed April 25, 2025, and the appeal was withdrawn. A nonprofit with standing took a decade out of a multibillion-dollar site's water runway without winning a single courtroom ruling. That spring brought a coda: Georgia EPD fined the plant $30,000 for construction-phase wastewater violations, per The Current GA's May 2025 reporting, putting an enforcement entry in the same public file the next permit renewal will be read against.

The two cases rhyme in one specific way. In Oregon, a settle-and-deny payment created a financial template that plaintiff economics will extend to comparable basins. In Georgia, a consent agreement became the governing document for the site's water operations for the next 15 years. Neither instrument is a zoning approval, and both now bind harder than most zoning conditions do.

Water risk does not attend the hearing.

How do you price water before it prices you?

Start by moving water out of the utilities appendix and into the risk section of the memo, alongside power and process. For industrial and mega-manufacturing sites, the record here suggests three concrete checks. Trace where the cooling or process water actually goes after you use it, because a return path into a shared regional aquifer is a shared-liability path, and closed-loop design has quietly become the political floor in stressed basins rather than a differentiator. Pull the state agency's full historical file on the basin rather than the current permit posture alone; 30 years of acknowledged contamination in Oregon and a named, regionally regulated aquifer in Georgia were both sitting in public archives before ground broke. And profile the opposition's institutional form: a riverkeeper with a documented appeal record is a different underwriting problem than diffuse neighborhood sentiment, because it can convert a permit into a negotiation years after the ribbon-cutting.

The Amazon settlement's arithmetic deserves one last look. $20.5 million buys wells and water treatment for affected residents; it does not buy an admission, and it does not end the case, since 16 defendants are still litigating and discovery in a multi-defendant basin case will keep producing public-record disclosures for years. Every one of those disclosures becomes political input to the next data-center or industrial application in the corridor. Settlement did not close the record. It funded its continuation.

I expect the Pearson template, a basin-wide class action naming the deepest industrial pocket alongside agricultural co-defendants, to get filed against at least one other data-center or mega-manufacturing corridor by the end of 2027, most plausibly in a Western basin where cooling water and irrigated agriculture share the same aquifer. If that filing never comes, I overread this record.

Until then, ask one question about your next water-hungry site before anyone drafts the traffic study: who else drinks from this aquifer, and what did the state write about it 20 years ago? The answer is in a file. Someone should read it before the plaintiffs do.

This analysis is a source-cited research summary drawn from public records, not legal advice. It can contain errors and should be verified independently before any investment decision.