While the industry spent eight months cataloguing supply-side plumbing — Visa Direct moving into production, Mastercard's Global Money Program reaching 130 countries, roughly $400 billion in reported stablecoin payment volume — a working paper from the Bank for International Settlements argued the binding constraint sits somewhere no integration roadmap can reach. Not throughput. Not compliance. The interest rate on an ordinary bank deposit.
That is the finding worth your attention, and it is not the one the market priced.
The ledger remembers what the hype forgets. Every announcement this year described a pipe getting wider. None of them described a customer getting a reason to fill it. A rail that clears in nine seconds is worthless to a treasurer comparing it against a 4.8% money market fund — and that comparison, not the rail, is what the BIS researchers measured. Their conclusion, stripped of econometrics: banks can slow stablecoin adoption simply by paying more for deposits. The incumbent's defense is a number on a statement, not a product launch.
The regulatory frame matters more than the technical one, so start there.
The GENIUS Act — the Guiding and Establishing National Innovation for U.S. Stablecoins Act — was signed in July 2025 and builds a federal framework for payment stablecoins: 1:1 backing in high-quality liquid assets, permits at the federal and state level, and, in Section 4(a)(11), a flat prohibition on issuers paying yield to holders. That prohibition takes effect January 18, 2027. Reserve interest accrues to the issuer. Holders get zero at the issuer level.
Everything else in the stablecoin stack has been remarkably static. The architecture of a fiat-backed token has not changed materially since 2014: a custodian holds short Treasuries and overnight repo, an issuer mints a claim, a redemption window converts it back. What changed this cycle was distribution — card networks wiring token settlement into merchant acceptance, banks testing on-ramps, wallets absorbing the UX. Visa and Mastercard did not make stablecoins better money. They made them reachable money.
That distinction is where the analysis usually collapses. Supply-side infrastructure solves acceptance. It does not solve demand. And demand, per the paper, is priced against an alternative that pays interest and carries deposit insurance.
It also matters because stablecoins are no longer a niche instrument. They are the base money of DeFi — collateral, quote asset, and settlement unit across lending markets and automated market makers — which means their supply answers to two different demand curves at once: a transactional curve that is nearly rate-insensitive, and a store-of-value curve that is acutely rate-sensitive. The measured elasticity is a blend of the two. Pure trading demand is far more resilient than the headline coefficient implies, which is why DeFi liquidity held better than the doomsday models predicted through the last tightening cycle.
Three findings, in order of how much they should change your model.
First: the constraint is demand-side, and it is elastic to rates. The paper's central claim is that deposit rates act as "stubborn gravity" on stablecoin growth. That is a direct rebuke to the infrastructure-determinism thesis that has organized this sector's narrative for three years — build the rails and adoption follows. It does not follow. It follows only when the rail's convenience premium exceeds the yield the holder surrenders. In a 5% rate environment, that premium is large and the surrender is painful.
Second: the yield ban is a mandatory redistribution of reserve income, not a consumer-protection measure. Read Section 4(a)(11) as an economist rather than a compliance officer and the mechanism is stark. Reserves generate interest — roughly 100% of issuer revenue, correlated with the policy rate and essentially uncorrelated with crypto activity. Before the ban, issuers competed for float by rebating part of that interest. After the ban, the rebate is illegal at the issuer layer, and the spread stays with the issuer and its distribution partners. Stablecoin issuance becomes a spread business, not a service business. That is a legislated margin transfer from holders to intermediaries, dressed as prudence.
Follow the money through the stack and the allocation becomes concrete. Reserve income — the interest on short Treasuries, overnight repo, and bank deposits — is generated at the asset layer and claimed by three parties: the issuer, the distributor, and the holder. Pre-ban, the holder had a claim, formal or informal. Post-ban, the holder's claim is zero by statute, the distributor's claim is negotiable and unconstrained, and the issuer's claim is whatever survives the revenue-share agreement. Circle's arrangement with Coinbase is the template here and the detail most analysts skip: it determines the actual split of reserve economics and therefore the actual profitability of the issuance business.
Third: the numbers in the underlying research do not survive a calculator, and you should say so out loud. The reported coefficient — deposit rates up 1 basis point, ten-week stablecoin growth down about 3 percentage points, with an instrumental-variable estimate at roughly triple the OLS result — is economically impossible as written. Run it: 100 basis points of tightening would imply a 300-percentage-point collapse in growth. The 2022–2023 hiking cycle delivered roughly 525 basis points. That would imply a 1,575-percentage-point swing. Total stablecoin market cap over that window moved from roughly $180 billion to roughly $125 billion — a drawdown near 30%, not a descent into negative infinity.
The likely explanation is a unit error in transcription — basis points conflated with percentage points, or a coefficient that corresponds to 100bp of tightening rather than one. The direction of the finding matches every practitioner's intuition: when cash pays, float loses. The magnitude is unusable. Based on my audit experience cross-referencing whitepaper tokenomics against deployed contract logic, I have learned to treat a number that fails dimensional analysis as a finding in itself — it tells you the source was summarized rather than read. Anyone feeding "3 percentage points per basis point" into a TAM model is building on a typo.
There is a methodological problem too. Using "the federal funds rate crossing 3%" as an instrument for deposit repricing fails the exclusion restriction almost by construction. That same variable moves risk appetite, DeFi lending yields, and every crypto beta asset simultaneously. The instrument is not isolating deposit competition; it is capturing the entire tightening regime.
Now the structural squeeze, which the paper implies but never states. Issuers are pinned from both ends. In a high-rate regime, deposit competition caps growth. In a low-rate regime, growth resumes but unit revenue per token compresses, because reserve income falls with the policy rate. You cannot simultaneously maximize float growth and spread per unit. That is not a market cycle; it is the shape of the business.
And value capture is migrating. The ban binds issuers only. Exchanges, wallets, and third-party platforms remain free to run reward programs, fee rebates, and points schemes that function as yield by another name. So the entity holding the customer relationship — Coinbase, not Circle — keeps the retention tool, while the issuer degrades toward a low-margin settlement utility. Bridging the gap between code and community means naming this plainly: the protocol is not where the power sits. The distribution layer is.
Here is the angle almost nobody reported.
The yield ban is widely read as a defeat for stablecoins. It is better read as the price of a much larger victory. The Howey test turns on four elements, and the decisive one is expectation of profit from the efforts of others. By legally foreclosing yield at the issuer level, Section 4(a)(11) removes that element from the instrument's definition. A payment token that cannot pay interest is very hard to characterize as an investment contract. The ban is a bank-charter protection measure, yes — and simultaneously the cleanest non-security opinion the industry has ever been handed, for free, in statute.
The second blind spot is bigger. The paper never mentions tokenized deposits. Banks issuing yield-bearing deposit tokens backed by bank credit, potentially inside deposit insurance, running on the same 24/7 rails, is not a competitor to stablecoins — it is a downgrade of them. If JPMD-class products reach production at scale, the addressable market for non-yielding stablecoins narrows to exactly the population banks do not want: offshore users, unbanked households, and machine-to-machine settlement.
Which points at the third omission: the rate cycle. The sample window is a hiking plateau. Cut rates and the "stubborn gravity" weakens on its own. The paper's finding may partially reverse without a single line of code changing.
And treat that $400 billion payment figure with suspicion. No methodology is disclosed. If it includes exchange-to-exchange transfers and market-maker inventory shuffling — and these figures usually do — terminal commerce is a fraction of the headline.
Mark January 18, 2027. In the months before that date, the real action is not in token design; it is in the rulemaking, where the OCC and Treasury will decide how broadly they define "disguised yield." That definition determines whether the ban has teeth or whether every exchange quietly becomes an interest-bearing account.
Watch four signals: Circle's quarterly circulating supply regressed against the effective fed funds rate; the first mainstream bank to launch a yield-bearing deposit token; any enforcement action against a platform reward program; and a payment-volume series that strips internal transfers. Narratives move markets faster than blocks — but the ledger settles the argument eventually.
The question worth sitting with is not whether stablecoins can scale. It is whether a payment instrument legally forbidden from paying its holders can ever compete for the store-of-value demand that actually drives float — or whether the industry spent a decade building the world's fastest settlement layer for a customer who was never coming.