Dallas Fed Paper Warns Tokenised Deposits Could Shrink Bank Lending by $580bn

The Federal Reserve Bank of Dallas has published research examining how widespread adoption of tokenised deposits would affect bank balance sheets, concluding that it could meaningfully reduce banks’ capacity to fund long-term lending while forcing them to hold larger buffers of liquid assets. The authors are explicit that they take no position on whether large-scale adoption will occur; their contribution is a conditional stress test of the balance-sheet consequences if it did.

The central mechanism is deposit stickiness. In conventional banking, demand deposits behave in practice like long-term fixed-rate funding because depositors rarely move money even when rates shift elsewhere, and banks pass on only a fraction of rate changes. The paper estimates that deposits underpin roughly 80 percent of interest rate risk in the US banking system. Tokenised deposits, settling instantly and eventually switchable by AI agents with minimal friction, would erode that stickiness. The authors calculate that a 10 percent reduction in deposit duration would remove approximately $580 billion of lending capacity measured in ten-year Treasury equivalents, and would also compress banks’ appetite for interest rate risk more broadly.

A secondary effect runs through liquidity management. Around-the-clock instant outflows would make deposit balances more volatile, requiring banks to hold larger reserves and government bond buffers against unexpected withdrawals. Brazil’s Pix instant payment network offers an empirical preview: research on Pix found that heavy usage pushed banks toward more government bonds, less lending, and higher-risk loan books. The Dallas Fed suggests tokenised deposits could trace a similar arc in the United States.

The paper is a carefully reasoned conditional analysis rather than a forecast, and its mechanics are not seriously disputed. Two assumptions worth carrying into any follow-on work are that it treats deposits as a single aggregate pool and considers tokenised deposits in isolation from other structural changes, both of which the source itself flags as simplifications a careful reader should note.