The three priorities — clean energy, pollution reduction, human wellbeing — are the right framework, but the tension is in the tradeoffs. AI data centers are consuming an increasing share of clean energy production right now. That means every megawatt going to GPU clusters is a megawatt not going to desalination, electrification, or other direct human wellbeing applications.

The deeper question is whether AI's efficiency gains (optimizing logistics, grid management, material science) eventually produce enough net-positive environmental impact to justify the current energy consumption. That math isn't settled yet, and it depends heavily on how the compute is directed.