Why TCO Doesn’t {Quite} Work for Comparing On-Premise and Cloud-Enabled HPC

Total Cost of Ownership (TCO) is traditionally used to determine the financial implications of owning assets like on-premise High-Performance Computing (HPC) systems by adding up all related expenses over their deployment period. However, this static financial tool becomes problematic when applied to dynamic environments like cloud-enabled HPC systems, where expenses are reactive and computing power is virtually limitless. While on-premise systems involve predictable expenses and limited computing power, cloud-based systems offer scalable resources with expenses linked to demand, allowing for greater flexibility and potentially more innovative research outcomes. The challenge arises when attempting to compare these two vastly different systems using TCO, as it fails to account for the benefits and rewards associated with cloud computing, such as access to the latest infrastructure and the elimination of queue times. This discrepancy highlights the inadequacy of TCO in evaluating cloud solutions, suggesting that a different approach is necessary to accurately assess the financial and operational benefits of cloud-enabled HPC.