Neo4j Aura provides a DBaaS (Database As A Service) through public cloud providers. It runs on container technology and this allows for the AuraDB Instance Instance to allocate dedicated CPU resources. A CPU physically means a vCPU or CPU thread.
As Neo4j AuraDB Instances provides high available 7x24 database service. Behind the scenes, an AuraDB Instance consists of 3 database nodes Causal Cluster to support users' access. Each database node runs on its own containers.
In this case, on the AuraDB Instance METRICS page, It shows the AVERAGE CPU Usage (%) for 3 Causal Cluster nodes. During a typical access mode, the 3 nodes will have similar CPU usage.
When the cluster is hit by a heavy WRITE (for example, data loading jobs) workload, the leading node of the cluster will have a much bigger workload and much high CPU usage (might hit 100% of CPU usage). At this time from the Aura CPU usage (%) page, it might show a CPU usage of about 35%.
So in this case, it shows you the average CPU usage is well below the high CPU usage threshold, in fact, the AuraDB Instance is running lack of CPU resources. You need to consider expanding your AuraDB Instance and assign more CPU resources to it.