Backplane

Asset Class: Closed Paper Mills

Closed Paper Mills: Power, Water, and Structure in One Site

Paper manufacturing is both power- and water-intensive, and mills were built accordingly — heavy electrical service, large water intake and discharge systems, and industrial structures designed to support very heavy process equipment. When a mill closes, those same fundamentals make it a compelling data center candidate.

Why mill sites line up with data center needs

Paper mills typically operate substantial on-site power infrastructure to run pulping, drying, and processing equipment around the clock — which means real grid capacity is already provisioned at the site, not something that has to be built from scratch.

Just as importantly, mills have significant water rights and infrastructure for their processes. That same water access is directly useful for data center cooling, particularly for higher-density AI deployments where liquid cooling is increasingly standard.

Other characteristics worth noting

Beyond power and water, former mill sites often bring:

  • Large, heavy-load industrial buildings with high floor-load ratings
  • Existing substations and heavy electrical distribution
  • Rail or river access for logistics
  • Rural or semi-rural locations with room to expand and fewer neighboring-use conflicts

What we help navigate

Mill sites can carry environmental considerations tied to legacy processing chemicals and wastewater handling — again, well-trodden diligence territory for industrial redevelopment, not a unique obstacle. Equipment removal and structural retrofit are typically the larger practical workstreams, and both are factored into how we structure financing for the buildout.

What this means for owners

If you're holding a closed or closing paper mill, the power and water infrastructure built for pulp and paper processing may be exactly what makes it valuable for AI infrastructure. Share the site's location, power and water access, and current condition, and we'll assess fit against real compute demand.