A Dallas Fed warning puts the funding economics of tokenized deposits in sharper focus

A fresh analysis from economists at the Federal Reserve Bank of Dallas has added needed realism to the tokenized-deposit debate. The paper's core point is straightforward: if bank deposits become easier to move instantly, program automatically and compare continuously across institutions, they may become less stable as funding. That would not mean tokenized deposits are a bad idea, nor would it imply banks stop lending. But it would mean the economics of always-on bank money are more complicated than the industry's current product rhetoric often suggests. For a market increasingly focused on onchain cash, that is an important intervention.

The Dallas Fed analysis by Rosie Levy and Srini Ramaswamy argues that tokenized deposits could increase the sensitivity of deposits to rates and shorten how long funds remain with any single bank. In the bank funding model used in the note, those changes matter because deposits do more than provide payment utility; they also support maturity transformation, allowing banks to fund longer-dated loans with comparatively sticky liabilities. The authors present scenarios rather than forecasts, but the magnitudes are still notable. A 10% reduction in deposit weighted-average life would lower aggregate maturity-transformation capacity by roughly $580 billion in 10-year equivalents, while a 10% increase in deposit rate sensitivity would reduce banks' duration-risk appetite by about $700 billion on the same basis.

Those estimates do not translate one-for-one into a collapse in lending, and the Dallas Fed note is careful on that point. Banks could preserve lending by changing other parts of the balance sheet, especially through greater reliance on term debt or wholesale funding. The tradeoff is cost. If more lending has to be supported by more expensive liabilities, some of that cost would likely show up in credit pricing for households and businesses. The argument therefore is not that tokenized deposits break banking, but that the convenience gained by depositors may reprice the funding model underneath traditional intermediation. That is a more useful frame than the binary narrative of innovation versus resistance.

What makes the note especially relevant is that it is not purely theoretical. The authors connect the argument to the broader development of faster payment infrastructure, including the Federal Reserve's FedNow service, which the Federal Reserve describes as enabling eligible depository institutions to deliver instant payment services to customers. The concern is that once settlement becomes immediate and software-driven, frictions that historically kept deposits relatively sticky begin to weaken. The Dallas Fed note goes a step further by highlighting how programmable deposit tokens and agentic AI could automate bank switching for yield-seeking users, compressing the time banks have to react to funding outflows.

The empirical backdrop comes from outside the US but is highly relevant. The Dallas Fed authors cite research built on Brazil's Pix system, and the underlying Banco Central do Brasil working paper provides concrete evidence from a large-scale instant-payment network. That study found that greater Pix usage was associated with higher demand for liquid assets, especially government bonds, and lower liquidity transformation on bank balance sheets. In the paper's estimates, a one-standard-deviation increase in Pix usage led to a 12.7 percentage point rise in the ratio of demandable deposits and a 15.4 percentage point increase in liquid assets, with the asset shift driven primarily by government bond holdings. In other words, once money moves faster, banks appear to protect themselves by holding more immediately usable collateral and fewer longer-duration assets.

That backdrop matters because tokenized deposits are often marketed as a clean upgrade to payments and settlement, especially for treasury operations, wholesale transfers and machine-driven commerce. Those benefits may prove real. Yet if the liability side of the bank changes at the same time, the asset side cannot remain untouched. More liquidity buffers, more collateral readiness and more expensive term funding would all be plausible responses. For onchain finance builders, that suggests the winning tokenized-deposit models may be the ones designed with balance-sheet economics in mind from day one, rather than assuming banks can simply wrap existing deposits in new rails without changing anything important.

The practical implication for the RWA market is that tokenized cash products are entering a more policy-intensive phase. Stablecoins, tokenized deposits and tokenized money funds are no longer just competing on speed or programmability; they are competing on who absorbs the cost of liquidity, compliance and capital efficiency. The Dallas Fed note qualifies as a strong signal because it comes from inside the US central banking system and grounds the discussion in funding mechanics rather than ideology. As institutions push further into always-on settlement, the next chapter will not be about whether tokenized deposits are possible. It will be about whether they can scale without materially changing the price and structure of bank credit.