B2B Crypto Payments: How Settlement Between Two Companies Actually Works

A joint study by the BIS Committee on Payments and Market Infrastructures and SWIFT gpi, published in February 2022 on data collected in late 2020, measured what happens to a cross-border payment after it leaves the sending bank. Intermediary banks processed 78% of payments in under five minutes. Beneficiary banks managed 33%. End to end, the share of payments completing in under five minutes fell to 25%. The average processing time was 8 hours 36 minutes; the median was 1 hour 38 minutes.

That gap between mean and median is the whole story of company-to-company payments — averages hide it, and the tail is what breaks deadlines. Half the time it is quick. The other half drags a tail long enough to blow a closing date, and the sender has no visibility into which half a given payment is in.

Settling the same invoice in a stablecoin removes the tail and the correspondent chain. It also removes a set of protections that most finance teams have never had to think about because they were always there. The trade is real in both directions, and it is worth being precise about which parts of it are legal, which are operational, and which are simply different.

Finality is the whole point, and it cuts both ways

In a conventional funds transfer, “sent” and “final” are different events separated by a process. That process contains something valuable: an unwind path.

Under UCC 4A-402(c), the sender’s obligation to pay is excused where the funds transfer is not completed, and 4A-402(d) gives a sender who has already paid the right to a refund with interest. Together they are the money-back guarantee. Under UCC 4A-207, where a payment order identifies the beneficiary by both name and account number and the two do not match, the article sets out who bears the loss. Those provisions exist because a payment order travels through intermediaries and can fail in the middle.

An on-chain transfer has no equivalent of either. Once a transfer is included in a block and confirmed, the tokens are controlled by whoever holds the key to the destination address. There is no beneficiary name to mismatch against the account number, because there is no name. There is no unwind.

Final. That word does more work in a crypto payment than in most rails a finance team uses — which is why every control in this article sits before the transfer rather than after it.

The upside of the same property is that settlement risk between the two companies disappears at confirmation. There is no period during which the payment is en route and could be recalled, frozen at an intermediary, or returned for a missing field. For a supplier releasing goods against payment, that is worth something concrete.

What the recipient legally receives

The 2022 amendments to the Uniform Commercial Code introduced Article 12, dealing with controllable electronic records. The structure matters for anyone building a payment process on tokens.

§12-102 defines a controllable electronic record. §12-105 defines control, which is the concept that does the work Article 9 does for possession of tangible property. §12-104 sets out the take-free rule: a qualifying purchaser who obtains control for value, in good faith and without notice of a competing claim, takes free of that claim. A conforming §9-107A defines when a secured party has control of such a record, by reference to §12-105; perfection by that control runs through §9-314.

Adoption has been running state by state and is uneven, so the sensible way to talk about it is by structure rather than by count. Most states have enacted a version, several have not, and where your counterparty and your collateral sit determines which law applies. That is a question for counsel with your actual facts rather than something to settle from a table.

The practical translation for a payment: a company that receives tokens for value, in good faith, without notice of a competing claim, and takes control of them is in a strong position against a later claimant, in a state that has enacted Article 12. It is a weaker position in a state that has not. Either way, “without notice of a competing claim” is a phrase that rewards screening the counterparty and the address before accepting payment rather than after.

The protections you are giving up, listed plainly

Read that as a list of jobs that moved rather than a list of losses — every one of them still gets done, by you. Each missing protection corresponds to something a bank was doing on your behalf: verify the destination, screen it, confirm the network, and keep the record. The failure surface of a batch run and the specific controls are set out in crypto payouts.

What actually improves

Timing certainty. A confirmed transfer is confirmed. The mean-versus-median problem in the BIS and SWIFT data does not have an analogue, because there is no chain of intermediaries each of which can hold a payment.

Third-party verifiability. A transaction hash lets a counterparty, an auditor or a court confirm that a stated amount moved from a stated address to a stated address in a stated block, without asking either party for cooperation. Few conventional payment records have that property.

Cost independent of amount. A network fee is charged for the transaction, not for the value moved. On 27 August 2026, with Ethereum base gas at 0.039 gwei and ETH at $2,044, a USDT ERC-20 transfer cost about $0.03, whether it carried $500 or $500,000. On the same reading a USDT TRC-20 transfer cost $2.15 to $4.50. Gas is volatile: the same measurement on 26 August 2026 read 0.77 gwei, twenty times higher, so any fee figure needs its date attached. The network comparison sits in ERC-20 vs TRC-20.

Weekend and holiday settlement. Networks do not observe banking calendars, which matters more for month-end and quarter-end than it does on an average Tuesday.

For a benchmark that comes from a standard setter rather than a vendor, the Financial Stability Board’s 2025 progress report on the G20 cross-border payments roadmap found the share of wholesale payments credited within one business day “hovering above 90%”, with more than 50% within one hour, and concluded that efforts “have not yet translated into tangible improvements for end-users at the global level”.

The invoice is where most of the compliance risk lives

Settling in tokens does not change what an invoice has to contain, and the details trip people up in a consistent way.

Article 226 of Directive 2006/112/EC lists the particulars a VAT invoice must carry. Article 230 is the separate provision that matters here: invoice amounts may be expressed in any currency, but the amount of VAT payable must be expressed in the national currency of the member state. Conversion is governed by Article 91. Those are three different provisions and citing 226 for the currency rule is a common error worth avoiding.

The practical consequence: an invoice denominated in USDT with the VAT amount stated in USDT is defective in an EU member state. Denominate in the currency, state the VAT in the national currency, and treat the token as the settlement method rather than as the unit of account.

Article 226(11a) requires the mention “Reverse charge” where the customer is liable for the tax. The Court of Justice addressed the consequences of getting the triangulation mention wrong in Case C-247/21, Luxury Trust Automobil, and the answer was not forgiving: the mention is a substantive condition, not a formality that can be corrected retroactively at will.

Two older cases are worth knowing because they shape how a token payment is characterised. Case C-264/14, Hedqvist held that the exchange of traditional currency for bitcoin units, and back, is a supply of services exempt under the financial transactions exemption. Case 230/87, Naturally Yours Cosmetics established that where consideration is non-monetary, the taxable amount is the subjective value the parties actually attributed to it rather than an objective market figure. Note the citation form: cases decided before 1989 carry no “C-” prefix.

MiCA is the other live constraint for an EU-facing payer. The transitional period for crypto-asset service providers ended no later than 1 July 2026, and earlier in member states that elected a shorter window. A dollar-referenced token of this kind falls within MiCA’s e-money token definition, and Article 48 permits only an authorised credit institution or electronic money institution to offer one to the public in the Union. Tether holds no such authorisation, which is the operative point rather than the label. For a company inside the EU choosing a settlement token, that points toward USDC or a euro-denominated equivalent rather than USDT. The broader position is in MiCA compliance for crypto payments, and the mechanics of a compliant document are in crypto invoicing.

Counterparty and address diligence, before the first payment

In a bank transfer, the beneficiary bank has done onboarding on your counterparty and will decline a payment to a blocked party. On a chain rail, that work is yours.

Three checks, in this order.

Screen the destination address. Sanctions and risk exposure, before the transfer rather than after. The consequences of getting this wrong are not commercial. Sanctions liability in the United States is strict, and the procedural obligations are specific: under 31 CFR 501.603, blocked property is reported within 10 business days, with an annual report due by 30 September covering property blocked as at 30 June. Blocking and rejecting are different actions. The mechanics are in cryptocurrency address screening.

Confirm the address out of band. Business email compromise adapted to crypto is address substitution in an invoice, and it works because a 42-character string does not get re-read. Confirm by voice, or against an address already on file from a prior settled payment. A change of banking or address details on an invoice is the highest-risk event in the whole accounts payable process and deserves a written procedure.

Confirm the network. Store the network with the address as a pair on the counterparty record, never as a batch-level assumption. A payment sent to a custodial deposit address on a network the platform does not support is often unrecoverable, because the recipient holds no key. Where the counterparty self-custodies the position is better, and the recovery outcomes are set out in crypto payouts.

Where the GENIUS Act has actually got to

Any 2026 article that states the GENIUS Act’s rules as current law is ahead of the facts, so here is the position with its horizon attached.

The Act is Public Law 119-27, enacted 18 July 2025. Its effective date is set by a mechanism rather than a fixed date: effective “on the earlier of the date that is 18 months after July 18, 2025, or the date that is 120 days after the date on which the primary Federal payment stablecoin regulators issue any final regulations”. Eighteen months lands on 18 January 2027.

As at 27 August 2026, no final implementing regulations had been issued. The OCC issued proposed rules in March 2026, Treasury and the banking agencies proposed anti-money-laundering rules in April 2026, and Treasury published its main proposed rule on payment stablecoin issuance, offer and sale on 18 August 2026. Proposals do not start the 120-day clock; final rules do. So the working assumption for planning purposes is 18 January 2027, subject to being pulled earlier if final rules land before roughly late September 2026.

Two dated provisions are worth separating, because they get merged in commentary.

§3(b)(1), codified at 12 USC 5902(b)(1), makes it unlawful from 18 July 2028 for a digital asset service provider to offer or sell a payment stablecoin in the United States unless it was issued by a permitted payment stablecoin issuer. That is a three-year date, not the effective date.

§3(b)(2), at 12 USC 5902(b)(2), addresses foreign issuers: their stablecoins may not be offered in the US by a digital asset service provider unless the issuer “has the technological capability to comply, and will comply, with the terms of any lawful order and any reciprocal arrangement pursuant to section 18”. The reciprocal arrangement is part of what the issuer must be able to comply with, rather than a separate standing condition. This one applies from the Act’s effective date rather than from 2028.

Why your treasury cannot earn yield on a payment stablecoin from the issuer

§4(a)(11) of the GENIUS Act, codified at 12 USC 5903(a)(11), provides that no permitted payment stablecoin issuer or foreign payment stablecoin issuer “shall pay the holder of any payment stablecoin any form of interest or yield (whether in cash, tokens, or other consideration) solely in connection with the holding, use, or retention of such payment stablecoin”.

Two words carry the weight — and both narrow the prohibition. “Solely” limits the prohibition to yield paid for holding, rather than to every payment an issuer might make. And the prohibition binds issuers, not exchanges, wallets or platforms, which is why yield on stablecoin balances, where it is offered at all, is offered by intermediaries rather than by issuers.

For a treasury function, the planning consequence is that a stablecoin balance is a settlement instrument rather than a yield instrument, and the decision about how much to hold in it is a working-capital decision rather than an investment one. The wider treasury framing is in crypto treasury management.

The accounting, and the one thing people get backwards

Two separate obligations, often conflated.

Measurement. ASU 2023-08 added ASC 350-60, requiring fair value measurement for crypto assets that meet six scope criteria. A fiat-backed stablecoin that gives its holder an enforceable claim on the issuer is widely read as failing one of those criteria and therefore falling outside the scope of 350-60. That is the prevailing reading rather than a settled answer for every instrument, and it is worth having your auditor confirm the analysis for the specific token you hold rather than adopting a conclusion from a summary.

Basis. Treas. Reg. §1.1012-1(j) requires basis to be tracked per wallet or account from 1 January 2025. Rev. Proc. 2024-28 is an optional transitional safe harbour for allocating basis unused before that date. A great deal of published commentary describes the Rev. Proc. as the requirement, and that is backwards: declining to use a safe harbour does not relieve you of a regulation.

Transaction costs are addressed in Treas. Reg. §1.1001-7. And moving assets between wallets you control is not a disposition, except as to any digital assets used or withheld to pay for the transfer itself. That sounds obvious until an intercompany transfer between two of your own addresses shows up in the expense account. The matching process that catches this is in crypto reconciliation, and the record-keeping framework in crypto bookkeeping.

Payment terms, netting, and what to write into the contract

Faster settlement changes what payment terms can reasonably say, and most B2B crypto arrangements inherit terms drafted for wires without revisiting them.

Six clauses earn their place.

Unit of account and settlement asset, stated separately. The invoice is denominated in a currency. The token is how it is settled. Say which token, on which network, and what happens if the token depegs materially between invoice and payment.

The rate mechanism. Which price source, which moment, and who bears the difference. The candidates are the invoice date, the payment initiation time and the block timestamp of confirmation. The block timestamp has the advantage of being independently verifiable.

Who bears the network fee. Gross or net. Left unsaid, the supplier receives slightly less than the invoice and the payable never closes cleanly.

Address change procedure. A named process with out-of-band confirmation, and a stated consequence if it is not followed. This is the clause that pays for the whole exercise.

Evidence of payment. The transaction hash is the artefact. Say so, and say that provision of the hash discharges the obligation, so that a delivery failure on the recipient’s side does not become an open payable on yours.

Screening and sanctions. A representation from each side that it is not a blocked person, and an acknowledgement that a payment may be held where screening flags a destination.

Model language for the contractor equivalent of these is in the crypto contractor agreement; the same structure adapts to a supplier arrangement.

Netting deserves a note. Where two companies invoice each other, faster settlement makes gross settlement cheap enough that netting stops being worth the reconciliation complexity it creates. At an Ethereum network fee measured in cents, the administrative cost of running a netting agreement can easily exceed what it saves.

Building the operational side

The controls that matter are unglamorous and they are the same ones that make month-end close short.

Separate the person who edits a counterparty record from the person who releases a payment. This one separation defeats both the inserted payee and the edited address, which are the two attacks that adapt most readily from bank payments to on-chain ones.

Keep counterparties in an address book rather than pasting addresses. A stored address that has settled a prior payment is a far better artefact than a string copied from an email.

Set per-transaction and per-day limits by role. A limit that forces a second approver above a threshold costs nothing on ordinary payments and catches the unusual one.

Retain the file for anything sent in bulk. A bulk run with no retained file leaves a lump sum that has to be explained from memory.

Record the transaction hash on the payable at the moment of sending. Not reconstructed later from a block explorer.

That set is worth more than any single piece of software. Where a platform helps is by making it the default rather than a discipline: VaultNow runs payouts of up to 100 transactions from a CSV or address book, invoicing, and address screening from a single dashboard, with team permissions that let preparation and release sit with different people, at $0.50 per transaction plus gas.

Where stablecoin settlement does not fit

An honest page names the cases where the answer is a wire.

Where the counterparty cannot hold the asset. A supplier whose bank will close its account over crypto activity, or whose jurisdiction restricts receipt, is not a candidate regardless of what the rail can do. Turkey is the clean example: the Central Bank’s “Regulation on the Disuse of Crypto Assets in Payments”, published in the Official Gazette of 16 April 2021 at No. 31456 and in force from 30 April 2021, provides that “crypto assets shall not be used directly or indirectly in payments” and that payment service providers shall not build business models that do so. Receiving and converting is a different question from paying, but a supplier in that market will read the rule before you do.

Where a currency other than the dollar is the natural unit. Dollar-referenced stablecoins dominate liquidity. Settling a euro-denominated or sterling-denominated obligation in a dollar token pushes an FX exposure onto one side of the deal, and somebody has to own it explicitly rather than by accident.

Where the payment must produce a specific banking artefact. Several export and exchange-control regimes require evidence that funds arrived through an authorised banking channel. A wallet does not produce that document. India is the widely encountered case, where service export proceeds are expected to be realised through the banking channel and evidenced by the certificate that channel produces. A wallet transfer produces no such certificate. Where a counterparty needs that paper, the rail is chosen for them.

Where the amount is large enough that operational risk dominates fee savings. At a network fee of a few cents, the saving on a single large payment is a rounding error against the cost of one irreversible mistake. The rational threshold is not about the fee; it is about how confident you are in the destination address.

Where the accounting or audit position is not settled internally. Adopting a new settlement rail two weeks before a year-end audit, without an agreed measurement and basis-tracking approach, converts a payments decision into an audit finding.

None of these are arguments against the rail. They are the cases where the honest answer to “should we pay this supplier in stablecoin” is “not this one, and here is why”.

Frequently asked questions

What are B2B crypto payments?

Payments between two companies settled in a digital asset, usually a fiat-referenced stablecoin, rather than through the banking system. The invoice, the tax treatment and the contract work as they would on any other rail; what changes is the settlement mechanism and the finality of the transfer.

Is stablecoin settlement final?

Once a transfer confirms, the tokens are controlled by whoever holds the key to the destination address, and there is no mechanism to unwind it. Conventional funds transfers have one: UCC 4A-402(c) and (d) entitles a sender to a refund with interest where a transfer is not completed. On-chain settlement offers no equivalent.

Can a B2B crypto payment be reversed or charged back?

No. This is not a card rail, so chargebacks do not apply, and there is no bank-level return of a confirmed transfer. Address validation and screening before sending are the only real controls.

Can I invoice in USDT?

In an EU member state, denominating the invoice in a token is a poor idea. Article 226 of Directive 2006/112/EC lists the required particulars and Article 230 requires the VAT amount to be expressed in the member state’s national currency, with conversion under Article 91. Denominate in the currency and treat the token as the settlement method.

Is the GENIUS Act in force?

Not as at 27 August 2026. The Act takes effect on the earlier of 18 January 2027 or 120 days after the primary federal payment stablecoin regulators issue final regulations, and as at that date the relevant rules were still at the proposal stage. A separate prohibition on offering non-permitted payment stablecoins applies from 18 July 2028.

Can a company earn yield on stablecoin held for payments?

Not from the issuer. Section 4(a)(11) of the GENIUS Act, at 12 USC 5903(a)(11), prohibits a permitted or foreign payment stablecoin issuer from paying a holder any interest or yield solely in connection with holding, using or retaining the stablecoin. The prohibition binds issuers rather than intermediaries.

How are stablecoins measured on the balance sheet?

ASU 2023-08 added ASC 350-60, which requires fair value measurement for crypto assets meeting six scope criteria. Fiat-backed stablecoins carrying a claim on the issuer are widely read as falling outside that scope, which leaves them measured under other guidance. Confirm the analysis for your specific instrument with your auditor.

What should a B2B crypto payment clause cover?

Unit of account separately from settlement asset, the rate source and moment, who bears the network fee, an address change procedure with out-of-band confirmation, transaction hash as evidence of payment, and mutual sanctions representations.

Take your three largest supplier relationships and check one thing in each: whether the address you would pay today has ever settled a payment before. Any that has not is a first payment, and first payments are where address substitution succeeds. Confirm those out of band this week. If your counterparties, invoices, screening and payouts already live in one dashboard with entry and release split by permission, as they do in VaultNow, that check is a report rather than an afternoon.

This article is general information, not legal, tax or accounting advice. Positions, rules and fees are stated as at 27 August 2026 and change. Check your own facts with your own adviser.