Coupling strategies in Earth System Models significantly influence their computational performance and resource efficiency. While coupling costs are traditionally evaluated in the context of concurrent implementations, the cost implications of sequential coupling approaches remain poorly quantified. In this work, we define and formalize the sequential coupling cost (<em>C</em><sub>seq</sub>): the additional computational resources required for a sequentially coupled ESM to achieve performance parity with an optimized concurrent setup.</p> <p>We develop an analytical framework that utilizes individual component scalability profiles, derived directly from model execution logs, to diagnose inefficiencies inherent in sequential coupling architectures. To demonstrate the versatility of the metric, we apply it to three structurally distinct coupling scenarios, using models widely used by the community: atmosphere–ocean (IFS–NEMO), ocean–sea ice (NEMO–SI3), and atmosphere–aerosols (IFS–M7).</p> <p>Our analysis reveals that sequential coupling imposes a substantial, yet often overlooked, efficiency deficit. By forcing all components to share a fixed resource pool, this approach ignores potential component-level parallelism and creates an exclusive reliance on domain-decomposition as a scaling strategy. While scaling solely via domain decomposition is sustainable in linear scaling regimes, it accelerates the loss of efficiency as soon as components enter sub-linear scaling regimes. We demonstrate that to achieve target throughputs, sequential setups often necessitate significant over-provisioning of computational resources, leading inefficiencies that traditional metrics fail to capture. The proposed <em>C</em><sub>seq</sub> metric quantifies these structural overheads, offering a quantitative basis for informing the design and architectural choices of next-generation exascale Earth System Models.