This article examines the potential of Wasserstein Generative Adversarial Networks (WGANs) for generating tabular energy data for high performance computing (HPC) nodes. Generating data that retains statistical properties similar to real data is especially valuable in scenarios where data access is costly. In this regard, information on the usage of HPC nodes in data centers is expensive to collect, mainly due to the high security levels maintained in these infrastructures. Four HPC nodes are studied through three different types of estimation models. Results indicate that the generated data preserve the statistical properties of the real data when evaluated using relevant techniques for distribution similarity analysis. Moreover, the model estimation quality remains at acceptable levels when replacing real data with generated data. In fact, improvements are observed when using generated data in combination with real data.