Google Cloud, Nvidia to Launch Vera Rubin NVL72 A5X Instances Next Month

The A5X launch matters less as a hardware milestone than as a signal that inference and reinforcement-learning workloads are becoming the organizing principle for cloud AI capacity, not training alone. The multi-site scale Google Cloud describes, plus a named customer already running large-scale RL research, suggests demand is being lined up ahead of general availability rather than after it. Whether the European expansion follows on the same timetable, and whether inference-focused instances become the default tier for agent workloads, is the open question worth tracking.