Key Takeaways

  • Residents questioned how the $16B project could affect the Saline River, groundwater, energy supplies, noise levels, wetlands, and local infrastructure.
  • Michigan’s review could involve wastewater, water withdrawals, stormwater, air emissions, and other permits, depending on the final design.
  • The dispute highlights a broader business risk for AI infrastructure developers as communities seek clearer facility-level environmental data.

Residents used a virtual town hall held by the Michigan Department of Environment, Great Lakes, and Energy to scrutinize the proposed $16B OpenAI-Oracle data center in Saline Township. Wastewater discharge into the Saline River emerged as a central concern, but the discussion ranged well beyond a single permit.

Participants also raised questions about groundwater withdrawals, electricity demand, noise, air emissions, wetlands, and pressure on local infrastructure. Together, those issues illustrate how quickly a large AI computing project can become a regional planning matter rather than a conventional industrial development.

The immediate regulatory focus involves wastewater associated with onsite water treatment. Any discharge governed by the federal Clean Water Act would typically operate through the National Pollutant Discharge Elimination System, or NPDES. Permit conditions can address where treated water goes, which pollutants are monitored, acceptable limits, testing frequency, and reporting.

That distinction matters. A permit does not amount to a broad declaration that the entire OpenAI-Oracle development has no environmental effects. It covers a defined activity under a particular law. Other parts of the project may fall under separate state or local processes.

According to Michigan EGLE, its potential oversight of data centers can include large water withdrawals, wastewater, wetlands, stormwater, air emissions, and public drinking-water systems. The project’s design and site conditions will determine which approvals apply. Local governments may separately address zoning, roads, emergency services, construction impacts, and noise.

Michigan’s Natural Resources and Environmental Protection Act also provides a framework for reviewing significant groundwater use. Part 327 requires large withdrawals to avoid adverse resource impacts. For residents, however, technical compliance may be only the starting point. They are also asking how sustained demand could affect neighboring wells, waterways, and seasonal water availability.

AI facilities concentrate several resource questions in one place. High-density computing hardware consumes substantial electricity and produces heat. Removing that heat may require water, depending on the cooling architecture, while electricity generation can add an indirect water footprint far beyond the facility boundary.

United Nations University findings reported by GovTech estimated that global data centers consumed about 448 TWh of electricity in 2025, an amount comparable to the demand of the world’s 11th-largest national electricity user. Consumption could exceed 945 TWh by 2030.

While global figures do not predict the environmental performance of one site, residents and regulators increasingly want facility-specific information, including expected peak power demand, annual water use, cooling technology, wastewater composition, backup generation, and plans for drought or grid constraints.

Cooling choices could become especially important. Microsoft, Google, and Amazon are among the operators developing lower-water or closed-loop approaches, although performance and disclosure vary by facility and ownership arrangement. For OpenAI and Oracle, explaining the selected system in practical terms could help address uncertainty. Is water recirculated? How much is lost through evaporation? What treatment chemicals could enter the waste stream?

Energy presents a parallel challenge. New transmission equipment, substations, backup generators, and utility capacity can influence costs and timelines. Developers may secure enough power contractually while still facing questions about grid upgrades, emissions, and who ultimately pays for supporting infrastructure. That said, the answers depend heavily on utility agreements and the project’s operating profile.

For enterprise technology leaders, the Saline debate is a reminder that compute capacity is increasingly tied to environmental disclosure and community acceptance. Capital budgets and construction schedules can be affected by incomplete water data, fragmented ownership, or unclear responsibility among developers, operators, cloud customers, and utilities.

The EGLE process will not settle every question residents raised. It can, however, establish enforceable conditions within the agency’s authority and create a public record around the proposed wastewater system. OpenAI and Oracle now face a broader commercial task as well: showing that a $16B AI infrastructure investment can operate within measurable local limits, not merely meet growing demand for computing power.