I use the neo-classical growth model to study financial intermediation in the U.S. over the past 130 years. I measure the cost of financial intermediation on the one hand, and the production of financial assets and liquidity services on the other. The model suggests that the US financial system has become less efficient over time: the unit cost of intermediation is higher today than it was a century ago. Improvements in information technology seem to have been cancelled by increases in trading activities whose aggregate social value is difficult to measure.