More than thirty banks, with combined assets above $10 trillion, have joined the Cari Network, a tokenized-deposit network expected to be fully operational by the fourth quarter. A survey American Banker published this month found nearly two-thirds of banks offering or developing tokenized deposits for corporate clients. A tokenized deposit is a claim on money at a licensed bank, recorded on a blockchain instead of the bank's own ledger. It is supposed to be a deposit in every way that matters, with the plumbing upgraded underneath. That is the whole pitch, and it is also the whole risk.
The new product has borrowed the trust of the old one: the insurance, the supervision, the settlement finality. All of it was built for a ledger that closes at night, and none of it has been tested on rails that never close. The question beneath this year's pilots is whether the new plumbing can deliver what the old plumbing promised, or whether the borrowed advantage gets withdrawn.
The token inherits the whole trust kit
Start with what a tokenized deposit legally is. It is a deposit: a liability of a licensed bank, on the bank's balance sheet. It creates no new money and stays on the books; what changes is the record of ownership, which moves to a distributed ledger. Everything else stays.
Because it remains a deposit, it keeps the whole trust kit. Balances are covered by the FDIC up to $250,000 per depositor per bank. The bank keeps its Federal Reserve discount window access. The deposit can pay interest. Stablecoins can claim none of this. The GENIUS Act, signed in July 2025, excludes tokenized deposits from the definition of a payment stablecoin, bars issuers from marketing their products as insured, and forbids stablecoins from paying interest.
Regulators have underlined the distinction. In a statement on the GENIUS Act proposal, FDIC Chairman Travis Hill reaffirmed that a deposit remains a deposit regardless of the technology or recordkeeping used, while proposing to block stablecoins from pass-through deposit insurance. His position compresses into five words: "a deposit is a deposit."
None of this is accidental: the insured tokenized deposit is positioned as the regulated answer to the uninsured stablecoin. A Treasury advisory committee estimated last year that $6.6 trillion in transactional deposits could migrate to stablecoins, though critics argue the figure counts the whole stock rather than a realistic outflow. The insurance advantage is the product, loaned to it by regulators and the deposit's own history.
The guarantee was designed for a ledger that closes at night
Here is the difficulty: the guarantees being borrowed were built around the old ledger's rhythm. Deposit insurance pays out against a record: the bank's books, settled at known moments. Liquidity supervision assumes a bank can see its position at a cut-off and forecast redemptions at day's end. Even the word "run" belonged to that world, a thing that happened during banking hours.
A tokenized deposit changes the plumbing beneath all of it. Money can move on a Saturday night. There is no close of business. The record of who owns what now lives across a bank's core system, a virtual ledger, and a blockchain network, but bank controls were built for a single system of record. M&T's head of enterprise innovation says the liquidity risks feel familiar but heightened when customers can move money at any hour. First Horizon's data and strategy director says the bank weighed whether tokenization changes the pace at which liquidity can move, and concluded runs are driven by the customer base, not the technology. SouthState's enterprise payments leader warns against automating flawed legacy processes onto the new rails, which only produces a faster bad process.
None of these executives is a skeptic; all are describing controls built for a rhythm that no longer exists.
The record has never met a stress test
Deposit insurance is a promise about the worst moment, and its machinery is unproven on the new rails. Hill made the same point about stablecoins: settle the pass-through question by regulation, not for the first time when a bank holding stablecoin reserves fails. The logic applies one level down. No insured bank with tokenized deposits has ever failed, so no one has processed an insurance claim against a distributed ledger, established which record is authoritative, or reconciled the on-chain and off-chain ledgers at the moment it mattered.
The state supervisors who charter four in five of the nation's banks know the machinery is untested. In a November 2025 letter to the FDIC, the Federal Reserve, and the OCC, the Conference of State Bank Supervisors asked the agencies to confirm that a deposit recorded on a distributed ledger is covered by federal deposit insurance to the same extent as a traditional deposit, and to set expectations for recordkeeping, reconciliation, and claim processing on those ledgers. The word "confirm" is doing real work: you confirm what you believe but cannot yet demonstrate.
The FDIC's answer has been technology-neutral reassurance: a proposed rule reaffirming that tokenized deposits are insured like any other, which the major bank trade groups, including The Clearing House, publicly endorsed. That may be the right answer, and it is unusual: ordinary deposits do not have their coverage re-issued in public. The rulemaking itself is a reminder that the guarantee is being renegotiated as it migrates, and renegotiation can narrow as well as broaden. The borrowed advantage is only as firm as the guidance that extends it.
Always-on money moves a run the same way it moves a payment
The second borrowed guarantee is the settlement rhythm itself. The spring of 2023 showed what withdrawals look like at phone speed: three regional banks failed within weeks, uninsured deposits fled faster than the old models expected. Tokenized rails remove the speed bumps the old system kept: no cut-off, no weekend, no branch queue, nothing between a treasurer's decision and the transfer.
Banks split on how much this matters. First Horizon's argument, stated plainly: the technology does not cause runs, the customer base does, and a bank funded by sticky deposits is safe at any speed. The state supervisors take the opposite precaution, asking for liquidity expectations that treat always-on redemption as a risk, including stress tests and contingency funding plans built for round-the-clock withdrawal.
This analysis takes no position on which view is right. The common ground: both sides agree the liquidity machinery must change, and neither can verify the change will hold. Whatever the true run risk, the lending and insurance promises depend on a bank's ability to see and fund its position continuously, and no bank has yet had to prove it could.
Even settlement stops at the bank's own wall
The quietest gap is interbank settlement. A tokenized deposit moves freely inside its issuing bank, but moving between banks is different. The largest live programs remain intra-bank: JPMorgan's Kinexys has processed trillions in volume, almost all inside one institution, while roughly $2 trillion a day still clears over CHIPS and more than $4 trillion over Fedwire. Analysts note that true interbank tokenized settlement does not exist at production scale. Every interbank transfer still ends on the old ledger.
The industry knows it. The same banks are building a shared tokenized deposit network through The Clearing House, targeting the first half of next year; Wells Fargo announced this month it will run its own platform this fall alongside that work. Two systems for one promise is a hedge, and a hedge is an admission. The money-between-banks claim is borrowed from a network that does not exist yet; the money-inside-my-bank claim is real today.
Interoperability was the old system's great achievement, the reason a payment from anywhere settles anywhere. The new networks begin as islands. SouthState's Tara Edmonds put it plainly: "The opportunity grows when we can interact across different banks." Until they do, the tokenized deposit's defining advantage, being a real deposit, is also its boundary: a deposit at one bank, and no further.
Loans get called, and this one is still on loan
The borrowed advantage can be withdrawn three ways. By performance: the first real failure on the new rails, when claims processing, reconciliation, and liquidity forecasting are tested at once and the answer is unknown. By rule: guidance that conditions the promise, on recordkeeping, on marketing, on which networks qualify. By the market: depositors and creditors who remember the advantage was provisional and move back to the tested system.
None of this argues the technology is a deception, and none of it argues for weaker supervision. Every request in this story, from state supervisors asking for joint guidance to the FDIC writing rules, runs toward more clarity, and that is the right direction while the machinery is unproven. For a retail depositor, coverage to $250,000 does not depend on the ledger; what is open is whether the record and the claims machinery hold when coverage is needed. The industry's own executives expect years of careful work. M&T's Matt McAfee says "anybody who thinks that they're missing the boat" underestimates how early it all still is.
The product is being sold as a deposit with an upgrade, and the honest description leans the other way: a deposit whose guarantee was built for another machine and loaned to this one on the strength of a name. The loan converts into ownership only at the first stress test, the first run, the first claim processed against a distributed ledger, and no one knows what that will look like. Until then, the borrowing terms are still being written, by examiners, by rulemakers, and by whatever the next failure teaches.
Primary sources
- American Banker's "Tokenized deposits are here. Banks need to manage the risks," Jasmine Ni, August 2026, for the Cari Network details, the survey finding, and the interviews with Matt McAfee of M&T, Brian Mellone of First Horizon, and Tara Edmonds of SouthState.
- The Conference of State Bank Supervisors' November 2025 letter to the FDIC, the Federal Reserve, and the OCC for the request to confirm insurance coverage for deposits recorded on distributed ledgers and the recordkeeping, reconciliation, claims, and liquidity asks.
- The FDIC statement by Chairman Travis Hill on the GENIUS Act proposal for the "a deposit is a deposit" position and the stablecoin pass-through insurance stance, and the GENIUS Act itself for the exclusion of tokenized deposits from the stablecoin definition and the ban on marketing stablecoins as insured.
- The Clearing House's June 2026 advocacy piece for the trade associations' endorsement of the technology-neutral approach, and reporting on Wells Fargo's dual-track launch for the interbank settlement gap figures and the Treasury Borrowing Advisory Committee's $6.6 trillion estimate.