We develop and estimate a structural model of competition between banks and non-banks in the UK unsecured SME lending market under asymmetric information. Because risk-based capital regulation applies exclusively to banks, it ties a bank's loan capital charges directly to its internal risk assessment, creating a buffer-dependent shadow cost that enters bank loan prices. We document a novel empirical pattern in SME credit supply: bank lending behavior shifts sharply at loan-size caps corresponding to application channels, transitioning from automated online screening on small loans to human-mediated evaluation on large ones, whereas non-bank lenders exhibit no such channel segmentation. Combining loan-level originations with confidential supervisory data, we estimate lender-specific cost structures and screening precisions, allowing bank precision to vary across application channels. We find that banks screen less precisely than non-banks on smaller online loans, but overcome this disadvantage on larger loans routed through relationship managers. Counterfactual policy simulations show that tightening minimum capital requirements by 2 percentage points increases the average bank-originated loan rate by roughly 83 basis points and reduces the bank share of originations by 7.6 percentage points.