We study the impact of higher capital requirements on banks' decisions to grant collateralized rather than uncollateralized loans. We exploit the 2011 EBA capital exercise, a quasi-natural experiment that required a number of banks to increase their regulatory capital but not others. This experiment makes secured lending more attractive vis-à-vis unsecured lending for the affected banks as secured loans require less regulatory capital. Using a loan-level dataset covering all corporate loans in Portugal, we identify a novel channel of tighter capital requirements: relative to the control group and after the shock, treated banks require loans more often to be collateralized but less so for relationship borrowers. We further nd this impact is stronger for collateral that saves more on regulatory capital.