How does that concern individual companies as AI consumers? The hyperscalers have no trouble utilizing their GPUs (and I would suspect they mathed it out to be profitable but, again, that's just a different topic).