Pierce Freeman and Richard Diehl Martinez connect denser AI computing with the need to remove heat. They compare air cooling, evaporative cooling and liquid-cooling approaches, using familiar analogies to explain where heat goes. The practical design matters: a recirculating chip-level loop can still transfer heat to equipment that consumes water elsewhere.
Pierce Freeman and Richard Diehl Martinez examine evaporation, mineral concentration, treatment and discharge. They distinguish water used at a facility from water returned to its surroundings, while recognizing that environmental effects depend on the site's cooling system and water source. Their informal runoff examples are not measurements of a specific facility.
Pierce Freeman and Richard Diehl Martinez separate three accounting boundaries: on-site cooling, water associated with electricity generation, and training costs allocated across later inference. They argue that a tiny per-response estimate can obscure aggregate or regional pressure, especially when demand is concentrated around already-stressed resources.
Pierce Freeman and Richard Diehl Martinez question whether data-centre tax incentives deliver benefits proportionate to public costs and resource demand. They contrast large capital investments with limited permanent jobs, asking how older economic assumptions should be reconsidered as AI infrastructure scales. Their policy critique is not a finding that every incentive or development is harmful.
Watch on YouTube
