Huawei on Thursday launched its new Flex:ai AI container software, introducing compute-splitting technology that divides a single GPU or NPU card into multiple virtual compute units with granularity as fine as 10%. The system enables a single card to support multiple AI workloads simultaneously while pooling idle XPU resources across cluster nodes into a shared compute pool. Huawei said Flex:ai will be open-sourced in the OpenEuler community following its release.