A Wisconsin horse farm owner living 1.6 miles from Meta's $1 billion data center under construction in Beaver Dam has gone public with alarming observations about drastic changes to a creek that has flowed consistently on her property for nearly 50 years. Maily Kocinski reports that since construction began upstream, the creek has repeatedly run dry outside of rainfall events, then surged back with milky-white, opaque water capable of severe bank erosion. At times it has emitted chemical odors. Kocinski paid hundreds of dollars out of pocket for independent water testing that revealed extremely high levels of aluminum, with the U.S. Geological Survey findings consistent with heavy construction activity such as rinsing concrete trucks. Some resident accounts also reference elevated strontium and indicators of deep groundwater disruption.

This testimony, delivered at local hearings, highlights not just immediate visible pollution but potential longer-term risks to groundwater, tributaries, Beaver Dam Lake, and the broader watershed. While the documented aluminum spikes appear tied to the active construction phase rather than full data center operations, the episode underscores a pattern of environmental externalities accompanying the explosive growth of hyperscale AI data centers. These facilities require enormous volumes of water for cooling—often millions of gallons daily—with periodic flushing of closed-loop systems that can introduce chemicals. A Milwaukee Journal Sentinel analysis detailed how Wisconsin's emerging data center boom, including Meta's Beaver Dam project alongside Microsoft and Vantage facilities, could consume water on a scale rivaling or exceeding entire cities, raising supply and quality concerns statewide.

Kocinski is not alone; she was joined by over a dozen Beaver Dam residents voicing worries over transparency, water impacts, and limited early community input in project approvals. Local news coverage from TMJ4 and WISN documented these public comments, with residents describing the changes as unprecedented and 'scary.' Meta has announced the facility as its 30th data center, emphasizing AI capabilities and some wetlands restoration efforts, but direct responses to the specific water complaints remain limited in public reporting.

Zooming out through the lens of the AI boom reveals systemic issues mainstream coverage often downplays. Data centers are multiplying rapidly—from roughly 2,600 in 2021 to over 4,000 today—frequently sited in or near marginalized or rural communities already facing pollution burdens. They create localized heat islands, drive up electricity demand, and risk chemical discharges or spills if wastewater is mismanaged. Federal policy is accelerating this: a 2025 executive order streamlined permitting and reduced certain EPA reviews, while proposed legislation like the Data Center Transparency Act and Community Impact Act seeks mandatory reporting on water, energy, and equity effects but has yet to pass. Virginia and other states are experimenting with tighter rules, yet Wisconsin's PSC process initially faced criticism for excessive redactions on energy and water projections for the Meta project.

The hidden costs extend beyond one creek. Construction-phase pollution can foreshadow operational strains as these AI 'brains' scale to meet hyperscale training and inference demands. Without rigorous, independent monitoring and updated regulations that match the pace of deployment, rural water systems risk becoming collateral damage in the race for computational dominance—eroding not only streams but public trust and long-term ecological health. Kocinski's self-funded tests and testimony serve as an early warning that demands deeper scrutiny, not dismissal as isolated growing pains.