Meta's $35B CoreWeave Commitment Sharpens the Energy Cost Argument

The structural detail that shifts this beyond a capex story is the inference framing. signals that the energy draw of AI is no longer a training-phase cost that diminishes once models are built — it is a permanent operational commitment proportional to usage. Every query served, every recommendation generated, every piece of content moderated runs on infrastructure whose energy and water costs are now contractually locked in through 2032. The labs have consistently framed energy consumption as a transitional cost on the…