TL:DR
For most of the last two decades, a cross-border payment meant one thing. Money left a bank, crossed a chain of correspondent institutions, and arrived somewhere else two to five days later, having paid a fee at each hop and an exchange rate spread nobody itemised. The system worked, it was universal, and it was the only option.
That is no longer true. By 2026 a business moving money across borders is choosing between three settlement layers that coexist, each with different economics, coverage and failure modes. This guide covers where each stands, what the widely quoted numbers actually measure, what regulation changed, and what is still unsolved.
The three settlement layers
The first layer is correspondent banking, still the default and still the broadest. It reaches essentially every country and every currency, and it remains the only option for many corridors. Its weaknesses are unchanged: multi-day settlement, fees applied at each intermediary, and exchange rate spreads buried in the rate rather than shown on the invoice. The World Bank's data puts the global average cost of sending money across borders at around 6 percent, with banks the most expensive channel at close to 15 percent [1]. Those figures describe remittances rather than corporate treasury, and enterprise pricing is considerably better, but the structure of the cost is the same.
The second layer is domestic instant rails, which now exist in most major economies and increasingly connect to each other. FAST in Singapore, UPI in India, Pix in Brazil, FedNow and RTP in the United States, SEPA Instant across the euro area. Within their own borders these settle in seconds at very low cost. Reaching them from outside means holding a local collection point in each market, which is what named global accounts provide without a local entity, and what our virtual account orchestration guide covers in depth. The unfinished work is interoperability between them, which is why a business still cannot treat these rails as a single cross-border network.
The third layer is stablecoin settlement, which moved from experiment to production infrastructure over 2025 and 2026. It settles in seconds, operates continuously, and removes the correspondent chain from the middle of a payment. Its constraint is that the cost and friction sit at the edges, in the conversion into and out of local currency, rather than in the transfer itself.
None of the three wins outright. Which one is cheapest and fastest depends entirely on the corridor.
What the volume numbers measure
The most quoted statistic in this market is also the most misleading if taken at face value. Stablecoin on-chain transaction volume reached 33 trillion dollars in 2025, up 72 percent year over year, which exceeded the combined annual payment volume of the two largest card networks for the first time [2].
That figure does not describe payments. On-chain volume includes trading, arbitrage, automated market activity and transfers between exchanges and wallets. The number that matters for a business is real-world payment volume, which doubled in 2025 to approximately 400 billion dollars, with an estimated 60 percent of that being business-to-business activity [2].
The gap between 33 trillion and 400 billion is the honest picture of the market: the infrastructure is enormous and the payment usage on top of it is still comparatively small, but growing quickly and concentrated in B2B. Anyone presenting the larger number as evidence of payment adoption is either confused or selling something. The smaller number is the one that has doubled, and it is doubling from a base that is now material rather than negligible.
Corporate survey data supports the same reading. EY-Parthenon's post-GENIUS Act research found that 77 percent of corporate stablecoin users cite supplier payments as their primary use case, and that 41 percent reported cost savings of at least 10 percent [3]. That is a narrow, practical pattern of adoption, not a broad replacement of the banking system.
Regulation arrived before adoption
The usual pattern in financial technology is that a market grows and regulation catches up years later. Stablecoins went the other way. Between mid-2025 and mid-2026, five major jurisdictions put frameworks in place, and adoption accelerated afterwards rather than before.
In the United States, the GENIUS Act was signed in July 2025, establishing the first federal framework for payment stablecoins. It requires one-to-one reserve backing in high-quality liquid assets, applies Bank Secrecy Act obligations, and prohibits issuers from paying yield simply for holding a balance [4].
In the European Union, MiCA's provisions for fiat-referenced tokens have applied since 2024, and the transitional period for service providers ended in July 2026. Compliant tokens are authorised as e-money tokens; non-compliant ones have been delisted from EU venues [4].
Hong Kong introduced stablecoin issuer licensing through the HKMA in August 2025, with eight licensed issuers operating by April 2026, and is expanding its wider virtual asset regime into dealing, custody, advisory and management. Singapore operates a single-currency stablecoin framework requiring full reserve backing and short redemption timelines. Canada requires FINTRAC registration as a money services business with virtual currency permission, alongside Bank of Canada registration under the Retail Payment Activities Act for retail payment activity.
The practical effect for a business is that the diligence question changed. It is no longer whether stablecoin settlement is permitted. It is whether the specific provider holds the specific permission in the specific market, and whether the entity providing stablecoin services is the same entity providing fiat services, which it frequently is not. Our guide to the global payment licensing landscape covers each regime in detail.
Enterprise infrastructure moved in
The clearest signal of where this is heading is not adoption data but capital allocation. Stripe acquired stablecoin infrastructure company Bridge for 1.1 billion dollars, then jointly launched a blockchain built specifically for stablecoin payments. Circle launched its payments network in 2025 as a coordination layer connecting licensed institutions for cross-border settlement. Visa, Mastercard, Klarna, Western Union, Intuit, Fiserv, Cloudflare and PayPal have all integrated stablecoin rails or announced plans to [2].
These are not experiments by companies looking for a use case. They are payment companies with functioning businesses spending significant money to add a settlement layer. That tells you more about the direction than any forecast does.
What it does not tell you is that the old rails are going away. The same institutions are simultaneously investing in ISO 20022 migration and instant payment interoperability. The realistic read is addition rather than replacement.
Where stablecoin settlement wins
The corridor determines everything, and averages hide the answer. Settlement through the stablecoin layer removes the correspondent chain, so the saving is largest exactly where that chain is longest and most expensive. On corridors into markets with high banking costs, slow settlement and volatile local currency, the difference is substantial. On a well-served corridor where a domestic instant rail already settles in seconds at low cost, the difference is marginal or negative once conversion costs are counted.
The Federal Reserve's analysis makes the mechanism explicit: moving the stablecoin itself costs close to nothing, and the real cost structure lives at the on-ramp and off-ramp, shaped by local regulation, liquidity depth and how competitive the provider market is in that destination [7]. That is why a business should model its own corridors rather than apply a global average, and why the sensible first deployment is the two or three corridors where existing rails perform worst.
The mechanics of how the fiat, stablecoin and fiat legs fit together, and the instrument itself, are covered in our guide to how stablecoins work.
There is a second benefit that shows up on the balance sheet rather than in the fee line. Paying across many currencies through banking rails requires holding a prefunded balance in each one, sized for the worst case and idle the rest of the time. Funding a payment at the moment it is sent releases that capital back into the business. For a platform disbursing across eight or ten currencies, the working capital released is frequently a larger number than the transaction cost saved, and it is the one a finance team can act on immediately.
Agent-initiated payments
The second shift underway is not about the rail but about who starts the payment. Agent-initiated commerce, where an autonomous system discovers, compares and pays under rules set in advance, moved from demonstration to shipped product across 2025 and 2026.
The forecasts are large and disagree on scope rather than direction. McKinsey estimates AI agents could mediate 3 to 5 trillion dollars of global commerce by 2030, with around 1 trillion in orchestrated United States retail revenue alone [8][9]. The card networks moved in parallel: Mastercard launched its agent programme in April 2025 and extended it to machine payments in June 2026 [10], while Visa shipped tokenised credentials and spend controls designed for agents.
For cross-border businesses the consumer checkout case is the least relevant. The consequential versions are procurement systems that source and settle with suppliers across borders, routing layers that select a rail per payment, and machine-to-machine payments where software pays software by the API call. Our agentic payments guide covers the category in full.
The machine layer, where software pays software by the API call, deserves a note of caution, because it is where the numbers are most often misread. The x402 protocol reported roughly 165 million transactions across about 69,000 active agents by April 2026, which sounds like a mature market. Chainalysis found that much of that growth was driven by meme coin farming rather than commerce, with real daily commercial volume far smaller [11]. The useful signal in the same data is the composition: transactions of a dollar or more rose from 49 percent of value transferred in early 2025 to around 95 percent [11]. The traffic is noisy, and it is getting less noisy.
What is changing on the bank rails
It would be a mistake to read 2026 as a story only about new rails. The existing ones are changing under two deadlines that affect any business sending cross-border payments.
The first is message format. Cross-border payments now travel as ISO 20022 messages, and from 14 November 2026 messages carrying fully unstructured postal addresses will be rejected at the network level. At minimum, town name and country must sit in their own structured fields. By SWIFT's own assessment roughly 65 percent of messages still carried unstructured addresses, and a large share of banks were behind schedule [6]. This propagates upstream: if beneficiary addresses are captured as free text at onboarding, the payment fails later.
The second is beneficiary verification. Under the EU Instant Payments Regulation, Verification of Payee has applied to euro-area providers since October 2025, with the rest of the EU required to comply by July 2027 [5]. The check confirms that the account and the name correspond before a transfer is authorised, returning a match, close match or no match. This is a structural change in how payments are made, and it matters more as payments become automated, because an automated payout has nobody looking at the beneficiary at the moment money moves. It sits alongside the existing obligation for originator and beneficiary information to travel with a payment under the Travel Rule, and together they point at verification becoming a default step rather than an optional one.
What is still unsolved
Four problems remain genuinely open, and a business planning for 2027 should treat them as constraints rather than details.
Accounting treatment. Stablecoins fall outside the scope of the fair-value rules introduced by FASB ASU 2023-08, because a fiat-backed token carries a redemption claim against the issuer's reserves. Classification therefore sits between cash equivalents, financial instruments and intangible assets, and US GAAP does not yet give a definitive answer [12].
Agent identity. Establishing that an agent is genuinely authorised to spend for a named business, with bounded authority, is unresolved across jurisdictions. Network programmes and mandate-based protocols each address part of it.
Liability and disputes. Card disputes assume four parties. An agent-initiated payment adds a fifth, the platform, and no settled framework exists for who bears loss when an agent acts within its instructions and produces a bad outcome.
Interoperability. Domestic instant rails still do not connect cleanly across borders, which is the single change that would most reduce the cost of cross-border payments for the largest number of businesses. Linking projects exist between individual markets, but there is no general solution, and until there is, a business reaching many countries still needs a provider to bridge them.
None of these is a reason to wait. They are reasons to choose infrastructure that can absorb the answers when they arrive, rather than infrastructure built around one assumption about how the market will settle.
What this means for 2027
The conclusion that follows from all of the above is unglamorous: no single rail wins, so the capability that matters is routing across all of them.
A business that commits entirely to correspondent banking pays more than it needs to on corridors where alternatives exist. A business that commits entirely to stablecoin settlement finds it uneconomic on corridors already served by fast domestic rails, and unavailable where local conversion is thin. The durable position is a layer that selects per payment, on the basis of corridor, currency, amount and urgency, and keeps compliance and verification consistent regardless of which rail is used.
That is also why the initiator shift matters less than it first appears. When an agent rather than a person starts the payment, the routing, verification and audit requirements underneath are the same ones a well-built cross-border operation already has. The initiator is new. The infrastructure problem is familiar.
Where Tazapay fits
Tazapay operates as the last-mile layer across these rails rather than as a bet on one of them. Collections run through named multi-currency global accounts in 40+ currencies, on the local rails a payer in each market expects. Payouts reach beneficiaries in 100+ countries, with local payouts in 70+ countries and 99 percent of payouts completing in under 15 minutes.
Stablecoin settlement, provided by Tazapay Canada Corp, is available on the corridors where banking rails are slowest, including participation in Circle Payments Network as a beneficiary financial institution. Payee verification is available as an opt-in check before a payout is initiated. Tazapay does not sell a live agentic product, and the honest position is that the category is still forming, but the routing, verification and multi-rail settlement that agent-initiated payments will require are live today.
Sources
[1] World Bank. "Remittance Prices Worldwide, Issue 54." September 2025.
[2] Bessemer Venture Partners. "Stablecoins: From DeFi Primitive to Global Financial Infrastructure." April 2026.
[3] EY-Parthenon. "Cost Savings and Speed Drive Stablecoin Adoption." 2025.
[4] K&L Gates. "Crypto in 2026: The Democratization of Digital Assets." January 2026.
[5] European Commission. "New EU rules make instant euro payments faster and safer." October 2025.
[6] J.P. Morgan. "Five payment trends to watch in 2026." 2026.
[7] Federal Reserve Board. "Payment Stablecoins and Cross-Border Payments." FEDS Notes, March 2026.
[8] McKinsey & Company. "The automation curve in agentic commerce." January 2026.
[9] Retail Dive (ICSC / McKinsey). "US agentic commerce revenue forecast to reach $1 trillion by 2030." May 2026.
[10] Mastercard. "Mastercard launches Agent Pay for Machines." June 2026.
[11] Chainalysis. "Agentic Payments Cross the Threshold: Inside x402's Path to Meaningful Adoption." 2026.
[12] Forvis Mazars. "Accounting for Stablecoins: Navigating Uncertainty Within US GAAP." November 2025.
