This study focuses on finding the impact of the Covid-19 pandemic on the cost efficiency of banks using Stochastic Frontier Analysis (SFA). Building on intermediation and risk-return theories, we contend that credit risk, profitability, capital buffers and liquidity management have different impacts on costs in crisis and normal regimes. Using SFA, we estimate a translog cost frontier and Battese-Coelli (1995) inefficiency-effects model on a balanced panel of the quarterly data of Indonesian banks over the period 2015-2024. Inefficiency is modeled using Non-Performing Loan (NPL), Return on Asset (ROA), Capital Adequacy Ratio (CAR), Loan-to-Deposit Ratio (LDR), a dummy of the Covid period (2020-2022) and interaction terms of structural change. Results show that average cost efficiency is high but significantly lower during the lives of the Covid compared to outside the pandemic and conventional banks are seen to be more efficient than Islamic counterparts, Muamalat being the least efficient. Crucially, determinant effects are state-dependent: before the pandemic NPL, ROA, and CAR are associated with less inefficiency and LDR with more inefficiency, whereas during the pandemic such associations become weaker or become opposite and LDR/FDR becomes efficiency-improving. Theoretically, we provide proof of a regime shift mechanism in efficiency drivers under disruption, which means that models of performance should take crisis interactions into account when testing banking efficiency and cross-model heterogeneity.