The AI infrastructure trade is increasingly shifting from training toward inference as models move from development into real-world deployment across the global economy. $NBIS CEO Arkady Volozh said “most of capacity was spent on training new and new models. But eventually, there will be more and more usage of these models, more and more inference.” That is a major reason why infrastructure tied to AI cloud platforms, networking, memory, and inference optimization continues becoming more strategically important. The important takeaway is that the next phase of AI demand may not just come from frontier labs training larger models, but from millions of enterprises and applications continuously running those models at scale every single day.