Written by Smartech Daily Team
This article has been originally published on Smartech Daily and republished at Dataconomy with permission.
By Sergey Kravtsov, Papaya Finance, CEO
Stablecoin payments spent 2025 collecting every win the fintech industry had spent five years asking for. Settlement volume crossed $33 trillion, more than Visa and Mastercard moved between them, and frameworks became law across the G7. Yet the share of global payment flows actually running on stablecoins stayed at roughly one percent — the same as in 2023 and 2024. In the same research, 86% of firms called their own infrastructure ready, while Visa executives scoring institutional adoption out of ten returned half a point.
Here is what that gap comes down to, and it is a fintech problem rather than a blockchain one. Chains solved settlement, the act of moving value. They did not build the layer above it, where two institutions agree in advance what a payment means, what it permits, and what happens when it fails. That layer is not a feature anyone ships alone — it is a set of standards, and it does not exist yet. Everything below is the same absence appearing somewhere else.
The blocker moved down the stack
Eighteen months ago the first question from a bank or a fintech on the other side of the table was whether any of this was legal. Today it has an answer wherever cross-border volume moves. The United States has a federal perimeter with agencies deep into implementation — the OCC’s proposed rules in February 2026, the FDIC’s in April, a joint Treasury proposal on anti-money-laundering and sanctions. MiCA has been fully applicable in Europe since late 2024, and Hong Kong licenses issuers.
The launches did not follow. Announcements did, and the asymmetry tells you where the work actually is: an announcement is a marketing decision priced in days, a launch an infrastructure decision priced in quarters. From outside it looks like an institution waiting on regulation; from inside it is the last mile between a stablecoin settling on-chain and a product someone can operate, audit, and explain to a customer. My own meetings shifted accordingly, from general counsel to integration engineers, treasury, and operational risk. Same institutions, same intent, different blocker.
A payment is an agreement, not a transfer
In card, ACH, and SEPA systems, moving the value is the smallest component. Around it sits an authorisation model, a mandate with a lifecycle, an exception-handling apparatus, and a reporting format that lets two institutions describe the same event identically. That structure is the agreement; the transfer just executes it.
The Travel Rule shows how little of that a chain can hold. FATF Recommendation 16 requires originator and beneficiary information to travel with the transfer. An ERC-20 transfer event carries a sender, a recipient, and an amount, and no field for anything else — so the data moves in a parallel standard, IVMS101, over out-of-band channels, matched back to on-chain transactions afterwards. Institutions run two systems reconciled against each other where a card scheme carries the same payload in one message.
This is also why swapping a wire for a stablecoin transfer is not modernisation: the same fixed instruction on a faster medium. A settlement instruction on-chain is a program and could carry the condition inside the payment — release against a signed attestation, accrue per second, draw against metered consumption. What stops that is not the rail. A salary accruing per second has no pay period, and withholding is defined on pay periods. The rail already expresses flows the surrounding machinery cannot record, which is the same absence one level up.
Where it bites hardest: recurring collection
A SEPA direct debit mandate encodes a persistent object with a unique reference and creditor identifier, a lifecycle covering amendment and revocation, a pre-notification obligation, a family of R-transactions for rejects and refunds, and an eight-week unconditional refund right. Cards encode a parallel apparatus, including decline codes precise enough to drive retry logic and account updater services for when the credential changes underneath the merchant.
On-chain, the equivalent was assembled from general-purpose primitives never designed for it, and each falls short differently. A standing token approval grants an unbounded claim with no ceiling, schedule, or expiry — custody wearing a different word. ERC-2612 permit replaces it with a signature carrying a deadline, which removes the payer’s gas burden but still authorises an allowance state rather than a schedule. EIP-3009 comes closest in production. Its signed-authorisation pattern, implemented in USDC, carries a validity window and a random nonce and can be submitted by a third party, which makes it genuinely a pull instruction. Even so, the amount is fixed at signing, so metered billing cannot use it, and revoking means racing the collector to spend the nonce.
The most recent step forward is EIP-7702, live since Pectra in May 2025, which lets an ordinary wallet delegate execution to code carrying a real spending policy. What it does not resolve is the residual key: the payer’s original key can re-delegate or reset the account at any moment, so a mandate stays silently revocable with no notification path to the collector. SEPA closed that hole with a rulebook obligation between institutions. A chain has nowhere to put an obligation.
What none of it provides is a mandate standard — no interoperable object describing ceilings, periodicity, and expiry that both sides read, no revocation carrying a proof a supervisor accepts as evidence, no decline codes, no retry semantics, no distinction between temporary and terminal failure. Every integrator builds this privately and incompatibly, so no institution can treat a mandate as portable and no supervisor can treat it as a control. Filed under subscriptions this looks niche, and it is not: payroll, insurance premiums, metered B2B billing, supplier drawdowns, and loan servicing rest on the same missing primitive, which is why fintechs building in any of those categories keep hitting the same wall.
And it is not a throughput problem
The tempting read is that this is a capacity problem, and it is worth killing that read directly. Executed naively, collection is N discrete transactions, each carrying 21,000 gas before doing anything, with storage writes costing 20,000 from zero against roughly 2,900 to update. Batching removes the overhead but not the state writes, and that floor comes down only by changing what gets written — holding balances in one contract’s own accounting and settling net positions outward, which is an accounting decision rather than a chain-selection one. Meanwhile PeerDAS, activated with Fusaka in December 2025, delivered roughly an order-of-magnitude increase in rollup data availability.
So the cost of one more payer is now a rounding error. Throughput solved the cheap half. No consensus upgrade delivers a mandate standard, a decline code, or a dispute process.
Three places the same gap surfaces
Ask an institution what stopped it and you get three answers that sound unrelated. They are the same answer.
The first is timing. A core banking system posts on cut-off windows and value dates; a chain does not stop at 17:00, so somebody has to decide what a continuous ledger looks like cut into periods, and most institutions have never written that policy down. Finality is not binary either: the confirmation a user sees on a rollup within a second is a promise from a single sequencer, and for optimistic designs the settlement that removes that exposure lands after a seven-day challenge window. Crediting a customer before it does is a credit decision and should be documented as one. Even the reporting format has no home for this — ISO 20022 offers no field for a chain identifier, a token contract, or a transaction hash. This is the least glamorous gap on the list and the one that kills the most launches.
The second is fungibility. Bank deposits settle at par because a central bank sits underneath and every institution reaches the same settlement asset. Two stablecoins from two issuers have no such mechanism; they clear through market liquidity and trade at a spread under stress. Redemption at par runs through a whitelisted account on the issuer’s banking calendar — exactly when a corridor cannot wait. So institutions pre-fund on every chain and issuer they support, which is a nostro account reinvented without a correspondent, and usually the largest item in the real cost stack, ahead of ramp fees, FX spread, and screening. Set that against the benchmark this technology is measured by: sending money across the G20 still costs an average of 6.39% as of Q3 2025, against a G20 target of 3%. Stablecoins win on gas and give it back at the edges — and because almost all edge cost is fixed per corridor and counterparty, it amortises over repetition and volume rather than speed, which is why the corridors that close first are the repeated ones.
The third is failure. Payment systems assume things go wrong at a predictable rate, which is why the apparatus for handling failure is most of the product — SEPA has its R-transactions, cards have chargebacks, evidentiary standards, and arbitration. On-chain, the only reversal that exists in practice is an issuer’s freeze function: a discretionary act with no published evidence standard, no appeal, and no obligation running to the counterparty. No institution can present that to a supervisor as a control.
Timing, fungibility, failure. In each case the primitive works and the agreement about what it means does not exist.
The case for waiting, and why it fails
Custody treatment, reserve composition, the yield prohibition, and the capital treatment of tokenised deposits remain genuinely open in most jurisdictions, and the current US federal texts are proposals rather than final rules. A bank waiting for final language is prudent, not slow. But the institutions that did receive clarity, with executive mandates behind them, did not ship faster. Regulation was necessary and never sufficient — and one percent of global payment flows is what that looks like from outside.
What gets built first
The first piece to close will be an interoperable mandate standard paired with batched collection on cross-border corridors, where the pain is sharpest and the regulatory path cleanest. Domestic billing collides with local-currency rules immediately; cross-border flows survive them, because there the stablecoin is only transport between two institutions with fiat at both ends.
My prediction: within twelve to eighteen months a regulated bank ships recurring cross-border collection — B2B billing and payroll — as a native feature in its own application, settling on stablecoin rails, with the customer never seeing the word crypto. Not in New York or Frankfurt first, but where correspondent relationships are thin and incumbents are light enough to move.
It will arrive as a standard rather than a product, because that is what has been missing the whole time. Which is why the round after it gets fought over decline-code taxonomies, idempotency, uptime commitments, and dispute policy. The fintech firms that win will look considerably more like acquirers than protocols.
