If we take the view that power becomes the ultimate bottleneck, the capex optimization constraint shifts to the one input China has in greater abundance vs US (who has relatively more chips) Then if you take the view that global token spend increasingly flows to cheap power and open weights, spot market is likely where China’s marginal impact would show up first via token dumping, esp if the future of work is mostly inference based not training (which also requires excess capacity with lower utilization rates) So the compute forward curve actually becomes an epic industrial duration battle where China competes at the front end and the US holds the long end. The question then becomes who is the dog and who is the tail, and which strategic advantage will roll down the curve faster- chips or power? I think anyone who understands the leverage of depreciation knows the answer. In other words this is a warning to US policymakers: the long end of the compute curve was never really about contracted chips security. It’s about energized megawatts and that’s basically the one duration the US is shortest vs its biggest geopolitical rival