Over the past ninety days, the aggregate supply of tokenized Treasury products expanded at a double-digit clip. On-chain transfer volume for those same instruments contracted. Two lines diverging is not noise. It is a disclosure.
I track that ratio the way equity desks track days-to-cover. Call it utilization: the share of tokenized money-market supply that changes hands in a given week. It has slid since the second half of 2025. Supply grows because issuance is administratively cheap. Volume falls because the assets are parked, not spent. The tokenization trade is a collateral trade wearing a payments costume, and the market has not repriced the gap.
The products themselves are familiar. Tokenized Treasury funds wrap short-duration government paper in a token, distribute it to allowlisted wallets, and settle redemptions in fiat through a transfer agent. The token is a mirror of a register. It is not the register.
That distinction determines everything downstream. A tokenized Treasury is a claim on a fund share. The cash leg — the actual dollars — still moves through correspondent banking on T+1. When a holder wants out, they do not get dollars from a smart contract. They get a wire. The on-chain layer handles ownership attestation and intraday transfer between parties who already trust each other. It does not handle settlement finality between institutions that do not.
Three cohorts hold these instruments. Stablecoin issuers parking reserve float. Prop firms and market makers posting margin. A thin slice of DAO and corporate treasuries diversifying out of idle USDC. None of them are trying to pay anyone. They are trying to hold something yield-bearing that can be pledged.
That is a collateral function, not a payments function, and it explains the utilization gap precisely: collateral is designed to sit still. The metric that matters is not assets under management. It is the velocity of the float.
Three years of RWA conference panels have promised a chain where a corporate treasurer moves between a money-market token and a payment token in one atomic step. The demos work. The production deployments do not, because the moment the counterparty is a regulated entity that does not share your transfer agent, atomicity breaks and the transaction reverts to a wire. What has scaled instead is the boring part: issuance. Issuance is a distribution problem, and distribution problems get solved. Settlement is a coordination problem, and coordination problems get scheduled.
Examine the incentives. An issuer earns a management fee on outstanding supply. Supply is maximized by distribution deals, not transaction volume. The bank partner earns on the fiat rails. The custodian earns on safekeeping. Every participant is compensated on stock, not flow. So the industry optimizes for the number that looks like adoption and ignores the number that would prove it. That is not a conspiracy. It is compensation design — the same failure mode I modeled in 2020, when I built Python simulations of early liquidity mining. Emissions rose, pools deepened, and swap demand never validated the emission schedule. Supply without velocity is a subsidy.
Now the layer where I have direct evidence.
In 2025 I ran a cross-border pilot for a Southeast Asian import-export client. USDC on Polygon, three regional banking partners, a five-developer team, two lawyers. The headline was a 60% reduction in transaction fees against the SWIFT baseline. That number is real. It is also the least interesting thing the pilot produced. What it revealed was prefunding drag.
To guarantee T+0 settlement, we had to pre-position USDC in every corridor the client used. Six corridors meant six funded accounts, each with an operational minimum. Aggregate float sat near four million dollars. At a 5.5% internal hurdle rate, that is roughly two hundred thousand dollars a year in carrying cost to generate about four hundred thousand in fee and FX savings. Net positive. Thin. And contingent on payment frequency.
The breakeven is a frequency calculation, not a technology calculation. A corridor with daily flow amortizes its prefunded balance across thousands of payments. A corridor with weekly flow amortizes it across dozens. The marginal seventh corridor — low volume, high compliance overhead — carried more float cost than fee savings. We shut it down and routed it back to the legacy rail, and the client's net benefit went up. Liquidity fragmentation is the bottleneck. Not throughput, not gas, not finality. Float.
Here the two threads converge. Tokenized Treasuries and stablecoin rails are attacking the same problem from opposite ends. One puts idle capital on-chain and lets it sit. The other needs capital on-chain and needs it to move. Neither has produced the piece that would join them: an on-chain payment-versus-payment mechanism letting two institutions exchange a tokenized claim for a stablecoin without trusting each other's books.
The plumbing exists in fragments. Custody APIs, allowlisted stablecoins, tokenized collateral schedules. What is missing is the legal wrapper that makes a smart contract's finality enforceable in a bankruptcy court. That is a documentation problem, and documentation problems move at the speed of law firms, not code.
Until that exists, the collateral layer and the settlement layer stay separate. Trust is verified, never assumed — but verification infrastructure is expensive, and in a sideways market nobody funds expensive infrastructure without a narrative attached.
Which brings me to what is mispriced. The tokens claiming exposure to tokenization capture almost none of the economics. Management fees accrue to the issuer. Custody fees accrue to the custodian. Float income accrues to whoever holds the reserve. On-chain, the take is a few basis points at most, frequently zero. The value creation is real. The value capture is off-chain and permissioned.
Meanwhile this sector prices on narrative. In a trending market that gap is forgivable, because liquidity papers over bad unit economics. In a sideways market it is not. Sentiment compresses first, and projects with no fee capture compress hardest. Strategy prevails where sentiment fails.
Regulation is the new liquidity engine — but it routes that liquidity into structures designed to be audited, not to be fast. A second blind spot follows from the first: the assumption that banks will eventually transact on public chains misreads what they are already doing. They use public chains as an attestation layer — a notary with a block explorer. Custody stays inside the permissioned perimeter. Cash stays on the legacy rail. What gets tokenized is the record, not the asset. That is a legitimate business. It is simply smaller than the phrase "on-chain finance" implies. The macro view reveals what the micro hides.
Watch two things. First, utilization — if the transfer-volume-to-supply ratio turns up while supply is flat, the collateral layer is starting to function as money. Second, the first production deployment of on-chain DvP between two regulated institutions that do not share a custodian. That is the unlock. Everything else is distribution.
Convergence is inevitable; timing is tactical. The float is already on-chain. It just has nowhere to go yet.