The Finger Lakes and Watershed-Scale Computing (mattwie.se)

At a contentious Lansing, NY town board meeting, local officials delayed a vote on a development moratorium after a packed public hearing that included representatives from crypto and AI data center operator TeraWulf. The company has leased Milliken Station, a retired coal plant on Cayuga Lake, and publicly touted the site’s “industrial-scale water intake system,” robust electrical infrastructure, and redundant fiber—assets attractive for large-scale compute. TeraWulf told residents cooling won’t use lake water; the author and many locals are skeptical, noting that while air or closed-loop cooling is possible, lake-sourced cooling would materially reduce both capex and opex. This matters for AI/ML infrastructure because ultra-large data centers need cheap, reliable power and efficient heat rejection; the Finger Lakes’ deep, temperature-stable waters (Cayuga max depth 435 ft; Seneca 618 ft) and extensive watershed make them especially appealing. Converting retired plants like Milliken or Greenidge (already a Bitcoin mine on Seneca) lowers build costs and accelerates deployment, but raises environmental, regulatory, and community conflicts. With climate-driven precipitation trends uncertain but possibly increasing regional water availability, investors may double down—putting local moratoria votes and NIMBY resistance at the front line of debates over where and how future AI compute capacity is sited.