Dallas Fed Warns Tokenized Deposits Could Strip $700 Billion From Bank Lending Capacity
Picture a depositor who never has to wait. Their savings sit in a tokenized account that settles instantly, pays a floating rate, and can be moved to whichever bank offers a marginally better yield the moment an algorithm notices the gap. No phone call, no waiting period, no friction — just code executing a preference in real time. That scenario is not a hypothetical dreamed up by a crypto founder. It is the central worry in a new research paper out of the Federal Reserve Bank of Dallas, and the number attached to it is large enough to get a bank CFO’s attention: $700 billion.
Dallas Fed economists Rosie Levy and Srini Ramaswamy published the analysis on August 25, arguing that large-scale adoption of tokenized bank deposits — combined with agentic AI capable of automatically chasing yield on a depositor’s behalf — could materially erode US banks’ capacity to hold long-term interest-rate risk. Their estimate: if tokenization makes depositors roughly 10% more sensitive to rate differences between institutions, banks could lose approximately $700 billion in the capacity to absorb interest-rate risk tied to long-duration loans and securities. A related scenario in the same paper, where deposits become 10% more likely to leave a bank sooner than they otherwise would, puts the figure at around $580 billion.
The mechanics matter here, and they are quietly different from the stablecoin debate that has dominated crypto-banking headlines for the past two years. Tokenized deposits are not stablecoins. They remain inside the regulated banking system, can pay interest the way stablecoins generally cannot, and are represented on a blockchain purely for programmability and settlement speed — not as an alternative to bank money. That regulatory pedigree is exactly why the Dallas Fed’s warning lands differently than typical stablecoin-risk commentary: this is a paper about banks’ own product roadmap potentially undermining the deposit stickiness banks have relied on for a century of maturity transformation.
Deposit stickiness is the quiet mechanism the entire paper turns on. Banks lend long and borrow short precisely because depositors, in practice, rarely move their money the instant a better rate appears elsewhere — inertia, loyalty, and simple friction keep funds in place long enough for banks to underwrite 15- and 30-year loans against short-term deposits. Tokenized deposits threaten to strip out exactly that friction. Instant settlement plus smart contracts plus an AI agent instructed to optimize yield removes the human hesitation that has quietly subsidized long-term lending for generations.
The paper stops short of forecasting a crisis. It frames the finding as a capacity constraint rather than a prediction — banks facing more rate-sensitive, faster-moving deposits would likely respond by shifting toward safer, more liquid assets and tightening lending terms, which in turn could raise borrowing costs across the economy rather than triggering an acute shock. Still, the timing is notable: major US banks, including JPMorgan, Bank of America, Citi, and Wells Fargo, have separately been reported building shared tokenized-deposit infrastructure explicitly to compete with stablecoins for payment volume, meaning the very institutions the Dallas Fed is warning about are among the ones racing to build the technology in question.
Nothing in the paper calls for slowing that build-out down. It reads more as an early flag for bank risk officers and regulators to start modeling now, well before tokenized deposits reach anything close to the scale needed to test whether $700 billion of lost lending capacity is a realistic ceiling or a worst-case tail.
This article discusses macroeconomic research and is not financial, legal, or banking advice. Figures cited are drawn from published Federal Reserve research and are subject to revision.
For a plain-language look at how blockchain settlement differs from traditional banking rails, see Coin680’s Bitcoin Academy.