Cloudspecs: Cloud Hardware Evolution Through the Looking Glass [pdf] (www.cs.cit.tum.de)

A recent paper from researchers at Technische Universität München, titled "Cloudspecs: Cloud Hardware Evolution Through the Looking Glass," offers a comprehensive analysis of cloud hardware trends from 2015 to 2025. The study focuses on the extensive variety of virtual machine instance types offered by major cloud providers like AWS, Microsoft Azure, and Google Cloud, which now number over 1,000 configurations. The authors reveal that while the variety has increased significantly, CPU and DRAM performance gains have stagnated, presenting new challenges in designing cost-efficient cloud-native systems. Additionally, the analysis highlights competitive pricing strategies of smaller cloud providers, which can undercut larger hyperscalers by up to 5x for commodity virtual machines. This research is significant for the AI/ML community as it underscores the importance of cost-efficiency in cloud computing, particularly for data-intensive applications that rely heavily on varying cloud configurations. By introducing an interactive visualization tool for exploring these trends, the authors aim to equip cloud system architects with the insights needed to optimize instance selection based on their performance and cost metrics. Key findings indicate that advancements in cloud hardware must be accompanied by economic considerations, as developers increasingly prioritize performance per dollar over raw throughput, potentially reshaping the deployment of AI and machine learning workloads in cloud environments.