Core Insights - The company has officially launched a new generation of GPU virtualization technology, which allows for the splitting of a single GPU's resources into multiple independent virtual computing units, with a minimum granularity of 10% [1] - This capability enables a single GPU to simultaneously support multiple AI workloads, significantly enhancing resource utilization in scenarios such as inference services, model development, and scientific teaching [1] - The newly released GPU virtualization technology has expanded its chip compatibility, now covering architectures from NVIDIA, Ascend, Cambricon, and others [1]