How money crosses borders

What this material is for

When a small business in Budapest pays 10,000 euros to a supplier in Istanbul, the money sets off on a journey most of us know nothing about. We press a button in online banking, and a day or two later the money arrives, a little less than what we sent. Along the way it passed through half a dozen institutions, and each of them kept a bit of it.

In recent years a second route has opened up. That route is the stablecoin: a digital token that moves on a blockchain and whose value is pegged to the dollar or the euro. Some see it as the successor to SWIFT, others as crypto risk. To judge which view is closer, you need to know both routes equally well.

Who is it for? Anyone curious about how money gets from one country to another, and what blockchain changes about that. Business owners and finance people who pay abroad or get paid from abroad. People working on public policy and regulation. University students. And anyone who has heard about stablecoins and wants to know what is behind the headlines.

No prior knowledge needed. You do not have to be an economist, a banker or a programmer. Every technical term is explained when it first comes up, and you can look it up any time in the glossary in Appendix A. The technical parts are introduced with an analogy.

How should you read it? The chapters build on each other, but each one also stands on its own. If you only have an hour, read this introduction, the “In short” at the end of each chapter, and the FAQ in Chapter 15. If you want to go deeper, go in order and take time with the diagrams, because each one draws a mechanism.

By the end you will understand:

• what money really is (an entry in someone’s ledger);
• how a forint gets to another bank, and a dollar to another country;
• what SWIFT does, and what it does not do;
• what a blockchain is, what a stablecoin is, and why it is classed as crypto;
• why a stablecoin transfer involves so many conversions;
• how much of a competitor stablecoins really are to the traditional system;
• who the players are, what licences they work under, and what it takes to move up a tier.

This material is not investment or legal advice. The goal is for you to see through the system, so you can judge news, offers and decisions with your own common sense.

The whole thing on one page

Picture two islands.

The left island is the traditional financial world. Here all money lives in the books of some institution. At the bottom sit the central banks (Magyar Nemzeti Bank, the European Central Bank, the US Fed); they keep the banks’ accounts. Above them are the commercial banks, which keep their customers’ accounts. The banks are connected by a messaging network, SWIFT, and by accounts they hold with each other (correspondent banking relationships). Further up come the payment institutions and e-money institutions (Wise, Revolut, Western Union) and the card networks (Visa, Mastercard), and at the top, us, the users.

The right island is the blockchain world. Here money lives in a shared ledger that anyone can see, kept in sync by thousands of computers. Crypto-assets (bitcoin, ether) and stablecoins (USDT, USDC, EURC) move on this ledger. The island is home to exchanges, wallets, custodians and analytics firms.

The two islands are connected by bridges:

1. The stablecoin issuer’s reserve. Behind every backed stablecoin are dollars or euros held in bank accounts and government bonds. This is the deepest bridge.
2. On-ramps and off-ramps. This is where forints become tokens, and tokens become forints again. Mostly crypto exchanges play this role.
3. Corporate stablecoin payment providers, which do both conversions on the customer’s behalf.
4. Card networks, which already settle in stablecoins too.
5. Banks’ tokenized deposits and SWIFT’s own blockchain ledger.
6. Central bank money on a blockchain (the ECB’s Pontes system, since September 2026).

The two islands are slowly growing together. Banks are becoming crypto service providers, crypto firms are getting banking licences, and SWIFT is building a blockchain. The material follows this process throughout.

Three ideas everything rests on

1. Every payment is a ledger change. In no payment does money travel down some wire. A number goes down in someone’s books, and a number goes up somewhere else. The only question is how many ledgers have to be reconciled along the way.

2. The message and the money are two separate things. SWIFT carries a message (“debit our account, credit X”). The money only moves when the banks rewrite the numbers in their own books. Most of what follows about international payments rests on this difference.

3. Cost arises at the borders. Every time you have to cross from one ledger to another, or from one currency to another, you pay for it in fees, spread and waiting. That holds in the traditional world and in the blockchain world alike.

1.1 Three kinds of money that look the same

The 1,000,000 forints you see in online banking is a number on a screen. Your Revolut balance and the USDC in your crypto wallet look exactly the same. Legally, though, they are three different things, because each one is a debt owed to you by a different institution.

Central bank money. This is a debt of the central bank, and it exists in two forms:

• cash (banknotes, coins) is the only central bank money an individual can hold;
• the other is the money in banks’ accounts at the central bank (reserves). In Hungary these accounts are kept by the MNB (Magyar Nemzeti Bank, the central bank of Hungary); in the euro area, by the Eurosystem’s T2 system.

In its own currency, central bank money is risk-free, because a central bank cannot run out of its own money.

Commercial bank money, i.e. deposits. Your bank balance is legally your bank’s debt to you. The vast majority of money in circulation is of this kind. Banks mostly create it by lending: when they pay out a loan, a new deposit appears in the borrower’s account. Deposits are protected by the National Deposit Insurance Fund (OBA, Hungary’s deposit guarantee scheme) up to 100,000 euros per bank and per customer.

Electronic money (e-money). A claim issued by an electronic money institution (EMI) or a bank against funds paid in advance, at 1:1. A prepaid card balance is e-money, as is a balance held with a fintech provider if the provider operates under an EMI license. An EMI cannot lend out the money paid in. It has to safeguard it separately and redeem it at face value on request. Deposit insurance does not cover e-money.

This third category deserves attention because under EU regulation (MiCA, see Chapter 11) any stablecoin pegged to a single official currency (euro, dollar, forint) is legally e-money. So only a bank or an e-money institution may issue one.

“One euro”, then, is really a whole family of debts of different institutions (the central bank, a commercial bank, an e-money institution, a stablecoin issuer). We can treat them as interchangeable because the system guarantees they convert into each other at 1:1. Economists call this the singleness of money. The Bank for International Settlements (BIS) faults stablecoins precisely for failing this test: every token carries the name and the risk of its issuer.

1.2 What does “fiat” mean?

“Fiat” is a Latin word meaning “let it be”, “so be it”. Fiat currency is money whose value comes from the state declaring it legal tender and from people trusting the issuing central bank. There is no gold or other commodity behind it that you could exchange it for. The forint, the euro, the dollar and the yen are all fiat money. The US cut the dollar’s last link to gold in 1971 (ending the Bretton Woods system), and since then every major currency in the world has been fiat.

In crypto jargon, “fiat” simply means traditional government money (HUF, EUR, USD), as opposed to crypto-assets. A “fiat on-ramp” is the gate through which traditional money enters the crypto world. And a “fiat-backed stablecoin” has traditional money (and short-term government bonds) behind it.

1.3 Every payment is a ledger change

A ledger records who owes whom how much. Every institution in the monetary system keeps one:

When you pay, no coin ever “travels” down any wire. A payment is a coordinated series of debits and credits in one or more ledgers:

• Within the same bank (an “on-us” transfer) a single ledger changes. The bank lowers one customer’s balance and raises another’s. It is instant and almost free.
• Between two domestic banks the payer’s bank debits the payer, the payee’s bank credits the amount, and the two banks settle what they owe each other on the central bank’s ledger. In Hungary this happens through VIBER or the instant payment system.
• Between two countries and two currencies there is no shared central bank ledger, so banks hold accounts with each other (correspondent banking, Chapter 3). More ledgers and more steps mean more time, more cost and more risk.
• On a blockchain a token transfer changes a single shared ledger. But getting money into that ledger, and back out of it, also requires changes to bank ledgers (Chapter 8).

This picture, of who keeps which ledger, is the most useful tool in the whole material. Whenever you get stuck, ask yourself: how many ledgers are in this story, and who keeps them?

2.1 Clearing and settlement

These two words often get blurred together, but they are two separate steps.

Clearing. In this step the instructions are transmitted and reconciled, and it is worked out who owes whom how much. This often uses netting. If bank A owes B 100 units and B owes A 80, then at the end of the day it is enough for A to pay B 20. This saves a great deal of liquidity, but until settlement the parties carry credit risk against each other.

Settlement. This is where the debt is actually discharged, by transferring the settlement asset. That asset is almost always central bank money. Settlement is final when it is irrevocable and unconditional. In the EU this is protected in payment systems by the Settlement Finality Directive (98/26/EC).

In gross settlement every payment is settled individually, immediately. In net settlement payments are collected and only the difference is settled.

How does netting save money? When many banks pay each other, the clearing house totals for each of them how much it has to send and how much it receives. This is called multilateral netting. By the end of the day every bank is left with a single net position: it either pays or receives. These net amounts are settled on the central bank accounts, so many gross payments turn into a few smaller transfers. Large and urgent payments, however, are not netted. They settle one by one, gross, in the RTGS (section 2.2).

2.2 The central bank’s role: RTGS

RTGS (real-time gross settlement) is the central bank’s own system. It settles large-value and urgent interbank payments one by one, in real time and in central bank money. Here the central bank acts as settlement agent: it keeps the banks’ accounts and moves money between them. Because central bank money is what moves, settlement carries no counterparty risk.

• Hungary: VIBER (the Hungarian real-time gross settlement system), run by the MNB. In Q1 2026 it settled about 785,000 items worth roughly HUF 750 trillion (750,000 billion). The amount is so large because interbank money market and securities transactions also run through it.
• Euro area: T2 (the successor to TARGET2), the Eurosystem’s RTGS. In 2025 it settled an average of EUR 1,932.8 billion a day, in about 431,000 payments.
• US: Fedwire Funds. In 2025 an average of USD 4.59 trillion a day went through it, in 869,000 transfers.

RTGS systems have opening hours; most are closed on weekends and public holidays. Part of the delay in international payments comes from this.

A word on big numbers. Throughout this material “billion” means a thousand million (10⁹) and “trillion” a thousand billion (10¹²). The Hungarian original uses “milliárd” and “billió” for these.

2.3 Retail and business payments: GIRO and instant payments

Most everyday transfers run through a clearing house. In Hungary this is GIRO Zrt., wholly owned by the MNB since 2014. It runs two systems:

• Intraday clearing (ICS, part of the Interbank Clearing System): traditional transfers go through it. In Q1 2026, about 108 million items worth HUF 70 trillion.
• AFR (Hungarian instant payment system): live since March 2, 2020. Any transfer up to HUF 20 million arrives in about 5 seconds, day and night, weekends included. In Q1 2026 it handled about 53 million instant transfers worth HUF 10.6 trillion. By Q3 2025 roughly every second transaction already went through AFR. Its add-on services include secondary identifiers (phone number, email address, tax number), request-to-pay, and qvik (QR-code and NFC payments, since September 2024).

AFR settles against the accounts held at the MNB, that is, in central bank money. So the payee’s bank does not have to take credit risk on the payer’s bank.

2.4 Euro area: SEPA and SEPA Instant

SEPA (Single Euro Payments Area) is the common rulebook for euro transfers. It covers 41 countries (the 27 EU member states, 3 other EEA countries and 11 others, including the United Kingdom and Switzerland). Including overseas territories, that is 57 countries and territories. Within SEPA an IBAN is enough to make a transfer, and a cross-border euro transfer costs the same as a domestic one.

• SCT: the traditional SEPA credit transfer, usually arriving the next business day.
• SCT Inst (SEPA Instant): an instant euro transfer, in about 10 seconds at most. Since the 2025 rulebook there is no scheme-wide upper limit; banks and customers can set their own limits.
• TIPS: the Eurosystem’s instant settlement service, which settles instant payments in central bank money, 24/7. In 2025, 2.47 billion such payments went through it (82.5% more than a year earlier).

The EU Instant Payments Regulation (2024/886) makes instant euro transfers mandatory. Euro-area banks had to be able to receive them from January 9, 2025, and to send them from October 9, 2025. In member states outside the euro area, Hungary included, for accounts held in euros it becomes mandatory from January 9, 2027 (receiving) and July 9, 2027 (sending). Under the regulation an instant transfer cannot cost more than a regular one, and it also makes Verification of Payee mandatory, meaning the bank has to check the name against the IBAN. Payment institutions and e-money institutions have slightly different deadlines (April 9, 2027; July 9 in non-euro countries).

Similar systems run in other parts of the world: PIX in Brazil, UPI in India, Faster Payments in the United Kingdom, FedNow in the US (since 2023; by July 2026 more than 1,800 banks and credit unions had joined).

2.5 A Hungarian peculiarity: the transaction levy

In Hungary the financial transaction levy makes every transfer more expensive. In 2026 transfers carry a levy of 0.45%, capped at HUF 20,000 per transaction. Currency exchange carries a 0.45% levy plus a 0.45% supplementary levy (since October 1, 2024), 0.9% in total. Banks typically pass it on to the customer. There are exemptions too, for example certain instant payments initiated by individuals via request-to-pay or qvik, or transfers between your own accounts at the same bank. The exact rules are in your bank’s published terms.

On a HUF 4 million business transfer this can come to as much as HUF 18,000, and more with currency exchange. In Chapter 9 this item comes up again when comparing the SWIFT and stablecoin routes, because forints sent to a crypto exchange also count as a transfer.

2.6 The lesson

Domestically, and within the EU in euros, payments are already almost instant and cheap today. What makes this possible is a shared central bank ledger on which every bank holds an account. So the speed stablecoins promise adds little that is new at home. The real problem is cross-border, cross-currency transfers. A 2026 study by the Banca d'Italia also concluded that instant payment systems and stablecoins are more complementary than competing.

3.1 Why are cross-border payments hard?

Domestically every bank holds its account at the same central bank, so there is a shared ledger on which they can settle with each other. Internationally there is no such thing. There is no world central bank, and no world ledger either. A forint account lives in the Hungarian system (ultimately settling at the MNB), a dollar account in the American one (ultimately at the Fed). If we want to get dollars from forints into a US account, we need someone who holds accounts in both systems and converts one currency into the other.

Banks fill in the missing shared ledger with an old trick: they open accounts with each other.

3.2 What SWIFT is, and what it is not

SWIFT (Society for Worldwide Interbank Financial Telecommunication) was founded in 1973. It is a cooperative under Belgian law, owned by its member banks, headquartered in La Hulpe, Belgium. It connects more than 11,500 institutions in more than 200 countries and territories. It is overseen jointly by the G10 central banks, with the National Bank of Belgium as lead overseer.

SWIFT is a messaging network. It is not a bank, not a payment system, it keeps no accounts, and it moves no money. It carries standardized, authenticated instructions between banks, such as: “debit our account with you by 10,000 euros and credit the account of Yılmaz Tekstil.” The money moves when the banks rewrite their own books on that basis.

An analogy: SWIFT is like a registered-mail service between banks. The postal service guarantees that the letter is authentic, intact and delivered, but it does not carry the money mentioned in it. The money “moves” in the banks’ books.

That does not make SWIFT irrelevant to costs. SWIFT messages trigger the accounting entries at the correspondent banks, but the fees and the spread are charged by the banks, not by SWIFT.

A few numbers on scale: in 2024 an average of 53.3 million messages a day went through the network, and in 2025 a single day set a record of more than 68 million. Because international interbank traffic runs on SWIFT, cutting a country’s banks off from SWIFT is a severe sanction (it happened to Iranian banks, and in 2022 to several Russian banks).

3.3 Identifiers: BIC and IBAN

• BIC (Bank Identifier Code, commonly called the “SWIFT code”), ISO 9362: 8 or 11 characters long, it identifies an institution. For example OTPVHUHB: OTP Bank (OTPV), Hungary (HU), location code (HB). The optional three-character part denotes the branch.
• IBAN (International Bank Account Number), ISO 13616: identifies an account. Structure: country code + 2 check digits + domestic account number. A Hungarian IBAN is 28 characters long, e.g. HU42 1177 3016 1111 1018 0000 0000.

Within SEPA the IBAN is enough. For a SWIFT transfer you usually have to give the BIC of the payee’s bank, the account number or IBAN, and the payee’s name and address. From November 2026, SWIFT no longer accepts unstructured (free-text) postal addresses.

3.4 The messages: MT103, MT202 and ISO 20022

For decades SWIFT worked with text messages in MT (Message Type) format. The two best known:

• MT103: a transfer initiated on behalf of a customer (payer → payee). Well-known fields: :20: the sender’s reference, :32A: value date, currency, amount, :50K: the ordering customer, :59: the beneficiary, :71A: the charges.
• MT202 / MT202 COV: a transfer between banks. This is how banks move the “cover” between themselves.

The user also needs to know the charges code:

• OUR: the sender pays all fees;
• SHA: shared; the sender pays their own bank’s fee, the payee pays the intermediaries’ fees and their own bank’s;
• BEN: the payee pays all fees, deducted from the amount.

With SHA and BEN, intermediary banks can deduct their fees along the way, so the payee receives less than what was sent.

The new global standard is ISO 20022. It is an XML-based format and carries far more structured data. Its equivalents: pacs.008 (customer transfer, the successor to MT103), pacs.009 (interbank transfer, the successor to MT202), pacs.004 (return), camt.053/054 (statement, notification), pain.001 (customer → bank instruction). For cross-border payment instructions, the coexistence of MT and ISO 20022 on SWIFT ended on November 22, 2025. Since then payment instructions go in ISO 20022, and more than 97% of traffic already does. The industry still says MT103, but what actually goes out today is a pacs.008 message.

Structured data (name, address, purpose of payment, remittance information) makes automated sanctions and anti-money-laundering screening easier, reduces manual repairs, and lets more payments run end to end without human intervention (straight-through processing). This only works if every player in the chain passes the data on in full.

Serial and cover methods. A transfer through correspondent banks can be run with two message flows. In the serial method the customer message (formerly MT103, today pacs.008) travels bank by bank along the chain, and each bank debits and credits in its own books. In the cover method the sending bank sends the customer message straight to the payee’s bank, and moves the money (the “cover”) with a separate interbank message (MT202 COV, today pacs.009 COV) through the correspondent banks. The COV variant also carries the original ordering customer’s and beneficiary’s details, so the correspondent banks can do their screening too. The example in section 3.6 follows the cover method.

The fields of the old MT103 are short codes with free text: in the :50K: field, for example, the ordering customer’s name and address each sit on a line of text. In ISO 20022 the same thing appears in separate, tagged XML elements: Dbtr (payer), Cdtr (payee), IntrBkSttlmAmt (interbank settlement amount), RmtInf (remittance information), and the address is split into parts too (street, building number, postcode, town, country). A machine can check this far more reliably.

3.5 Correspondent banks, nostro and vostro

A correspondent bank keeps accounts for other banks, often foreign ones, and makes or receives payments on their behalf.

A nostro account (Italian for “ours”) is the account that our bank holds with a foreign bank, in that bank’s currency. If OTP has a dollar account with a New York bank, from OTP’s point of view that is a nostro: “our account with them”.

A vostro account (“yours”) is the same account seen from the foreign bank’s side. The New York bank calls OTP’s account with it a vostro: “your account with us”.

So it is one and the same account, just seen from two sides. On OTP’s balance sheet it is an asset (a claim on the New York bank); on the New York bank’s balance sheet it is a liability (a debt to OTP).

Prefunding and trapped liquidity. For a bank to pay in a given currency, it has to keep money in its nostro accounts in advance, in many countries and many currencies. That money earns little interest and cannot be used for anything else. This is one of the hidden but large costs of the traditional system, and banks ultimately pass it on to customers. This is where blockchain-based solutions hope to save money: instead of prefunded nostro accounts, they offer on-demand settlement.

The shrinking correspondent network (de-risking). In the 2010s many banks terminated correspondent relationships, mainly with smaller countries or those seen as riskier, over money-laundering concerns. According to data from the BIS payments committee (CPMI), in the seven years to 2018 the number of active correspondent banking relationships fell by about 20%, and the number of corridors (country pairs) by about 10%. In Latin America and the Caribbean, by 30% since 2012. Fewer relationships mean longer chains and higher costs.

3.6 A transfer, step by step

Let’s take a concrete case. A Budapest SME (Kovács Kft.) pays 10,000 dollars to a supplier in Vietnam. The company pays from a forint account.

1. Order and conversion. Kovács Kft. enters the transfer in online banking. Its bank converts the forints into dollars at its own rate (the first spread).
2. Message. The Hungarian bank has no direct relationship with the Vietnamese bank. So it sends a pacs.008 message over SWIFT and arranges the cover separately through its dollar correspondent bank, a New York bank where it holds a nostro account.
3. Settlement in the US. The New York correspondent debits the Hungarian bank’s account with it and credits the amount to the Vietnamese bank’s correspondent. If that is another US bank, this usually happens through the US systems (Fedwire or the privately owned CHIPS). If both banks use the same correspondent, an internal transfer at that bank is enough, without Fedwire.
4. Credit in Vietnam. The Vietnamese bank’s correspondent credits the Vietnamese bank’s dollar account, and the Vietnamese bank credits the supplier, possibly converting to dong (the second spread).
5. At every step come sanctions and anti-money-laundering screening, possible fee deductions, cut-off times and time-zone differences.

So the message and the money travel separately. The message crosses SWIFT in seconds. The money moves as accounting entries; at every correspondent bank and in the US RTGS, a ledger changes.

3.7 Why is it slow and expensive?

No single cause explains why international transfers are slow and expensive. Seven factors add up:

1. Cut-off times. Every bank and every RTGS has a daily deadline. Anything that misses it is processed the next business day.
2. Time zones and opening hours. Business hours in Budapest, New York and Hanoi overlap only briefly. On weekends and public holidays settlement stops.
3. Compliance checks. Every intermediary does customer identification (KYC), anti-money-laundering (AML) and sanctions screening. If a name looks suspiciously like one on a sanctions list, or an address is missing, a manual review, a query back or a rejection follows.
4. Multiple intermediaries. Each one can charge a fee, and each one can cause a delay.
5. Spread. The difference between the mid-market rate and the rate given to the customer. For retail customers this is often the largest and least transparent cost (Chapter 4).
6. Prefunding. The cost of capital parked in nostro accounts.
7. The “last mile”. According to SWIFT’s own research, about 80% of the total time of an international transfer passes after the payment has left the SWIFT network: at the beneficiary bank, in local systems and in checks.

Prefunding up close. Item 6 on the list is less conspicuous, but it is a big one. An international bank pays in many currencies, so it keeps prefunded nostro accounts with many correspondent banks: dollars in New York, euros in Frankfurt, pounds in London, lira in Istanbul. These balances must never run dry, or the bank could not pay out. So the money sits there, earns little interest, and cannot be put to any other use. This is called trapped liquidity. Banks build its cost into their fees and spreads.

On-demand liquidity is the reverse. The destination currency is not held in advance but obtained at the moment of payment, for example through a shared ledger or a stablecoin. This is the real economic target of blockchain-based solutions (section 3.5).

3.8 SWIFT gpi and the speed-up

In 2017 SWIFT introduced gpi (global payments innovation). Every payment gets a unique identifier, the UETR (Unique End-to-end Transaction Reference), and a tracking system shows where it is at any moment, much like parcel tracking. The goal of gpi is for money to be credited the same day, for fees and exchange rates to be transparent, and for the remittance information to arrive unchanged.

The figures for October 2024: 90% of cross-border payments sent over SWIFT reach the beneficiary’s bank within an hour, but only 43% reach the end customer’s account within an hour. In 2025 SWIFT reported that 75% of payments arrive at the beneficiary bank within 10 minutes. Don’t confuse the two: money arriving at the beneficiary’s bank does not mean the customer has been credited yet, and the G20 targets are measured all the way to the end customer.

Tracking works like this. The UETR is a 36-character identifier that the sending bank assigns to the payment, and it travels unchanged in every message along the chain. Every bank the payment passes through updates its status in the central gpi Tracker: received, forwarded, fee deducted, credited. So the sending bank, and through it the customer, can see step by step where the money is.

SWIFT Retail Payments scheme. In June 2026 SWIFT launched a rulebook for retail and SME transfers. More than 50 banks support it, and a first group of more than 25 banks committed to going live by the end of June 2026 on corridors in 11 countries (including the US, the United Kingdom, Germany, Spain, India, China and Australia). Under the rules the full amount arrives (nothing is deducted along the way), the fee and the timing are known upfront, and the payment can be tracked.

3.9 SWIFT’s blockchain experiments, 2025–2026

SWIFT’s plan is to become the layer that connects the other systems, blockchains included.

• September 29, 2025 (Sibos conference, Frankfurt): SWIFT announced it would add a blockchain-based shared ledger to its infrastructure. The first use case is real-time, 24/7 cross-border payments. The prototype is being built by Consensys (one of the big software companies in the Ethereum ecosystem), and more than 30 institutions from 16 countries took part in the design. According to Consensys’s CEO, the prototype is built on the company’s Linea technology; SWIFT has not officially confirmed this.
• July 9, 2026: SWIFT announced that the ledger is ready for use and that 17 banks would begin piloting cross-border payments with tokenized deposits on it (including ANZ, BNP Paribas, BNY, Citi, DBS, HSBC, Lloyds, MUFG, Standard Chartered, UBS and Wells Fargo). No Central and Eastern European or Hungarian bank is among them.
• August 19, 2026 (date of the press release): the first live transaction took place between HSBC and Standard Chartered. The banks messaged each other over the ledger, recorded the obligations as tokenized deposits in their own books, SWIFT matched and netted them, and final settlement ran over the existing systems.

Tokenized deposits are covered in more detail in Chapter 10. For now it is worth knowing that SWIFT is building on digital money issued by banks, not on stablecoins from non-bank issuers.

4.1 The mid-market rate and the spread

The foreign exchange (FX) market is the largest market in the world. In April 2025 its turnover was 9.6 trillion dollars a day (according to the BIS triennial survey).

On the interbank market every currency pair has a bid price and an ask price. The midpoint between them is the mid-market rate, the one you see in search engines and on financial sites.

The customer always gets a worse rate than that, and the difference is called the spread. Banks often advertise "fee-free" conversion while applying a 2–4% spread. For retail customers the spread is typically 1–5%.

The hidden cost is easy to calculate:

hidden cost ≈ amount × |rate received − mid-market rate| / mid-market rate

An example. A company converts 1,000,000 forints to euros. Say the mid-market rate is 1 EUR = 400 HUF, and the bank sells euros at 408 forints (a 2% spread). At the mid-market rate the company would get 2,500 euros; at the bank it gets 2,451. The difference is 49 euros, about 20,000 forints. That is roughly the cap on the financial transaction levy, and both come on top of the transfer fee.

This is also where the value date comes in: the day the money actually becomes usable in the beneficiary's account. It can differ from the booking date, and in international transfers the gap itself is a cost, since the money sits idle until then.

4.2 Settlement risk in FX trades: CLS

An FX trade involves two payments: one party pays forints, the other pays euros. What happens if one party has already paid out its currency and the other goes bankrupt in the meantime? This is called Herstatt risk, after the 1974 collapse of the German Bankhaus Herstatt. That day the bank had already received the Deutsche marks, but never paid out the dollars, because it was shut down in between.

The fix came with CLS (Continuous Linked Settlement). It is a special-purpose bank launched in 2002 that settles both legs of an FX trade at the same time: either both happen or neither does. The principle is called payment versus payment (PvP). CLS works in 18 currencies; in the first half of 2025 it settled about 7.9 trillion dollars a day on average, with 76 settlement members. The forint has been a CLS currency since November 16, 2015 (it was the 18th). For Hungary this means that large interbank HUF/EUR and HUF/USD trades can settle without Herstatt risk. The Turkish lira is not a CLS currency.

When an SME converts forints to euros, that does not go through CLS; the bank handles it on its own books. CLS settles large FX trades between banks. Blockchain projects (Agorá, mBridge) are trying to make this atomic, "both or neither" kind of settlement universal (Chapter 10).

4.3 Card networks in brief

Visa and Mastercard are so-called four-party schemes. The four parties are the cardholder, the issuing bank, the acquiring bank (acquirer) and the merchant. The card network routes authorization messages in real time and performs clearing. The actual settlement happens later, a day or two afterwards, through settlement banks. The merchant pays a fee made up of the interchange fee, the scheme fee and the acquirer's margin. In the EU the interchange fee on consumer cards is capped: 0.2% for debit cards and 0.3% for credit cards.

When you pay by card abroad, the currency is converted either by the card network or by the merchant's terminal. The latter is dynamic currency conversion (DCC): the terminal offers to charge you in forints. That is usually the more expensive option.

A card payment happens in three steps. At authorization, the terminal asks the issuing bank over the card network whether there are sufficient funds, and the answer comes back within seconds. No money moves yet; the amount is only put on hold. At clearing, at the end of the day, the acquirer submits the merchant's transactions, and the network works out which bank owes how much to which. At settlement, a day or two later, the money actually moves through the settlement banks, and the merchant receives the amount minus the fee.

The card networks are taking stablecoins seriously too. Visa's stablecoin-based settlement passed an annual run rate of 20 billion dollars in September 2026 (more than 15 times higher than a year earlier). Mastercard closed its acquisition of the stablecoin payments company BVNK on August 3, 2026 (at an announced price of up to 1.8 billion dollars). More on this in Chapters 9 and 13.

4.4 How does Wise work?

Wise (formerly TransferWise) is built on a simple insight: the money often never crosses the border at all.

1. You pay forints into Wise's local forint account in Hungary. That is a domestic transfer, done in seconds via AFR (Hungary's instant payment system).
2. Wise pays the beneficiary from a local account in the destination country, for example from its own euro or dollar account, again with a domestic transfer.
3. Because many customers send money in both directions, Wise nets the flows and only periodically rebalances the remainder between the pools of money it holds in each country (through banks and the FX market).
4. In more and more countries Wise connects directly to domestic payment systems (for example the UK's Faster Payments, and in 2026 also in Brazil and Japan), so it bypasses correspondent banks.

Wise's numbers for the fiscal year ending March 31, 2026: about 243.5 billion dollars in cross-border volume, an average take rate of 0.52% of volume, and in the fourth quarter 75% of payments arrived instantly (within 20 seconds). In the first quarter of fiscal 2027 it had 11.9 million active customers.

In the EU, Wise operates through a Belgian company, Wise Europe SA, a payment institution supervised by the National Bank of Belgium. In the UK, Wise Payments Ltd is an electronic money institution.

The lesson for stablecoins: between the major currencies, Wise achieves stablecoin-like speed and sub-1% cost without any blockchain, purely through netting, local accounts and direct system access. In a 2026 study by the Banca d'Italia, the USDC route turned out cheaper than Wise on 3 corridors and more expensive on 4.

4.5 The traditional money transfer operators: Western Union, MoneyGram

Western Union and MoneyGram are built on agent networks. You can pay in cash at shops and post offices, and the beneficiary can collect it from an agent too, even in cash. They keep their own internal ledger, net the flows in-house, and pay out in the destination country from pre-funded money. They use SWIFT only for large treasury rebalancing. In the EU, Western Union International Bank GmbH is an Austrian credit institution, and MoneyGram International SA is a Belgian payment institution.

A traditional money transfer goes like this. The sender pays the amount and the fee at an agent, and the operator converts at its own rate. The system issues a reference number (at Western Union this is the MTCN, Money Transfer Control Number), which the sender passes on to the beneficiary. The paying agent in the destination country checks the number and the beneficiary's identity in the central system, then pays out from its own pre-funded money. So the message gets across instantly, but the money travels separately: the operator and the agents settle with each other later, in aggregate, through their bank accounts. The fee and the spread arise when the transfer is sent.

Both companies have stepped into the stablecoin world. On May 4, 2026, Western Union launched USDPT, a dollar stablecoin on the Solana blockchain (issued by the US-based Anchorage Digital Bank), along with a "Digital Asset Network" that connects crypto wallets to Western Union's cash payout points. In September 2025 MoneyGram launched a USDC-based app in Colombia.

4.6 Real costs and the G20 targets

The World Bank's Remittance Prices Worldwide database measures how much it costs to send a 200-dollar remittance (when someone abroad sends money home to their family).

• The global average cost is 6.36% (Q3 2025).
• By channel (Q1 2025): banks 14.55%, post offices 7.71%, money transfer operators 5.04%, mobile operators 4.97%. Digital remittances cost 4.85%, non-digital 7.16%.

According to IFAD (the UN's International Fund for Agricultural Development), 728.6 billion dollars in remittances flowed into low- and middle-income countries in 2025.

In 2021 the G20, coordinated by the Financial Stability Board (FSB), set targets for improving cross-border payments:

There is no cost target for large-value (wholesale) payments. According to the FSB's October 2025 report, the 2027 targets will probably not be met; the FSB chair said the industry is "far from the targets." This shortfall is part of what has made policymakers more open to new solutions.

5.1 The shared ledger

In Chapter 1 we saw that every bank keeps its own ledger, and a payment is a coordinated change to those ledgers. A blockchain solves the same problem differently.

A blockchain is a ledger of which many independent computers keep identical copies. Following a shared set of rules (the protocol), the machines decide without a central bookkeeper which new entry is valid.

Picture a village where every resident keeps a notebook of who has how much money. Whoever pays shouts it out in the main square, the others check whether they have the funds, and if the majority finds it in order, everyone writes it into their own notebook. You don't have to trust any single bookkeeper, because if someone falsifies their notebook, it won't match everyone else's.

• Bank ledger: owned by a single institution. You trust the bank and its supervisor.
• Blockchain: thousands of copies exist. You trust the protocol's math, the economic incentives, and the assumption that no one holds the majority of the validating power.

A blockchain is one kind of distributed ledger technology (DLT). Regulators (the ECB, ESMA) often prefer to say "DLT", especially when the ledger is permissioned.

5.2 Blocks and the chain

Transactions are collected into blocks. Every block contains the hash of the previous block, and that is how the chain is formed.

A hash function produces a fixed-length "fingerprint" from any data. Change a single character in the input and a completely different fingerprint comes out. So if someone rewrites an old transaction, that block's hash changes, and every later link in the chain "breaks", which means the forgery is exposed immediately. An Ethereum transaction ID (its hash) looks like this: 0x5c50…a1f3 (66 characters).

Every block has two parts. The header holds the previous block's hash, a timestamp, a summary fingerprint built from all the block's transactions (the Merkle root), and on proof-of-work chains also a number called the nonce, which miners keep changing until the block's hash satisfies the rule. The body is the list of transactions itself. A block's own hash is computed from its header. So if even a single transaction changes in block 2, the Merkle root and the hash of block 2 change too, and the "previous hash" stored in block 3's header will no longer match. The forger would have to recompute every block from block 2 onwards, and do it faster than the rest of the network.

5.3 Consensus: how do strangers agree?

Consensus is the rule by which the machines on the network decide who gets to write the next block.

• Proof of Work (PoW): this is what Bitcoin uses. "Miners" compete to solve a computational puzzle; the winner writes the next block and is rewarded with new bitcoin and fees. Rewriting the past would mean redoing all that work, which is extremely expensive. The method consumes a lot of energy.
• Proof of Stake (PoS): this is what Ethereum (since September 15, 2022), Solana and many other chains use. Validators lock up collateral in the chain's own coin, and they are the pool from which the protocol picks who proposes and who attests blocks. Anyone who cheats has part of their stake destroyed by the system ("slashing"). The Ethereum Foundation estimates that this needs about 99.9% less energy than PoW.
• Permissioned consensus: only approved participants (banks, for example) validate. It is faster and more closed, but you have to trust the group of operators.

Whichever method a chain picks, the goal is the same: every honest participant in the network accepts the same ledger, in the same order, and no one can spend the same money twice. What differs is what cheating costs. With proof of work, it costs the burned energy and the hardware; with proof of stake, the cheater's own stake.

5.4 Public and permissioned chains

In 2025–2026 the two worlds are moving closer together. Banks are bringing their tokenized deposits onto public chains (JPMorgan's dollar deposit token, JPMD, has been available to institutional clients on the Base network since November 2025). Payment companies are launching their own chains optimized for payments: Tempo, the chain from Stripe and Paradigm, went live on March 18, 2026, and Circle's Arc on September 16, 2026. On Arc, network fees are paid in USDC, and it is run by a curated set of validators that includes BlackRock, Visa and Mastercard.

5.5 Keys, addresses, wallets

For many readers this is the least familiar part, so after the definitions there is an analogy.

• Private key: a secret 256-bit number. Whoever knows it controls the money that belongs to it. You often meet it in the form of a seed phrase: 12 or 24 words (mostly English) from which the key can be restored.
• Public key: mathematically derived from the private key. Easy in one direction, practically impossible in reverse (elliptic curve cryptography).
• Address: a shorter identifier derived from the public key, a kind of blockchain IBAN. An Ethereum address starts with 0x and consists of 40 hexadecimal characters, for example 0x742d35Cc6634C0532925a3b844Bc454e4438f44e. Tron addresses start with T; Bitcoin addresses start with 1, 3 or bc1.
• Digital signature: when you send money, your wallet signs the transaction with your private key. Anyone can use your public key to verify that the signature came from you, yet no one learns your private key.

An analogy: an address is like a glass-walled mailbox in the main square. Everyone can see inside and anyone can drop something in, but only the person holding the key can take anything out. The signature, in turn, is a seal that anyone can verify but no one can copy.

A wallet is software or a device that stores the keys and signs transactions. There are no coins in it. The coins are entries on the chain; the wallet only holds the keys. There are several kinds of wallets: hot wallets (online), cold wallets (offline hardware), MPC wallets (the key is split among several parties) and smart contract wallets (with recovery options).

Self-custody and custodial wallets.

• Self-custody (unhosted wallet): you hold the keys. As the crypto saying goes: "not your keys, not your coins." There is no intermediary, so by default there is no customer identification either. But if you lose the key, the money is gone forever, and no one will reverse a transfer sent to the wrong address.
• Custodial wallet: an exchange or custodian holds the keys for you, and you have a claim against it, just like with a bank account. It is more convenient and regulated, but comes with counterparty risk, as the collapse of the FTX crypto exchange in November 2022 showed.

5.6 Network fees ("gas")

Every transaction requires paying a network fee to the validators. On Ethereum this is called gas. Every operation consumes a certain number of "gas units", and the fee is gas units × gas price (paid in ether). When the network is congested, the fee spikes, because users bid for space in the blocks.

Some typical values from 2026. On Tron a USDT transfer often stays under 1 dollar. On Ethereum, fees fell sharply after the 2025 capacity upgrades; in January 2026 the average transaction fee was about 0.15 dollars. A token transfer is somewhat more expensive than a plain ether send, and during congestion the fee can creep up to several dollars. On layer 2 networks (see below) and on Solana the fee is typically measured in cents, or even less.

The network fee is fixed per transaction and does not depend on the amount. Sending a million dollars on-chain costs the same as sending ten, so for large amounts the blockchain leg is cheap. But across the whole route there are also percentage-based costs on the way in and out (trading fees, spread, slippage), and exchange withdrawal fees can be percentage-based too (Chapter 8).

5.7 Smart contracts and tokens

A smart contract is a program stored on the blockchain that runs exactly as written when someone calls it. Its state (token balances, for example) is part of the ledger. "Smart" is a marketing word, and "contract" here means code, not necessarily a legal contract.

A token is an entry managed by a smart contract, as opposed to the chain's own coin (ether, for example). On Ethereum, USDC is not a "coin flying around". It is a table inside the USDC contract, with addresses and balances. A transfer consists of the contract rewriting two rows.

On Ethereum, ERC-20 is the most widespread token standard (for interchangeable, fungible tokens; USDC, USDT and EURC are examples). There are others, such as ERC-721 (unique tokens, NFTs) and ERC-3643 (permissioned security tokens with built-in identity). On Tron the equivalent is TRC-20, on Solana SPL.

Stablecoin contracts typically have admin functions built in: mint (create new tokens), burn (destroy them), pause (suspend), blacklist/freeze (freeze). So a centrally controlled token runs on a decentralized chain, and the issuer can freeze addresses (Chapter 14).

Let's see what happens inside the contract during a transfer. If A wants to send 100 USDC to B, A's wallet signs a transaction that calls the USDC contract's transfer function ("send 100 to B's address"). The network's validators run the code. The contract checks the signature and whether A's balance is at least 100, then subtracts 100 from A's row and adds 100 to B's row. If the funds aren't there, it rejects the transaction and nothing changes. The mint, burn and freeze functions can only be called by the issuer, from its own privileged address.

5.8 Finality

Finality is the point beyond which a transaction can no longer be reversed.

• Bitcoin (PoW): probabilistic finality. Every new block lowers the chance of a reversal; the usual benchmark is 6 blocks, about 60 minutes.
• Ethereum (PoS): a block arrives about every 12 seconds, and economic finality sets in after about 13–15 minutes. For small amounts most applications consider one or two blocks enough.
• Solana and many permissioned chains: a fraction of a second or a few seconds.

Legal finality, however, is a different matter. No public chain has been designated under the Settlement Finality Directive, so in a bank's eyes on-chain finality is not legally equivalent to the finality reached in VIBER or T2. That is the main reason central banks want blockchain transactions settled in central bank money (Chapter 10, Pontes).

5.9 Block explorers and transparency

Block explorers are websites (Etherscan, Tronscan, Solscan) that display the public ledger: every address's balance and every transaction (hash, sender, recipient, amount, fee, time). This works in both directions. Everything can be checked and traced, which helps the fight against money laundering (specialized analytics firms work on this: Chainalysis, Elliptic, TRM Labs). But if a company's address is known, its competitors can see its payments and balance too.

5.10 Layers, bridges, oracles

• Layer 1 (L1): the base blockchain, with its own consensus (Bitcoin, Ethereum, Solana, Tron, Arc).
• Layer 2 (L2): a system built on top of an L1 that processes transactions off the main chain and writes a compressed proof back to it. Examples are Base (a Coinbase initiative), Arbitrum, Optimism, zkSync and Linea. It is cheaper and faster, and largely inherits the L1's security (with some caveats).
• Bridge: moves tokens between chains. Typically the token is locked on chain A, and a "wrapped" token is issued in its place on chain B. The locked pool is a huge target: bridges are the most frequently hacked thing in the crypto world. A few cases from 2022: Ronin (about 620 million dollars), Wormhole (about 320 million), Nomad (about 190 million). According to Chainalysis, about 2 billion dollars was stolen from bridges in 2022. For stablecoins it is safer if the issuer mints natively on every chain or runs its own "burn and re-mint" protocol (such as Circle's CCTP), because then no wrapped token is needed.
• Oracle: brings off-chain data (exchange rates, proof of reserves) into smart contracts. If the oracle is wrong, the contract works with bad data.

5.11 Why "crypto"?

"Crypto" comes from cryptography. The system's security rests on two pillars: (1) digital signatures, which prove ownership without revealing the key, and (2) hash functions, which make the ledger tamper-proof. "Cryptocurrency" originally meant a money-like asset whose ownership and transfer are protected by cryptography rather than by an institution. Since then, everyday language has come to use "crypto" for the whole asset class and industry. The name does not mean the asset is secret or illegal. On a blockchain every transaction is explicitly public.

6.1 Bitcoin and ether

Bitcoin (BTC). The first cryptocurrency. Its design was published in October 2008 by an author using the pseudonym “Satoshi Nakamoto”, and the network went live in January 2009. Supply is capped at 21 million, and it runs on proof of work. It has no issuer, and it is not a claim on anyone. It is worth exactly what buyers are willing to pay for it. Today it is mostly held as a speculative store of value (“digital gold”).

Ether (ETH). The native coin of the Ethereum network (2015). It pays the network fee and is also used as collateral. Ethereum is today the most important platform for smart contracts and stablecoins.

Beyond these two there are thousands of other tokens (Solana’s SOL, Tron’s TRX and so on).

6.2 The volatility problem

Historically, bitcoin’s price has been many times more volatile than stocks or the major currencies. It has crashed 50–80% more than once (in 2018, in 2022). On October 10, 2025, more than $19 billion of leveraged positions were liquidated across the crypto market in a single day.

Why does this rule out payments? A means of payment has to be a stable unit of account. You cannot settle a €10,000 invoice with something that is worth only €9,000 by the time it arrives. No corporate treasurer will hold something that can lose 10% of its value in a day.

So the crypto world needed an asset that keeps the convenience of a blockchain token but has a value fixed against traditional money. That became the stablecoin. Crypto traders asked for it first: they wanted a stable “dollar” inside the exchanges, where they could park money between two bitcoin trades without wiring it back to the bank every time. Most crypto exchange trading is still quoted in stablecoins today (BTC/USDT, for example).

7.1 Definition

A stablecoin is a crypto-asset (a token on a blockchain) whose job is to hold a stable value against some reference asset. That is almost always a traditional currency (USD, EUR), sometimes gold or a basket of currencies. It tries to combine two properties: the mobility of a blockchain token (24/7, global, programmable, directly between two parties) and the price stability of traditional money.

The legal definition differs by jurisdiction:

• EU (MiCA): an e-money token (EMT) if it references a single official currency, and an asset-referenced token (ART) if it references another value or a combination of values (a basket, gold, several currencies).
• USA (GENIUS Act): a “payment stablecoin”. A digital asset used for payment and settlement that the issuer is obliged to redeem for a fixed amount of money.

7.2 Types

The fall of TerraUSD (May 2022). On May 7, 2022, UST lost its peg to the dollar and crashed to about $0.04. UST (about $18.7 billion) and its linked LUNA together wiped out more than $40 billion of value in a single week and set off a crypto credit crisis (Celsius, Three Arrows Capital, and eventually FTX). Regulators responded: in the EU and the USA an algorithmic “stablecoin” cannot get regulated stablecoin status. It was not explicitly banned, but it cannot be an EMT or an ART, so the protections that apply to stablecoins do not cover it. Only fully reserve-backed tokens can get that status.

Ethena USDe had grown to about $14.7 billion by October 2025. During the crash of October 10, 2025, a fault in an internal price feed on the Binance exchange briefly quoted it at $0.65; about $2 billion of it was redeemed within 24 hours, and by the end of 2025 it had shrunk to about $6.4 billion. In August 2026 its supply was about $4–4.5 billion.

7.3 Minting and redemption, step by step

Let’s follow the life of a fiat-backed stablecoin, USDC.

Primary market. Only vetted institutional clients (exchanges, fintechs, large companies) can mint and redeem tokens directly with the issuer. At Circle, this is what “Circle Mint” is for.

1. Onboarding. The client (say, a crypto exchange) goes through the issuer’s KYC/KYB checks.
2. Deposit. The client wires $10 million to the issuer’s bank account. (Bank ledgers change.)
3. Mint. The issuer’s smart contract creates 10 million new USDC at the client’s blockchain address. (The blockchain ledger changes; total supply grows by 10 million.) The issuer invests the $10 million in its reserve: short-term government securities, repo, bank deposits.
4. Circulation. The tokens move freely between addresses (exchanges, wallets, companies). This is the secondary market. Most users get their stablecoins here, by buying on an exchange, not from the issuer.
5. Redeem. A vetted client sends 10 million USDC back to the issuer.
6. Burn. The issuer destroys the tokens. (Total supply shrinks by 10 million.)
7. Payout. The issuer wires $10 million from the reserve to the client’s bank account.

How does the price stay at $1? Arbitrage keeps it there. If USDC trades at $0.99 on the exchanges, a vetted client buys it up cheaply and redeems it with the issuer for $1.00. The buying pushes the price back to $1. If it trades at $1.01, they mint at $1.00 and sell. The peg is only as strong as redemption is credible and fast.

In the EU, MiCA states that every EMT holder has the right to redeem their token with the issuer at any time, at face value and free of charge. By law, this right belongs to every holder. In practice retail users exercise it through an intermediary (a crypto exchange), and the issuer’s KYC is mandatory here too.

7.4 Reserves, attestations and audits

What is in the reserve? Typically bank deposits, short-term US government securities (Treasury bills, T-bills), overnight reverse repurchase agreements backed by government securities (reverse repo), and government money market funds. At Tether, also gold and bitcoin.

The rules:

• USA (GENIUS Act): the reserve may consist of cash, bank deposits, government securities with a maturity of up to 93 days, repo backed by government securities and government money market funds, at least 1:1, and the composition must be published monthly.
• EU (MiCA, EMT): at least 30% of the funds received must always be held in segregated accounts at credit institutions, the rest in secure, low-risk, highly liquid assets in the same currency. For electronic money institutions issuing a significant EMT, the ratio is 60% (under the EBA’s technical standard).

A real example: Tether’s report for Q2 2026 (as of June 30, 2026, attested by the accounting firm BDO):

• assets: $187.75 billion;
• liabilities (tokens in circulation): $183.64 billion;
• excess reserve: $4.11 billion (roughly half of the $8.23 billion in Q1);
• of which $114.96 billion in short-term US government securities, $18.84 billion in gold, $5.80 billion in bitcoin;
• quarterly profit of about $1.5 billion. Tether has hired KPMG for its first full audit; that is still in progress.

Attestation and audit. They are not the same thing.

• An attestation only checks whether the declared reserve existed at a given point in time and whether it exceeded the tokens in circulation. It is a limited examination.
• An audit reviews the financial statements, internal controls and processes of an entire period, and ends with a formal opinion. Stablecoin issuers used to publish only attestations. The GENIUS Act requires issuers above $50 billion to produce audited financial statements every year.

7.5 How does the issuer make money?

The model is simple. The issuer receives dollars, pays no interest to the token holder (both MiCA and the GENIUS Act prohibit it), and invests the reserve in interest-bearing government securities. In 2025 short-term US rates were around 4%. On a supply of about $185 billion, Tether earned about $1.5 billion in a single quarter. That makes issuers’ revenue extremely sensitive to the level of interest rates.

Circle shares a large part of its revenue with its distribution partners. According to figures published at its 2025 stock market listing, Coinbase receives 50% of the residual revenue from the USDC reserve, plus all of the revenue on USDC held on Coinbase’s own platform.

The ban on interest and rewards: how do the EU and the USA differ? In the EU, MiCA prohibits paying interest both for the issuer and for crypto-asset service providers (exchanges, for example). The ban covers any remuneration or benefit tied to the length of time the token is held, even if a third party pays it. So EU exchanges cannot offer the kind of “USDC rewards” American ones do. In the USA the GENIUS Act’s ban binds only the issuer, and exchange “rewards” became one of the cornerstones of the fight over the CLARITY Act (Chapter 11).

7.6 Depeg

We speak of a depeg when a stablecoin’s market price drifts away from its target ($1).

USDC and Silicon Valley Bank, March 2023. On Friday, March 10, 2023, California’s Silicon Valley Bank (SVB) failed. Circle announced that $3.3 billion (about 8%) of the USDC reserve was stuck at SVB. On Saturday, March 11, USDC fell to $0.87 on decentralized exchanges. On Monday, March 13, it climbed back to $1 after US authorities guaranteed all SVB deposits. Meanwhile USDC’s supply shrank from $43.4 billion to $36.9 billion.

There are two lessons. A reserve held in bank deposits carries bank risk. And redemption stops at the weekend because the banking systems stop too, while the token keeps moving on the chain. That is why MiCA requires the reserve to be spread across several banks, and why the United Kingdom regulates the share held as central bank deposits.

There have been other cases. On May 12, 2022 (during the Terra crisis) USDT fell to about $0.95 and recovered within a day. During the March 2020 crash DAI traded at $1.11, that is, above its peg. On April 2, 2025, FDUSD fell to about $0.87 after allegations of insolvency.

7.7 Market size and dollar dominance

Size (October 2026):

• Total stablecoin supply is about $309 billion according to Standard Chartered’s October 14 report, and it has moved between $300 and $320 billion over the past 12 months. Press reports citing DefiLlama data showed $302–306 billion on October 16–17. Growth has stalled since the end of 2025.
• For comparison: in January 2024 it was about $134 billion.
• USDT: about $183.5 billion (about 60%), USDC: about $72.6 billion (about 24%) (October 16, 2026). In early August they were still at $186.9 billion and $75.6 billion respectively.
• By chain (October 16, 2026): Ethereum about $144.5 billion (47%), Tron about $95.1 billion (31%), Solana about $16 billion (5%). USDT dominates on Tron (emerging markets, retail transfers); USDC dominates on Ethereum, Solana and Base.

Supply is flat, but the money moves faster and faster. On-chain stablecoin transfers averaged about $10 trillion a month in 2026 (versus $4 trillion in 2025), and set a record of $19 trillion in May 2026. On average a USDC changes hands about 80 times a month, a USDT about 7 times. Many people hold USDT as dollar savings; USDC is used more for settlement.

These raw on-chain volume figures, however, are mostly trading, exchanges’ internal movements and automated programs. According to McKinsey, real economic payments (for goods and services) came to about $390 billion annualized in December 2025. About 60% of that is business-to-business (B2B) payments, and the whole thing is about 0.02% of global payment volume.

Dollar dominance. About 99% of stablecoins are denominated in dollars. The combined supply of euro stablecoins was about $674 million on June 28, 2026 (up 128% in a year), about 64% of it EURC. Other euro stablecoins: EURCV (Société Générale-Forge), EURI (Banking Circle), EURe (Monerium). Qivalis, a European bank consortium (37 banks from 15 countries, including ING, UniCredit, BNP Paribas, KBC and RBI), received an electronic money institution license from the Dutch central bank on October 14, 2026. It plans to start issuing “in the coming months”, after its white paper is approved, so it has not launched yet. There is no Hungarian bank among the named members.

The EU’s first “significant” e-money token. On October 14, 2026, the European Banking Authority (EBA) classified USDC as the EU’s first significant e-money token (out of 45 notified EMTs). From November 12, 2026, the EBA supervises the issuer’s (Circle Internet Financial Europe SAS) own funds, reserve, liquidity, and recovery and redemption plans. The French supervisor (ACPR) remains responsible for governance, qualifying holdings, the fitness of management, and outsourcing. So a single issuer now answers to two supervisors.

7.8 Is a stablecoin crypto? Three answers

The answer depends on whether you look at it legally, technically or economically.

1. Legally (EU): yes. Under MiCA, a crypto-asset is any digital representation of a value or a right that can be transferred and stored electronically using distributed ledger or similar technology. A stablecoin is one, in the EMT or ART category. But an EMT is also legally e-money: crypto in form, e-money in substance. (In the USA, under the GENIUS Act, a payment stablecoin issued by a licensed issuer is neither a security nor a commodity, but a category of its own.)

2. Technically: yes. It is a token running on a blockchain, protected by the same cryptography (keys, signatures, hashes), held in the same wallets, moved by the same networks, and visible on the same block explorers.

3. Economically: mostly not. In economic terms it is a tokenized IOU from a private issuer, backed by bank deposits and short-term government securities. It is much closer to a money market fund share or an e-money balance than to bitcoin. Its price cannot rise on its own; its value comes from the dollar behind it and the issuer’s promise to redeem.

In one sentence: the packaging is crypto, the content is traditional money.

8.1 The example and the measurement

Anna lives in Budapest and has a forint account. She wants to send the equivalent of $200 to Bruno in São Paulo, who has a bank account in Brazilian reais. They use USDC.

The Banca d'Italia (the Italian central bank) measured transfers built exactly like this in a “mystery shopping” study. The paper is titled Are Stablecoins Efficient for Remittances? Evidence from a Mystery Shopping Exercise, by Di Iorio, Di Stefano, Mascioli and Trebeschi. It appeared as No. 86 of the Mercati, infrastrutture, sistemi di pagamento series on July 30, 2026. The tests were run on March 24 and 26, 2026, each time with 200 USDC, across 10 corridors: between Italy and Argentina, Brazil, South Africa, the United Arab Emirates and Japan, in both directions.

8.2 The five phases

8.3 Why so many conversions?

Because the blockchain is a separate ledger, independent of the banking system.

The money has to cross a system boundary twice. First it moves from the bank ledger into an intermediary’s books (1), then it turns into a token the chain can carry (2), crosses the chain (3), turns back into local money (4), and finally returns to the local bank ledger (5). Every ledger boundary and every foreign exchange step is a separate step.

In the SWIFT world the same conversions exist, just inside the banks, hidden: the sending bank converts the currency, and the correspondent banking entries run in the books. The stablecoin route makes these steps visible and splits them across several providers. That is why it looks more complicated, although the traditional route takes the same steps; we just don’t see them.

8.4 Where does the cost arise? The evidence

The Banca d'Italia’s findings:

• Total cost ranged from 0.30% to 8.96% of the amount sent.
• The on-chain leg averaged about 0.4% (including the exchange withdrawal fee), with a low of 0.01% (Brazil to Italy). Most of the cost and delay arises at the on-ramp and off-ramp.
• The extreme values are distorted. The 0.30% on the Italy→Argentina route only looked cheap because in Argentina USDC traded about 3% above the official exchange rate, and the 8.96% in the opposite direction comes partly from the same distortion. The 8.95% for the Emirates→Italy route is because only card deposits were available there, at a 3.8% markup.
• Compared with World Bank averages, the stablecoin route proved cheaper from Brazil (2.21% versus 9.96%) and from South Africa (5.44% versus 15.23%), but more expensive from the Emirates (8.95% versus 2.65%).
• Compared with Wise, USDC was cheaper on 3 corridors and more expensive on 4. On Italy→Brazil, for example, Wise cost 2.20% and USDC 2.70%.
• Speed: where both ends have an instant payment system (TIPS in Italy, PIX in Brazil, Transferencias 3.0 in Argentina), the money arrived within 20 minutes. In South Africa, where there is no instant leg, it took 1–2 business days.
• The authors’ conclusion: the stablecoin route has no systematic cost advantage. More could be saved if the payee could spend the stablecoin directly (without an off-ramp). But that raises the question of monetary sovereignty, the danger that dollar tokens crowd out local money.

One more note on the network fee. It does not depend on the amount, so for large amounts the on-chain leg becomes negligible. But the full route carries percentage costs at the entry and exit points (trading fee, spread, slippage). And the 0.4% the Banca d'Italia measured also includes the exchange withdrawal fee, not just the network fee.

8.5 When it is cheaper, and when it is not

The stablecoin route is typically better when:

• the corridor is poorly served by banks (few correspondent banks, “de-risked” countries, capital controls with a parallel market);
• the payee wants to stay in dollars (no off-ramp; an Argentine or Nigerian supplier, for example, hedges against inflation this way);
• the payment has to happen outside banking hours (weekends, holidays, across time zones);
• both parties already hold stablecoins (crypto firms, exchanges, market makers).

For large amounts the network fee is negligible. Large transfers, however, usually go through over-the-counter dealers (OTC desks) with an individually negotiated spread, and come with stricter compliance checks.

It is typically worse when:

• both ends have a cheap domestic instant payment system and a competitive money transfer operator (between EUR, GBP, USD);
• you can only pay in by card (1.5–4% fee);
• the local off-ramp market is thin (wide spread), or KYC and withdrawal limits get in the way;
• it is a small amount with a fixed withdrawal fee;
• there are regulatory obstacles. In Hungary, for example, a separate “validation” regime made crypto conversion harder until August 2026 (Chapter 11), and Japan strictly limits retail use of dollar stablecoins.

8.6 Speed: what is instant, and what is not?

• On-chain: seconds (Tron, Solana, L2s), about 13–15 minutes to full finality on Ethereum, day and night, every day of the year.
• What causes the rest of the delay: the bank deposit (cut-off times if there is no instant payment), exchange KYC and compliance holds (Travel Rule checks, risk scoring), withdrawal queues, weekend processing at the local bank, and the weekend FX market (wider spreads).

The middle is instant; the edges are only as fast as the local banking system. The token moves 24/7 in minutes, but getting money in and out depends on banks, exchanges and checks, and that can take minutes, or even days.

8.7 The “stablecoin sandwich”

The industry calls corporate stablecoin payments the “stablecoin sandwich”: traditional money → stablecoin → traditional money. A payment provider handles both conversions, and the client never even touches the token.

[ HUF in the bank ] --on-ramp (provider A)--> [ USDC on-chain ] --off-ramp (provider B)--> [ BRL in the bank ]
       bread                                       filling                                        bread

In the middle of the sandwich, the blockchain replaces the correspondent banking chain and the prefunded nostro accounts. It only pays off if the cost of the on-ramp and off-ramp, including the two foreign exchange steps, is lower than the correspondent banking cost. Between the major currencies (EUR, USD, GBP), where Wise-type competitors exist, that is often not the case. On more exotic corridors (for example dollars to Nigeria, Argentina, Turkey, Vietnam or the Philippines), where correspondent banks are few and local currency is scarce, it often is. The legal rules of the destination country must always be checked separately, though (we go into the Turkish example in Chapter 13).

If the payee keeps the stablecoin and does not convert it back to local money, the sandwich stays “open”. The second slice of bread disappears, and with it the off-ramp’s fee, spread and delay. According to the Banca d'Italia, this is where the savings would be largest, but it raises the question of monetary sovereignty (sections 8.4 and 14.10).

8.8 The intermediaries: on-ramps and off-ramps, exchanges, market makers

On-ramp: any service that converts traditional money into a crypto-asset (a stablecoin, for example). It is the “highway on-ramp”. Examples: buying USDC on an exchange with a bank transfer, buying by card in a MoonPay or Ramp widget inside a wallet, or using a bitcoin ATM.

Off-ramp: the reverse: a crypto-asset becomes traditional money, in a bank account, on a card or in cash.

On-ramps and off-ramps are where the regulated financial system meets the blockchain. This is where KYC and anti-money-laundering checks happen, where a license and a banking relationship are needed, and where most of the costs arise too. In MiCA’s language, on- and off-ramping is the crypto-asset service called “exchange of crypto-assets for funds”.

How does a centralized exchange swap, and how does a decentralized one? On a centralized exchange the client first deposits money or tokens, and from then on the exchange holds them. Trading happens in the exchange’s internal ledger: an order book matches buy and sell orders, and the exchange rewrites the two clients’ balances. Nothing happens on the blockchain at that point; there is an on-chain transaction only on deposit and withdrawal. A decentralized exchange has neither an order book nor a custodian. A smart contract holds pools of two tokens (USDC and ETH, say), filled by liquidity providers. The user sends one token from their own wallet, and the contract returns the other. The price is set by the ratio of the two pools: the more ETH is taken out, the more expensive the next one. This is called an automated market maker (AMM).

Liquidity, spread, slippage.

• Liquidity: how easily an asset can be bought and sold in large size without moving the price. Deep liquidity = tight spread.
• Spread: the difference between the best buy and sell prices. On a liquid USDC/USD market it can be as low as 0.01%; on a peer-to-peer USDT/Nigerian naira market, several percent.
• Slippage: a large order “chews through” the order book and fills at progressively worse prices.

The cost of the off-ramp depends mainly on how liquid the local market is between USDC/USDT and the local currency. That explains why the Banca d'Italia’s results varied so much from corridor to corridor.

9.1 The two routes side by side

Let’s stick with a single payment: an SME in Budapest pays €10,000 to a foreign supplier in a country outside SEPA.

SWIFT route: SME → Hungarian bank (conversion, fee) → correspondent bank (nostro debit, fee) → possibly RTGS → the correspondent bank of the supplier’s bank (fee) → the supplier’s bank (conversion, fee) → supplier.

Stablecoin route: SME → Hungarian bank (AFR/SEPA Instant) → EU crypto-asset service provider (on-ramp, fee, spread) → blockchain (network fee) → crypto-asset service provider in the destination country (off-ramp, fee, spread) → local bank (local instant payment) → supplier.

What disappears: the chain of correspondent banks, the nostro/vostro accounts, prefunding, the cut-off times of the middle leg, and fee deductions along the way.

What takes its place: two crypto-asset service providers, two conversions (into the token and back), a blockchain, an issuer with its own risk, the Travel Rule data exchange, screening by blockchain analytics firms, and the question of what the destination country’s crypto regulation allows.

9.2 Comparison table

9.3 How much of a competitor is it?

Its volume is still small today, but it is already forcing the incumbents to respond.

A few reference points:

• Real stablecoin payments come to about $390 billion a year (December 2025), about 0.02% of global payment volume (McKinsey).
• FXC Intelligence estimates the total addressable market for cross-border payments at about $208 trillion in 2025, mostly high-value interbank flows. Total stablecoin supply (about $0.3 trillion) is roughly one seven-hundredth of that. The comparison is only indicative, since it sets a stock against an annual flow.
• Behind SWIFT’s tens of millions of messages a day stands interbank system volume that is orders of magnitude larger (Fedwire alone settled $4.59 trillion a day, T2 €1.93 trillion a day in 2025).

We also have to be careful what we compare with what. SWIFT is a messaging network; a stablecoin is a means of payment. The real competition is between settlement assets: privately issued stablecoins, banks’ tokenized deposits and central bank money. SWIFT could even carry the messages for stablecoin transactions.

Stablecoins still put the system under heavy pressure, for two reasons:

1. they already dominate crypto market settlement, and they are growing fast in corporate payments on poorly served corridors (according to McKinsey, B2B stablecoin payments grew roughly eightfold, by 733%, in a year);
2. their technology (shared ledger, tokenized money, 24/7 operation) is being adopted by SWIFT and the banks too.

9.4 Where do they complement each other?

The Banca d'Italia concluded that stablecoins and instant payment systems are more complementary than competing. The stablecoin route is fastest precisely where domestic instant payments work at both ends. For 2026–2030 the realistic expectation is coexistence and gradual convergence:

• SWIFT becomes the connecting layer between banks’ tokenized deposits;
• stablecoins stay strong in retail flows in emerging markets, in the crypto world, in fintechs’ corporate payments and in machine-to-machine micropayments;
• in high-value settlement, central bank money moving on a blockchain (Pontes) and tokenized deposits come to the fore.

9.5 How are banks and card networks responding?

• Tokenized deposits. Examples are JPMorgan’s Kinexys platform and the JPMD dollar deposit token running on Base (November 2025), Citi Token Services and the HSBC Tokenized Deposit Service. A tokenized deposit is a liability of the bank, just like a traditional deposit (Chapter 10).
• SWIFT’s blockchain ledger. A pilot with 17 banks has been running on it since July 2026, and the first live transaction (HSBC and Standard Chartered) went through in August 2026 (section 3.9).
• Banks’ own stablecoins. Société Générale-Forge (EURCV, USDCV), Banking Circle (EURI), and Qivalis, which brings together 37 banks (license: October 14, 2026).
• Stablecoin settlement by card networks. Visa’s stablecoin settlement passed an annualized run rate of $20 billion in September 2026, and in December 2025 it opened USDC settlement to US banks. Mastercard offers USDC and EURC settlement to acquiring banks, and bought BVNK in August 2026.
• Money transfer operators. Western Union’s USDPT (May 2026) and MoneyGram’s USDC app (2025).

So the two islands are growing together at the institutional level too.

10.1 What does it mean to tokenize an asset?

In tokenization, a right attached to an asset (money, a bond, a fund unit, real estate, a receivable, a commodity) is represented as a token on a programmable shared ledger. From then on it can be held, transferred and used in smart contracts just like a stablecoin.

There are two models:

• Native digital issuance: the asset is issued on the ledger from the start (for example, a digital bond under German or Luxembourg law).
• Wrapping / representation: a custodian holds the traditional asset, and the tokens represent it (most tokenized government bonds and gold tokens work this way).

Legally, the token has to be tied to enforceable rights (securities law, registry law). Otherwise it is nothing more than a database entry.

Here is what the wrapping model looks like for a tokenized fund. (1) The issuer, say an asset manager, sets up the fund, which buys government bonds. (2) The asset is held by a custodian or by a separate legal entity created for this purpose (an SPV). (3) A tokenization platform (Securitize, in the case of BlackRock BUIDL) deploys a smart contract on the chain and mints as many tokens as the fund units investors have bought. The contract only allows tokens to be sent to identified addresses on a whitelist. (4) The token shows up in the investors’ wallets. (5) The token can be passed on between whitelisted addresses, and every transfer also updates the register. (6) On redemption the token is destroyed and the investor gets paid out of the fund. If the token embodies a financial instrument, capital markets rules apply to every step (section 10.7).

RWA (real-world assets) is the crypto industry’s term for tokens that represent an off-chain asset (government bonds, private credit, equities, commodities, real estate). Technically, stablecoins make up the largest RWA class, but they are usually counted separately.

10.2 Tokenized money market funds and government bonds

According to rwa.xyz, on October 17, 2026 tokenized US Treasury funds were worth about USD 14.7 billion. The largest include Circle USYC (2.39 billion), Ondo USDY (2.36 billion), BlackRock BUIDL (2.17 billion; launched on Ethereum in March 2024, Securitize is its registrar, it now runs on 9 chains and is accepted as collateral in several places), and Franklin Templeton’s BENJI/iBENJI funds (about 2.4 billion combined). The Franklin Templeton fund was the first fund registered with the US securities regulator to run on a public chain.

The total market for tokenized real-world assets (excluding stablecoins) is somewhere around USD 30–40 billion, depending on the definition (mid-2026). These numbers are methodologically messy; every source counts differently.

What do tokenized funds have to do with payments? They serve as interest-bearing collateral and reserve assets. Stablecoin issuers hold government bonds, and money market funds available on-chain offer a “parking spot” for stablecoins that are not being used at the moment.

10.3 Why do institutions care?

1. Faster settlement. Securities trades today settle one day after the trade in the US (T+1, since May 2024) and two days after in the EU (T+2; T+1 from October 11, 2027). On a shared ledger the asset and the money can change hands at the same moment, with no settlement risk. This is delivery versus payment (DvP).
2. Mobile collateral: a tokenized government bond can be moved as collateral from one venue to another 24/7, in minutes.
3. Programmability: coupon payments, corporate actions, or who is allowed to hold the asset can all be written into code.
4. Less reconciliation between separate ledgers (custodian, depository, registrar, bank).

This leads to the “cash leg” problem. Tokenized assets need a tokenized settlement asset. If a government bond token is on the chain, the purchase price has to be there too. Three candidates are competing for this role.

10.4 Stablecoin, tokenized deposit, central bank digital currency

According to the BIS Annual Economic Report of June 2025, stablecoins fail three tests: singleness (money must always be exchangeable one-for-one), elasticity (the money supply must be able to grow with the needs of the economy) and integrity (protection against money laundering). The BIS says the future system should be built on tokenized central bank reserves, tokenized deposits and tokenized government bonds, on a unified ledger. That is the central banks’ position; the stablecoin industry disputes it.

10.5 Central bank digital currency and the digital euro

Central bank digital currency (CBDC) is issued directly by the central bank. It comes in two kinds: retail, which anyone can hold, and wholesale, which only banks and market infrastructures use.

Where does the digital euro stand (October 2026)?

• On October 30, 2025 the ECB Governing Council moved the project into its next phase: a pilot is coming in 2027, and first issuance in 2029 at the earliest, provided the EU legislation is adopted in 2026.
• The Council of the EU adopted its position on December 19, 2025, and the European Parliament on July 9, 2026 (416 in favor, 169 against, 22 abstentions). The first three-way negotiation (trilogue) took place on July 13, 2026, and talks have been ongoing since.
• The holding limit is still an open question. A limit of about €3,000 has been floated, but nothing has been decided. The point of the limit is to stop people from moving their bank deposits into digital euros en masse.
• As planned, the basic service will be free, will also work offline, and will pay no interest.

Other examples. By the end of November 2025, China’s e-CNY had processed a cumulative total of more than 3.4 billion transactions and about 16.7 trillion yuan (about 2.3 trillion dollars). mBridge is a platform shared by several central banks (China, Hong Kong, Thailand, the United Arab Emirates, Saudi Arabia). The BIS left it in October 2024, officially because the project had “matured”. In the background were disputes over sanctions and moving away from the dollar.

Project Agorá. A joint project of the BIS, eight central banks (among them the Banque de France on behalf of the Eurosystem, the Bank of England, the Bank of Japan and the New York Fed) and more than 40 private banks. They want to build a “unified ledger” that links tokenized wholesale central bank money with tokenized commercial bank deposits to enable atomic cross-border settlement. According to the BIS update of May 27, 2026, the prototype is complete and a real-value test has been run: 28 institutions, about CHF 800,000 in six currencies, 30 transactions, about 80 seconds on average from initiation to settlement. No decision has been made on going live.

10.6 Pontes: central bank money on the blockchain

On July 1, 2025 the ECB adopted a two-step plan for settling blockchain transactions in central bank money:

• Pontes (Latin for “bridges”) is the short-term solution that connects blockchain platforms with T2.
• Appia is the long-term, integrated ecosystem; its blueprint is expected for 2028.

Pontes went live on September 21, 2026. The first participants include Deutsche Bank, Santander, Société Générale, the European Investment Bank, KfW, and four blockchain operators (Axiology, Cashlink, Clearstream, SWIAT). Sources differ slightly on the number of launch participants (13 or 18, depending on how you count). Full rollout is expected by 2028. The ECB announced that it will invest part of its own funds in tokenized euro public sector debt, settled through Pontes. It is not investing in crypto-assets or stablecoins.

With Pontes, in wholesale tokenized markets the EU is offering central bank money as the “cash leg” instead of private stablecoins. This also solves the problem of legal finality (section 5.8).

10.7 The EU DLT Pilot Regime

The DLT Pilot Regime (Regulation 2022/858, in force since March 23, 2023) is a regulatory “sandbox”. It lets market infrastructures trade tokenized financial instruments (shares, bonds, fund units) and settle them on a blockchain, with exemptions from certain EU capital markets rules. Its license types: DLT MTF (trading), DLT SS (settlement), DLT TSS (both). Its limits: shares of issuers under €500 million, bonds under €1 billion, €6 billion in total per infrastructure. Interest so far has been modest.

The European Commission’s package of December 4, 2025 proposes that all financial instruments be eligible, that the €6 billion ceiling be raised to €100 billion, and that the time limit be dropped. For now this is only a proposal.

A separate case: a tokenized security is not a MiCA crypto-asset. It falls under the Markets in Financial Instruments Directive (MiFID II), because it is a “financial instrument” regardless of the technology. MiCA only regulates crypto-assets that are not financial instruments, deposits and so on.

11.1 MiCA: what does the name mean?

MiCA stands for Markets in Crypto-Assets Regulation, that is, the EU regulation on crypto-asset markets, number (EU) 2023/1114. It is a regulation, not a directive, so it applies directly, with the same text, in all 27 member states.

• The rules on stablecoins (EMT, ART) have applied since June 30, 2024.
• Everything else (crypto service providers, other crypto-assets, market abuse) since December 30, 2024.
• Transitional period: firms already operating under national rules could keep working without a MiCA license until July 1, 2026 at the latest. Member states could shorten this. Hungary (together with the Netherlands, Finland, Latvia and Slovenia) chose the shortest transition, 6 months, which ended on June 30, 2025. France, Malta and Luxembourg granted the full 18 months.
• Since July 1, 2026, crypto services in the EU may only be provided by a MiCA-licensed crypto service provider, or by an already licensed financial institution (such as a bank) that has notified its supervisor.

11.2 The three categories of crypto-assets

The white paper is the mandatory disclosure document for a crypto-asset, a relative of the securities prospectus.

A common mistake is to think that only euro stablecoins are EMTs. In fact every stablecoin that refers to a single official currency is an EMT, the dollar included. USDC is an EMT in the EU, and it is the one the EBA classified as significant.

What falls outside MiCA: financial instruments (MiFID II covers those), deposits (including tokenized deposits), e-money not in token form, unique NFTs, fully decentralized services with no human intermediary, and central bank digital currency.

11.3 Significant tokens and the EBA

An EMT or ART is significant if it meets at least three of seven criteria. Examples: more than 10 million holders, a value or reserve above €5 billion, more than 2.5 million transactions and over €500 million in turnover per day, the issuer being a gatekeeper platform, significant international use, or interconnectedness with the financial system. Issuers of significant tokens face a stricter capital requirement (3% of the reserve) and liquidity rules, and the EBA takes over part of the supervision.

On October 14, 2026, USDC became the first significant EMT (section 7.7). The EBA did not say which criteria were met.

11.4 A cap on stablecoins in non-EU currencies

Article 23 of MiCA (for ARTs, and through Article 58(4) also for EMTs denominated in a non-EU currency) provides the following. If a token’s estimated quarterly average daily use “as a means of exchange within a single currency area” exceeds 1 million transactions and €200 million, the issuer must stop issuing and submit a plan within 40 working days. The EU explicitly wants to prevent dollar stablecoins from pushing the euro out of everyday payments.

11.5 CASP: the crypto service provider

A CASP (crypto-asset service provider) is a legal entity that provides one or more crypto services under a MiCA license. The ten services:

Minimum capital: Class 1 €50,000, Class 2 €125,000, Class 3 €150,000, or a quarter of the previous year’s fixed overheads, if that is higher. On top of that you need an EU registered office and at least one director resident in the EU, fit and proper management, a governance system, complaints handling, segregation of client assets, IT security (under DORA, the Digital Operational Resilience Act), an anti-money laundering program and compliance with the Travel Rule.

Banks and other financial institutions (credit institutions, central securities depositories, investment firms, electronic money institutions for custody and transfer of their own EMTs, as well as fund managers and market operators) can provide crypto services by notification (Article 60), without a new license.

How many are there? According to ESMA’s interim register (processed by casptracker.eu), on October 17, 2026 about 375 licensed CASPs were operating in the EEA. Most are in Germany (106), followed by France (36), the Netherlands (29), Cyprus (25) and Malta (23). Hungary shows 0 in the register, even though the MNB granted a license to CoinCash on July 20, 2026. The register update is probably lagging. Hungarian customers can be served by 177 CASPs arriving from other member states with a “passport”.

Passporting. A CASP licensed in one member state can also provide services in the others if it notifies its own supervisor. One license thus opens the whole EU market.

Reverse solicitation. A non-EU firm may only serve an EU customer if the customer approached it exclusively on their own initiative. This is interpreted very narrowly; any marketing breaks it.

The future: under the Commission’s proposal of December 4, 2025, supervision of CASPs would move to ESMA (except for banks and small players). This is still only a proposal.

11.6 The ban on interest and rewards

Under Article 50 of MiCA, neither the issuer nor a crypto service provider may pay interest on an EMT. Interest means any remuneration or benefit linked to the length of time the token is held, even if a third party provides it. So in the EU this is not a gray zone (section 7.5).

11.7 Where USDT stands in the EU

Tether did not apply for a MiCA license, because in its view the reserve rules clash with its business model. Two steps followed:

1. January 17, 2025: according to a statement by ESMA and the Commission, offering non-compliant ARTs and EMTs and admitting them to trading had to stop by the end of Q1 2025. EU exchanges then delisted their USDT trading pairs (Coinbase Europe as early as December 2024, Crypto.com in January 2025).
2. October 8, 2026: according to an ESMA opinion, crypto service providers must stop all services (custody, transfer, advice, portfolio management) linked to stablecoins that are not MiCA-compliant, must prevent new purchases, and must wind down remaining positions (by selling, converting, withdrawing or transferring out) by January 8, 2027 at the latest. ESMA did not name a token, but USDT and PYUSD are the main ones affected.

That still does not mean MiCA “banned USDT”. MiCA regulates firms, not private individuals, so it does not stop anyone from holding USDT in their own wallet. But regulated EU service providers are exiting it step by step.

11.8 Other EU rules around MiCA

TFR, the “Travel Rule”. The Transfer of Funds Regulation ((EU) 2023/1113) on the information accompanying transfers has applied since December 30, 2024. Crypto service providers must send and receive the sender’s and the beneficiary’s data with every crypto transfer, with no minimum amount (bank transfers have a €1,000 threshold). If more than €1,000 goes to or comes from a self-custody wallet, the provider must verify that the wallet really belongs to its customer. The name refers to the data “traveling” with the money. The regulation transposes Recommendation 16 of the Financial Action Task Force (FATF).

The AML package and AMLA. The new Anti-Money Laundering Regulation (AMLR, (EU) 2024/1624) applies from July 10, 2027, and prohibits crypto service providers from keeping anonymous accounts and wallets. AMLA (the Anti-Money Laundering Authority, the EU’s anti-money laundering authority, based in Frankfurt) started operating in 2025, and from 2028 it will directly supervise about 40 high-risk, cross-border financial entities (banks, possibly large crypto firms).

DAC8. The eighth amendment to the Directive on Administrative Cooperation ((EU) 2023/2226): crypto tax reporting. It applies from January 1, 2026. Crypto service providers report their customers’ crypto transactions to the tax authority (purchases, sales, transfers, aggregate values, identification data), including stablecoins and domestic transactions. The first data exchange, covering 2026 data, will take place by September 30, 2027. This is the EU implementation of the OECD’s Crypto-Asset Reporting Framework (CARF). In Hungary the data goes to NAV, the national tax authority.

DORA. The Digital Operational Resilience Act (2022/2554, since January 17, 2025) requires IT risk management from crypto service providers and EMT issuers too.

11.9 Payment licenses: PSD2 and EMD2

Alongside MiCA, the older EU payment rules decide who may handle client money.

• Payment institution (PI). Under the Payment Services Directive (PSD2, 2015/2366) it can execute payments, issue payment instruments, accept card payments and do money remittance. It cannot take deposits or issue e-money, and it must keep client money segregated. Initial capital: €20,000 (money remittance only), €50,000 (payment initiation only), €125,000 (other services). In Hungary, under the payment services act, it is HUF 37.5 million in the general case, HUF 6 million for money remittance only, and HUF 15 million for payment initiation only.
• Electronic money institution (EMI). Under the E-Money Directive (EMD2, 2009/110/EC) it can issue e-money and provide payment services. It cannot lend out client money or pay interest on it. Initial capital is €350,000, and ongoing own funds must reach 2% of the average outstanding e-money issued. (Sources contradict each other on the forint amount in the Hungarian act; the MNB’s guidance mentions €350,000.)
• Credit institution (bank). Under the Capital Requirements Directive and Regulation (CRD/CRR) it can take deposits and grant loans. The EU minimum initial capital is €5 million, and in Hungary you need at least 4 billion forints of starting capital to found a bank (300 million forints for a cooperative credit institution). Significant banks in the euro area are supervised by the ECB, the rest by the national supervisor, in Hungary the MNB.
• PSD3 and PSR. Political agreement on the new Payment Services Directive and Regulation was reached on November 27, 2025, and the final texts were published on April 23, 2026. They are expected to apply from around 2028. The E-Money Directive will be repealed, and electronic money institutions will become a subcategory of payment institutions.

Why does a stablecoin issuer have to be a bank or an EMI? Because MiCA treats an EMT as e-money, and in the EU only these two types of institution can issue e-money. That is why Circle has a French EMI license, and why Qivalis applied for a Dutch EMI license.

A new rule since March 2, 2026. According to an EBA opinion (EBA/OP/2026/01), executing transfers of EMTs is a payment service, whether custodial or not. So if a CASP transfers stablecoins on behalf of its customer, it needs, in addition to its MiCA license, a payment institution license or a payments partner. The transitional “tolerance” period ended on March 2, 2026. Since then, CASPs without such a license have had to stop EMT payment services (those with a pending application could continue under strict conditions, without new customers).

11.10 United States

GENIUS Act (Guiding and Establishing National Innovation for U.S. Stablecoins Act), signed on July 18, 2025. Its main elements:

• only a permitted payment stablecoin issuer may issue tokens to US persons. That can be a subsidiary of an insured bank, a federal issuer approved by the bank regulator under the Treasury (the Office of the Comptroller of the Currency, OCC), which may even be a non-bank, or a state issuer, if it has no more than 10 billion dollars’ worth of tokens outstanding and the state regime is “substantially similar”;
• 1:1 reserves in permitted assets, monthly disclosure, and above 50 billion dollars an audited annual financial statement;
• the issuer may not pay interest or yield;
• a permitted issuer’s payment stablecoin is neither a security nor a commodity;
• if the issuer goes bankrupt, token holders get access to the reserve ahead of all other claims, in an expedited court procedure;
• the issuer must be technically able to freeze and burn tokens on a lawful order from the authorities, and anti-money laundering obligations apply to it in full;
• a foreign issuer’s token (Tether’s, for example) may only be offered by US platforms if the Treasury recognizes the home country’s regime as comparable, the issuer is registered with the OCC, holds sufficient reserves in the US and is able to comply with US lawful orders. Three years after enactment (around July 18, 2028), US service providers may no longer offer payment stablecoins from non-permitted issuers to US persons.
• Effective date: the earlier of two dates: 18 months after enactment (around January 18, 2027), or 120 days after final implementing rules are issued.

Implementation (October 2026): the regulators missed the statutory deadline of July 18, 2026. On October 15, 2026 the head of the OCC said the office would issue its final rule by the end of October, and that the Fed, the FDIC and the Treasury were also working on their parts.

Tether’s US token. On January 27, 2026 Tether launched USAT, a GENIUS-aligned stablecoin issued by Anchorage Digital Bank, which is supervised by the OCC. The first attestation (January 31, 2026) showed about 17.6 million dollars in reserves, so for now it is a small item.

OCC national trust bank charters. On December 12, 2025 the OCC conditionally approved two new national trust banks (Circle’s and Ripple’s), as well as the conversion of BitGo, Fidelity Digital Assets and Paxos (according to BitGo itself, its approval was unconditional). Circle’s bank (First National Digital Currency Bank, N.A.) received final approval on July 10, 2026. Trust banks can do custody, settlement and payments, but they do not take deposits and have no deposit insurance.

CLARITY Act (on market structure, that is, which US regulator supervises crypto-assets). The House passed it on July 17, 2025 (294–134), and the Senate Banking Committee advanced it on May 14, 2026 (15–9). On September 15, 2026, however, the cloture vote to open debate failed in the Senate, 49–50 (60 votes would have been needed). Formally the bill is not dead, but in 2026 it is effectively stalled. One of the sticking points was the stablecoin “rewards” paid by exchanges.

Other: a money transmitter license is needed in almost every state, and in New York a separate BitLicense on top. The EBA and the New York State Department of Financial Services (NYDFS) signed a cooperation agreement on the supervision of cross-border stablecoins on June 2, 2026.

11.11 Other major hubs in brief

• United Kingdom: the Financial Conduct Authority (FCA) published the final rules of its crypto regime, including those on stablecoin issuance, on June 30, 2026. The Bank of England finalized its rules for systemic stablecoins on June 22, 2026: during a transitional period, an issuance cap of 40 billion pounds per stablecoin, at least 30% of the backing in non-interest-bearing central bank deposits, and at most 70% in short-term government bonds. Holding limits were dropped. Regulated stablecoins can operate from 2027.
• United Arab Emirates: under the central bank’s June 2024 regulation, only licensed dirham tokens may be used for domestic payments. Foreign tokens (USDT, USDC) only from a registered issuer, and only for buying virtual assets. In Dubai VARA licenses crypto service providers, in Abu Dhabi’s financial free zone the ADGM FSRA.
• Singapore: the central bank (MAS) finalized its framework for single-currency stablecoins in August 2023 (full reserves held in segregated trust, redemption within 5 business days, capital requirements, disclosure; only licensees may use the “MAS-regulated stablecoin” label). The draft bill was published on September 1, 2026, and the consultation closed on October 16.
• Hong Kong: the Stablecoins Ordinance has been in force since August 1, 2025. The first two issuer licenses went to HSBC and Anchorpoint (a joint venture of Standard Chartered’s Hong Kong unit, HKT and Animoca) on April 10, 2026.

11.12 Hungary

Supervision. The competent authority under MiCA is the MNB (Magyar Nemzeti Bank, the central bank of Hungary). It licenses crypto service providers and ART/EMT issuers, and it supervises whether crypto service providers comply with anti-money laundering rules. This was laid down in Act VII of 2024 (the Hungarian MiCA implementation act). Hungary chose the shortest transitional period, so unlicensed providers had to stop operating by June 30, 2025. The MNB’s new recommendations for crypto service providers apply from October 1, 2026.

The 2025 validation regime. In 2025, Act LXVII of 2025 introduced a Hungarian rule that exists nowhere else:

• every crypto→fiat and crypto→crypto exchange would have had to be checked by a licensed validator. Validators were licensed by the Supervisory Authority for Regulated Activities (SZTFH), and the detailed rules appeared in an SZTFH decree of October 27, 2025;
• two new criminal offenses were added to the Criminal Code, “misuse of crypto-assets” and “providing unauthorized crypto-asset exchange services”, and these could even affect users.

The exact timing: the two criminal offenses took effect on July 1, 2025. The validation obligation, however, only had to be applied in practice 60 days after the SZTFH decree, around December 27, 2025. The first validator (Caduceus Zrt.) was appointed on December 19, 2025; the second was Mikroháló Kft.

The effects: Revolut (used by about 74% of Hungarian crypto users in 2025) froze its crypto services in early July 2025 because of the criminal law uncertainty. MoonPay and others also pulled out, CoinCash suspended operations in December 2025, and Bitpanda built in validation and stayed. According to a PwC survey (June 2026), the number of active users fell by about 38% from roughly 210,000 before. At the end of January 2026 the European Commission opened infringement proceedings against Hungary (INFR(2025)2174).

The repeal. After the April 2026 election, the new government announced a U-turn. On July 28, 2026 Parliament passed Act XXXVIII of 2026 (143 in favor, 46 against, 1 abstention), which took effect on August 7, 2026. The act abolished the validation obligation and the two criminal offenses, and stripped the SZTFH of its power to license validators. The SZTFH had to close pending proceedings by August 12, 2026, and must delete the exchange data it collected.

After the repeal, Lightyear launched its crypto service in Hungary on August 7, 2026, eToro came back, Interactive Brokers opened crypto to Hungarian customers in September 2026, and Bitpanda kept operating throughout. As of fall 2026 Revolut has not restarted its Hungarian crypto service, even though it holds a Cypriot MiCA license. Binance cannot enter (it has no MiCA license).

CoinCash: the first Hungarian-based MiCA crypto service provider. CoinCash (Tiwala Solutions Kft.; founded in 2017, started with bitcoin ATMs, bought MrCoin in 2021) applied for a license in March 2025 and on July 20, 2026 received its MiCA license from the MNB for custody, crypto↔fiat and crypto↔crypto exchange, transfers, advice and portfolio management. From September 2026 it gradually restarted its services, beginning with exchange. As of mid-October 2026 it was not yet in ESMA’s register.

Crypto tax. Since January 1, 2022, income from crypto-assets has been separately taxed income: 15% personal income tax is due on the annual net gain, but no social contribution tax.

• Only exchanging into money, or into a non-crypto product or service, is taxable; crypto→crypto swaps are not.
• Loss offsetting: the tax on a given year’s crypto income can be reduced by 15% of the transaction losses declared in that tax year and the two preceding years, provided the loss was reported in the return for the year it arose.
• Small transaction exemption: no tax is due if (a) the revenue from the transaction (not the gain!) does not exceed 10% of the monthly minimum wage, (b) there was no other such revenue-generating transaction that day, and (c) the annual total of such revenues does not exceed the monthly minimum wage.
• Because of DAC8, from the 2026 tax year crypto service providers report Hungarian customers’ transactions to NAV.

Summary box: Hungary in October 2026

• The validation regime is gone (August 7, 2026).
• The MNB is the MiCA authority. CoinCash is the first Hungarian-based crypto service provider licensed by the MNB (July 2026).
• Hungarian customers can be served by any EU CASP with a passport (177 of them as of October 2026).
• Tax: 15% personal income tax on the annual net gain, with no social contribution tax. DAC8 reporting starts with 2026 data.
• Hungarian payment service providers must be able to receive instant payments on euro accounts from January 9, 2027, and to send them from July 9, 2027.
• The financial transaction levy (0.45%, max HUF 20,000; 0.9% on currency exchange) also applies to forints transferred to a crypto exchange.

12.1 Why a ladder?

Financial regulation rests on a simple principle: the more damage you can do with other people’s money, the more responsibility, capital and oversight you have to take on.

• If you only write software or provide information, you never touch other people’s money, so you don’t need a license either.
• If you hold client money or client keys, you can lose them, have them stolen, or embezzle them yourself. So you need a license, capital, segregation and supervision.
• If you issue money (e-money, a stablecoin), whether that money is worth anything depends on your promise. You are expected to hold reserves and to be obliged to redeem.
• If you lend out of deposits (a bank), you create money, and if you fail, you put the whole economy at risk. That is why the strictest capital and liquidity rules apply to you.

At every rung of the ladder, risk, responsibility and capital rise together. The biggest jump is where someone first touches other people’s money or keys (between rungs 2 and 3).

A warning: the statutory minimum capital is only the floor. In practice, the cost-based capital requirement (for example, a quarter of fixed overheads) and compliance staff (anti-money laundering officer, risk manager, internal auditor, data protection officer) cost a lot more. The time and cost estimates below are [practical estimate], not official figures; they only give a sense of the order of magnitude.

12.2 The rungs of the ladder

Rung 0: user, self-custody

• What they can do: hold their own crypto-assets and stablecoins and send them from their own wallet; buy from licensed crypto service providers.
• What they can’t: provide services to others for a fee.
• License, supervision: none, but there is a tax obligation (NAV, 15% personal income tax).
• Capital: none.
• Stepping up: you need a company and a business model that serves others.

Rung 1: software, information site, affiliate, educator, analytics firm, non-custodial wallet developer

• What they can do: build software, publish information, refer customers to licensed firms for a commission, sell software services (SaaS) to crypto service providers, and do blockchain analytics.
• What they can’t: hold client money or keys, execute transactions, give personalized investment advice, or advertise unlicensed firms to EU customers.
• License: not needed under MiCA (consumer protection, advertising and data protection rules do apply). Watch the boundaries carefully, because “receiving and transmitting orders” and “advice” are already CASP services. Referral and comparison sites (“aggregators”, such as Onramper) can typically operate without their own CASP license because they never touch the customer’s money and route customers to licensed providers.
• Time, cost [practical estimate]: weeks; legal review €5–20k.
• Stepping up: contracts with licensed partners, compliance know-how, and a decision about which regulated activity comes next.
• Examples: MetaMask, Ledger (software/hardware), Chainalysis, Elliptic, TRM Labs (analytics), Notabene, Sumsub (Travel Rule and identity verification), Fireblocks (technology).

Rung 2: agent, distributor, white-label partner

• What they can do: acquire and onboard customers under the principal’s license, and run a branded front end.
• What they can’t: operate outside the principal’s license, or hold money in their own name.
• License: registered by the principal’s supervisor (PSD2 agent register; in Hungary the MNB). MiCA has no separate agent regime; there, outsourcing or a white-label setup under the CASP’s license is the way to go. Liability lies with the principal.
• Capital: none (the principal’s is what counts).
• Time [practical estimate]: 1–3 months, including the principal’s due diligence.
• Stepping up: your own governance system, anti-money laundering officer, capital, IT security; in other words, your own license.
• Examples: Western Union and MoneyGram agent locations (shops, post offices), white-label crypto apps.

Rung 3: CASP, crypto service provider

• What they can do: depends on the class. Class 1 (€50,000): advice, order transmission, execution, placing, portfolio management, transfers. Class 2 (€125,000): on top of that, custody, crypto↔fiat and crypto↔crypto exchange (on- and off-ramp). Class 3 (€150,000): on top of that, a trading platform. With a passport, they can serve the whole EU.
• What they can’t: issue e-money or EMTs, or take deposits. They typically handle the customer’s traditional money through a bank or e-money institution partner. Since March 2, 2026, executing EMT transfers also requires a payment license or a payments partner.
• License, supervision: MiCA license; in Hungary issued by the MNB and recorded in ESMA’s register. (Under a proposal, ESMA would supervise in the future.)
• Capital: €50k / €125k / €150k, or a quarter of fixed overheads, if that is higher.
• Time, cost [practical estimate]: 6–12+ months; total cost €100–500k (lawyers, compliance staff, IT, audits).
• Stepping up: a payment or e-money license to handle traditional money, more capital, segregation systems.
• Examples: Coinbase (Luxembourg supervisor), Kraken (Irish), Bitvavo, MoonPay, Banxa (Dutch), Ramp (Irish), Bitpanda (German), Revolut Digital Assets Europe (Cypriot), CoinCash (MNB).

Rung 4: payment institution or e-money institution

• What they can do: a payment institution executes payments, sends money and accepts cards. An e-money institution can also issue e-money (and therefore EMTs), and keeps client money segregated. With a passport, they can operate across the whole EU.
• What they can’t: lend out client money, take deposits, or pay interest on e-money.
• License, supervision: PSD2 / EMD2 license; in Hungary the MNB. (From around 2028, PSD3 folds the EMI into payment institutions.)
• Capital: payment institution €20,000 / €50,000 / €125,000 (in Hungary HUF 6 / 15 / 37.5 million); e-money institution €350,000 + 2% of e-money outstanding.
• Time, cost [practical estimate]: 9–18 months; €0.5–2 million.
• Stepping up to stablecoin issuance: an EMT white paper under MiCA, reserve management, working redemption.
• Examples: payment institutions: Wise Europe SA (Belgian), MoneyGram International SA (Belgian), Remitly Europe (Irish). E-money institutions: Circle Internet Financial Europe (French), Stripe Payments Europe (Irish), Qivalis (Dutch), AllUnity (German), Quantoz (Dutch), BVNK (Maltese), Zerohash Europe (Dutch).

Rung 5: stablecoin issuer

• What they can do: mint and burn EMTs (as an e-money institution or a bank), invest the reserve (at least 30% in bank deposits), and live off the yield on the reserve. As an ART issuer, it can also issue basket-based tokens.
• What they can’t: pay interest; an algorithmic design cannot become a regulated stablecoin; and for a token in a non-EU currency it cannot exceed the Article 23 limits (1 million transactions and €200 million per day as a means of exchange).
• License, supervision: MiCA Titles III and IV + an EMI or banking license; the national supervisor (ACPR, DNB, MNB, etc.), and for a significant token the EBA too.
• Capital: EMI capital + 2% own funds (3% of the reserve for a significant token); for an ART, the highest of €350,000, 2% of the reserve, and a quarter of fixed overheads.
• Time, cost [practical estimate]: 12–24 months; several million euros.
• Stepping up: size decides. If the token becomes significant, EBA supervision follows, along with 60% bank deposits in the reserve and recovery plans.
• Examples: Circle (USDC, EURC), Société Générale-Forge (EURCV, USDCV), Paxos (USDG, Finnish EMI), Banking Circle (EURI, bank). Outside the EU: Tether (El Salvador), Anchorage Digital Bank (USAT, USDPT, US).

Rung 6: bank (credit institution)

• What they can do: take deposits, grant loans (and thereby create money), issue tokenized deposits and EMTs, provide crypto services by notification, and hold a central bank account (with access to the RTGS).
• What they can’t: within prudential limits (capital, liquidity and risk-taking), practically anything goes.
• License, supervision: CRD/CRR banking license. In the euro area, significant banks are licensed and supervised by the ECB (the Single Supervisory Mechanism, SSM); in Hungary by the MNB.
• Capital: EU minimum €5 million, in Hungary HUF 4 billion of starting capital (HUF 300 million for a cooperative credit institution), plus the Basel capital adequacy ratios and limits on crypto exposure.
• Time, cost [practical estimate]: 2–3+ years; €10 million+.
• Stepping up: systemic importance, direct RTGS and CLS membership, a correspondent network.
• Examples: OTP, JPMorgan, HSBC, Citi, Deutsche Bank, Banking Circle, Revolut Bank UAB (Lithuanian) and Revolut Bank S.A. (French, August 10, 2026). US: national trust banks (Anchorage, BitGo, Circle National Trust).

Rung 7: systemic infrastructure and central bank

• What they can do: operate the financial “rails”, set standards, settle in central bank money. Only the central bank can issue central bank money and central bank digital currency.
• Supervision: central bank oversight (SWIFT by the G10 central banks led by the National Bank of Belgium, CLS by a cooperative arrangement led by the Fed, card schemes by the ECB); central banks themselves are governed by statute and the EU treaties.
• Capital, time: not applicable; decades of work and the network effect.
• Examples: T2, TIPS, EBA CLEARING, VIBER, GIRO (AFR, ICS), CLS, SWIFT, Fedwire, MNB, ECB.

12.3 The rule of thumb for stepping up

Every step up increases five things:

1. whose hands the customer’s money is in along the way (nobody’s → the partner’s → yours);
2. capital;
3. governance and compliance staff (anti-money laundering officer, risk management, internal audit, data protection officer);
4. regulatory reporting;
5. the time it takes to get licensed.

The biggest gap is between not touching other people’s money or keys (rungs 0–2) and actually holding them (rung 3 and up).

The US ladder (simplified): software (no license) → registration with the Financial Crimes Enforcement Network (FinCEN) + state money transmitter licenses → New York BitLicense / state trust company → OCC national trust bank → full-service bank → Federal Reserve.

13.1 How to read the map

The map has two axes.

• Layer: where the player sits in the payment chain, from the shared infrastructure at the very bottom up to the app in the user's hand.
• License: what it is legally allowed to do, and who supervises it (following the ladder in Chapter 12).

A corporate group often holds several licenses through several legal entities. Circle, for example, has a French electronic money institution, a US national trust bank, New York licenses, and licenses in Bermuda and Abu Dhabi too. The rule of thumb: the closer a player gets to holding customer money or issuing money, the harder the license is to get.

13.2 Layer by layer

Central banks and shared infrastructure (rung 7)

Regulators and standard setters (they don't move money)

Messaging and card networks (rungs 7/6)

Banks (rung 6)

Money transfer operators and fintechs (rung 4, sometimes a bank)

Stablecoin issuers (rung 5)

Blockchains (the token's "rails")

Blockchains themselves are not licensed. The issuers, crypto service providers and custodians that use them are the ones that need a license.

Exchanges, crypto service providers (rung 3). As of October 17, 2026, about 375 MiCA-authorized providers were operating in the EEA.

On-ramps and off-ramps, aggregators (rung 3 or 1)

Corporate stablecoin payment infrastructure (rungs 3/4)

Custodians and wallets (rungs 0–3, depending on who holds the key)

The decisive line runs at the key. If the company holds the key, it is custody (MiCA class 2 or a bank/trust charter); if the user holds it, it is software (rungs 0–1).

Tokenization platforms

Compliance and analytics players (rung 1)

Chainalysis, Elliptic, TRM Labs (blockchain analytics, transaction monitoring), Notabene, Sumsub, Global Travel Rule (Travel Rule data exchange, identification). They hold no financial license; the responsibility stays with the licensed crypto service provider and the bank.

Besides the layers, the license is the other axis. The matrix below lines up the rungs of the ladder from Chapter 12: what EU license each level needs, who supervises it, what the minimum capital is, and who the typical players in this course are.

13.3 Who connects to whom

Now we can populate the two islands from Chapter 0 with the players.

On the left island, at the very bottom, stand the central banks and the RTGS systems (T2, TIPS, Fedwire, VIBER). Only banks have accounts here (plus, through TIPS, a few payment and electronic money institutions). Above them come the clearing houses (EBA CLEARING, GIRO) and CLS. Banks are connected by SWIFT for messages and by nostro/vostro accounts for money. Higher up are the payment and electronic money institutions and money transfer operators (Wise, Western Union, Revolut, PayPal), which keep segregated customer funds at banks, and the card networks, which settle through banks. At the outer edge work the agents.

On the right island, at the bottom, are the public chains (Ethereum, Tron, Solana, Base) and the new payment chains (Arc, Tempo). Stablecoins and tokenized funds run on these. Around them sit the wallets and custodians, the exchanges and crypto service providers, and around everyone, the analytics firms watching.

The bridges (now in more detail than in Chapter 0):

1. Issuer reserves: behind every fiat-backed stablecoin stand commercial bank deposits, government securities and repos, and minting and redemption run through the issuer's bank accounts. Because of MiCA, a large part of EMT reserves sits at EU banks.
2. On- and off-ramps: MoonPay, Ramp, Banxa and the exchanges' cash desks connect card networks, SEPA and TIPS to the blockchain. The money side needs a bank or payment institution partner.
3. Corporate stablecoin providers: BVNK, Bridge, Zerohash and Circle Payments Network take in the money on local rails, carry it on-chain, and pay it out on local rails.
4. Card networks: Visa and Mastercard also settle with banks in stablecoins, and with a crypto card the stablecoin can be spent anywhere cards are accepted.
5. Bank tokenized deposits and shared ledgers: JPMD, Kinexys, Citi Token Services, HSBC, the SWIFT ledger, Agorá.
6. Central bank money on blockchain: Pontes, later Appia, plus the digital euro (retail, 2029 at the earliest).
7. Hybrid money transfer operators: Western Union (USDPT + cash payout) and MoneyGram (USDC + agent network) could become the stablecoin's "last, cash mile."
8. Banks as crypto service providers and issuers: EU banks with a MiCA notification (BBVA, Commerzbank, German cooperative banks), SG-Forge, Qivalis and Banking Circle; and in the US, crypto companies as trust banks.

13.4 Three journeys

(a) A Hungarian SME pays a Turkish supplier over SWIFT

A Budapest furniture-making SME pays 10,000 euros to an upholstery fabric supplier in Istanbul.

1. Order. The SME submits the transfer in online banking at OTP (a credit institution licensed by the MNB).
2. Foreign exchange. If the SME pays from a forint account, OTP converts the forints to euros on its own books, with its own spread. This conversion does not go through CLS, which is for interbank trades. On top comes the 0.9% financial transaction levy on the conversion (max. 2 × HUF 20,000).
3. Message. OTP sends a pacs.008 (ISO 20022) message over SWIFT, with gpi tracking. (Its old name was MT103.) No money passes through SWIFT.
4. Correspondent bank. If OTP has no direct account at the Turkish bank, the transfer goes through a euro correspondent bank (typically a large EU bank). There OTP's nostro account is debited, and the correspondent bank credits the Turkish bank, or its correspondent, on its own books.
5. Settlement. If both banks involved hold accounts at the same correspondent bank, it is a simple internal book transfer. If not, settlement can happen in T2, where euro area banks hold accounts. Hungary is not in the euro area, so OTP gets euro liquidity through its correspondent banks or its euro area units.
6. Credit. The Turkish bank (for example Garanti BBVA or İşbank, licensed by the Turkish banking supervisor, BDDK) credits the amount to the supplier, possibly converted to lira.
7. Supervisors involved: the MNB (OTP), the euro correspondent bank's supervisor (ECB/national), the National Bank of Belgium and the G10 (SWIFT), the BDDK and the Turkish central bank (TCMB). Anti-money laundering and sanctions screening runs at every bank.

Typically: it arrives the same day or within 1–2 business days (often same day with gpi), with several fees (sending bank, correspondent bank, receiving bank), an FX spread and cut-off times.

(b) The same with a stablecoin, and why it collides with Turkish law

First, a legal warning. Under the Turkish central bank's regulation of April 16, 2021 (in force since April 30, 2021), in Turkey crypto-assets may not be used for payments, directly or indirectly, and payment and electronic money institutions may not transfer money to or from crypto platforms. If a Turkish supplier sells goods and receives the price in USDC or EURC, that is exactly the indirect payment use the ban targets. On the Turkish side, then, this route is legally problematic. The description below shows the mechanism; it is not a recommendation. The lesson holds in general: licenses and bans have to be checked at both ends. The same route to a country where accepting stablecoins in commerce is allowed can be legally fine.

The mechanism:

1. Deposit. From its OTP account, the SME sends forints or euros by AFR or SEPA Instant to an account in its own name at a MiCA-authorized crypto service provider (for example Kraken, Bitvavo, Coinbase or CoinCash). The crypto service provider holds customer money at its partner bank or electronic money institution. The forint transfer is subject to the financial transaction levy.
2. Customer due diligence and conversion. The crypto service provider performs corporate due diligence (KYB) and anti-money laundering checks, then converts the euros to EURC or USDC (both MiCA-compliant EMTs, issued by the French Circle Internet Financial Europe SAS; from November 12, 2026, USDC comes under joint EBA–ACPR supervision). USDT would not be offered to an EU customer, because it is not MiCA-authorized. The crypto service provider supplies the token from its own inventory or (as an institutional client) has Circle mint it.
3. On-chain transfer. The crypto service provider sends the token on-chain (for example on Ethereum, Solana or Base). Since March 2, 2026, this requires a payment institution license or a payment partner in addition to the MiCA license. The Travel Rule data (sender and beneficiary) reaches the receiving provider (through Notabene, Sumsub or similar), and the addresses are screened by analytics firms (Chainalysis, Elliptic, TRM).
4. Receipt. On the Turkish side, a crypto service provider (BtcTurk, Paribu and others operate under the transitional regime of the Turkish capital markets board, the SPK; according to the SPK, being on the list does not mean a license) or a self-custody wallet would receive the token. This is the point where the transaction collides with the Turkish payment ban.
5. Off-ramp. The supplier would convert the token to lira and withdraw it to its Turkish bank account (over the FAST instant system).
6. In the background the fiat reserve does not move. It stays at Circle's reserve banks and money market funds; only the token moves on-chain.
7. Corporate version: instead of a crypto exchange, the SME uses a stablecoin payment provider (for example BVNK or Bridge), which handles the on- and off-ramp in the background (the "stablecoin sandwich").

Typically: the middle leg completes in minutes, 24/7. The cost is the trading fees and the FX spread at both ends, plus a network fee of a few cents. Risks: legal bans in the destination country (see Turkey), accounting and tax treatment, Travel Rule compatibility and issuer risk.

(c) Remittance home: money transfer operator or stablecoin app

A migrant worker in Hungary sends the equivalent of 200 dollars home to their family every month.

(c1) With a traditional money transfer operator

• Western Union: the worker pays in cash at an agent (a shop, a post office; rung 2), or pays in the app by card or bank transfer. Western Union (in the EU, an Austrian bank) works from pre-funded money in the destination country and its own internal ledger. The family picks up the amount in cash at a Western Union agent, or receives it to a bank account or mobile wallet. Western Union uses SWIFT only for large treasury settlements.
• Wise: the worker pays into Wise's local account (for example by AFR or SEPA to Wise Europe SA, a Belgian payment institution). Wise nets the flows and pays out in the destination country from its own local account, on the domestic rails (in India, for example, over UPI; in the Philippines, over InstaPay). An individual transfer needs no correspondent banking chain.
• Cost benchmark: the average cost of digital remittances is about 4.6–4.85%, of non-digital ones about 7.2–7.3% (IFAD for 2025 and World Bank 2025 Q1, respectively).

(c2) With a stablecoin app

• The worker uses a wallet or an app. In Hungary, Revolut's crypto service is still suspended in fall 2026, so they would turn to another EU crypto service provider operating via passporting (for example Bitpanda) or to CoinCash, or perhaps top up a self-custody wallet with a MoonPay or Ramp widget. In the EU they buy USDC (a MiCA EMT), but on corridors outside the EU many people use USDT on Tron.
• The stablecoin arrives in the family's wallet in seconds, for a fee of a few cents.
• The family cashes it out: on a local exchange or a peer-to-peer market, at a MoneyGram or Western Union agent that accepts stablecoins (MoneyGram–Stellar–USDC, Western Union–USDPT), or spends it with a Visa or Mastercard card topped up with stablecoins.
• Players involved: wallet/app (rung 1 or 3), on- and off-ramp provider (rung 3), issuer (rung 5: Circle, Tether, Anchorage), blockchain (Tron, Solana, Stellar), cash payout network (rungs 2–4), analytics and Travel Rule providers.
• The balance: the on-chain step is cheap and fast, but the cost migrates to the edges (on- and off-ramp fees, FX spread, cash payout). The family has to manage a key or trust an app, and the destination country's rules may differ too.

14.1 Issuer risk

A stablecoin is a private company's liability. If the reserve is lost, badly invested or frozen, the token loses value. Tether's excess reserve melted to 4.11 billion dollars in the second quarter of 2026, half of what it was, and part of the reserve is gold and bitcoin, i.e. volatile assets. On top of that, its reports so far have only been attestations; no full audit has been done yet.

14.2 Depeg and run

If the market price drops below 1 dollar and holders all want to redeem at once while the reserve is not liquid enough, a run breaks out, just like a bank run. Two examples: during the SVB collapse (March 2023), USDC fell to 0.87 dollars, and the collapse of UST (May 2022) wiped out more than 40 billion dollars. A separate risk is the failure of the bank holding the reserve: 3.3 billion dollars of USDC reserves got stuck at SVB.

A run feeds itself. A piece of news (say, a reserve bank failing) shakes confidence, and many people want to redeem at once. To pay them, the issuer has to sell the reserve, turning Treasuries and repos into cash. If the reserve is liquid and holds its value, the payout happens and the loop stops. But if part of it gets stuck (as at SVB), or can only be sold at a loss, the token's market price falls below 1 dollar, and that triggers more redemptions. That is why regulation puts brakes on several points of the loop. Under MiCA, at least 30% of an EMT's reserve (60% for a significant token) must be bank deposits, spread across several banks, and redemption is due at any time, at par, free of charge. The GENIUS Act requires Treasuries of at most 93 days and monthly disclosure, and the Bank of England expects systemic stablecoins to hold at least 30% in central bank deposits. Fire sales by the big issuers can also feed back into the Treasury market (section 14.11).

14.3 Hacks, smart contract and exchange risk

• Smart contracts and bridges: code bugs and their exploitation. In 2022, about 2 billion dollars were stolen from bridges alone.
• Exchanges and custodians: bankruptcy, fraud or hacks. FTX collapsed in November 2022. In February 2025, about 1.5 billion dollars' worth of ether was stolen from the Bybit exchange; the attack was attributed to North Korea's Lazarus Group (this is a widely reported figure; our sources did not re-verify it).
• Operational and oracle risk: faulty price feeds, halted chains, congestion. On October 10, 2025, USDe traded at 0.65 dollars on one exchange because of a fault in that exchange's internal price feed.

14.4 Lost keys and user error

Lost seed phrase = lost money. A transfer sent to the wrong address or the wrong chain cannot be undone. Without a custodian, there is no "forgot password" button.

14.5 Sanctions, money laundering, illicit finance

Stablecoins are an attractive tool for money laundering, sanctions evasion, fraud and ransomware. According to Chainalysis, at least 154 billion dollars flowed to addresses linked to illicit activity in 2025, and 84% of illicit volume was in stablecoins. At the same time, that is less than 1% of all on-chain volume. A7A5, a ruble-pegged stablecoin, moved more than 93 billion dollars in its first year. The US Treasury's sanctions office (OFAC) sanctioned it on August 14, 2025, and the EU on October 23, 2025. On March 3, 2026, the FATF issued a dedicated report on stablecoins and unhosted wallets.

14.6 Freezing cuts both ways

Centralized issuers can blacklist addresses. That is handy for law enforcement, but for users it means censorship and counterparty risk. By March 2026, Tether had frozen about 4.2 billion dollars in total, blacklisted 7,268 addresses, and worked with more than 310 authorities in 62 jurisdictions. The GENIUS Act explicitly requires the issuer to be technically capable of freezing.

14.7 Privacy

On a public chain, every transaction of an address is visible. If a company's address is known, it can leak whom it pays and how much, suppliers and employees alike. Meanwhile, because of the Travel Rule, regulators mandate collecting the sender's and beneficiary's data. The two requirements pull against each other.

14.8 Fragmentation

There are dozens of chains and tokens: USDC runs on more than 20 chains, there are "wrapped" versions, and many stablecoins compete with each other. Liquidity gets fragmented, and the singleness of money suffers (one of the BIS's three tests).

14.9 Regulatory arbitrage

Issuers and providers pick the most lenient jurisdiction, then serve customers from there via "passporting" or globally. Offshore exchanges serve EU users. This includes the disputes around reverse solicitation, or Binance withdrawing its Greek MiCA application (June 24, 2026).

14.10 Monetary sovereignty

If a country uses dollar stablecoins on a mass scale, they can crowd out the local currency, weaken the central bank's interest rate policy, and drain bank deposits ("deposit substitution"). This is the risk of dollarization.

• The ECB's concerns: on May 8, 2026, ECB President Christine Lagarde said that even for euro stablecoins, the financial stability risks outweigh the benefits, and she considers tokenized central bank money (Pontes, Appia) the better solution. Bundesbank President Joachim Nagel, on the other hand, came out in favor of euro stablecoins in February 2026. So there is no single position within the Eurosystem.
• Dollar stablecoins: about 99% of the market is dollars. The cap in MiCA Article 23 (1 million transactions and €200 million a day as a means of exchange within a currency area) is meant precisely to prevent the euro from being crowded out.
• The Banca d'Italia also warns: the stablecoin route would only be truly cheap if the beneficiary spent the token directly, but that would lead to the local currency being crowded out.
• forecast According to Standard Chartered, up to 1 trillion dollars of emerging-market bank deposits could flow into dollar stablecoins by 2028.

14.11 The link to the Treasury market and interest rate dependence

The big issuers are major buyers of US Treasury bills: Tether alone held about 115 billion dollars of them in June 2026. A run could force fire sales in this market. The issuers' business model also hinges on short-term rates. If rates fall, the temptation to make riskier investments grows.

14.12 Legal uncertainty

What a token holder's claim is worth differs by jurisdiction. It is an open question how to treat a token issued in parallel in several jurisdictions (for example USDC, issued by both Circle's US and French entities), and what happens in insolvency proceedings. The Commission's MiCA review (consultation May–September 2026, report in mid-2027) touches on these questions too, but meaningful change before 2028 is unlikely.

1. What is fiat currency?
“Fiat” is Latin for “let it be.” Fiat currency is money issued by the state with no commodity behind it; its value comes from the law and from trust in the central bank. The forint, the euro and the dollar are all fiat currencies. In crypto usage, “fiat” simply means traditional money, as opposed to crypto-assets. → Section 1.2

2. What exactly does SWIFT do?
SWIFT is a network of standardized, authenticated messages between banks. It carries instructions (“debit our account, credit X”), but it is not a bank, it keeps no accounts, and it moves no money. Based on those messages, the money moves in the books of correspondent banks and in central bank systems. The fees and the FX spread are charged by the banks. → Sections 3.2–3.6

3. Why does a cross-border transfer take days?
Because there is no shared world ledger. The money travels through correspondent banks, and at every link in the chain a ledger changes. Along the way, cut-off times, time zones, weekend closures and compliance checks slow it down. According to SWIFT, about 80% of the time passes after the payment has left the SWIFT network. Since gpi was introduced, 90% of payments reach the beneficiary’s bank within an hour, but only 43% reach the end customer that fast (2024). → Sections 3.7–3.8

4. How much does a cross-border transfer cost?
Besides the fee, the FX spread is the biggest and often hidden item: a 2% spread on one million forints is about HUF 20,000. The global average for sending a $200 remittance is 6.36% (Q3 2025), 14.55% through banks and 4.85% through digital services. In Hungary, the financial transaction levy comes on top (0.45%, capped at HUF 20,000; 0.9% on currency exchange). → Sections 4.1 and 4.6, Section 2.5

5. Why are there so many conversions in a stablecoin transfer?
Because the blockchain is a separate ledger, independent of the banking system. The money has to leave the bank ledger, become a token, cross the chain, turn back into local currency, and finally land in a bank account again. That is five phases with two or three conversions. The SWIFT route has the same conversions, they just happen inside the banks, out of sight. → Sections 8.2–8.3

6. What is the advantage of a stablecoin?
On-chain, the token moves 24/7, in seconds or minutes, without a correspondent banking chain, for a network fee that does not depend on the amount, and it is programmable. It helps where the banking system is slow, expensive or hard to reach, and on weekends and outside business hours. It adds little where instant domestic payments and cheap fintech transfers already exist. The 24/7 part is only true on the chain; on-ramp and off-ramp move at the pace of banks and exchanges. → Sections 8.5–8.6

7. Is a stablecoin crypto?
Legally, in the EU, yes: under MiCA it is a crypto-asset, but its single-currency variety (EMT) is also legally e-money. Technically, yes too: it is a token running on a blockchain, protected by the same cryptography and held in the same wallets. Economically, though, it is closer to a prepaid liability of a private issuer, backed by bank deposits and government bonds. In one sentence: the packaging is crypto, the contents are traditional money. → Sections 5.11 and 7.8

8. Can stablecoins replace SWIFT?
For now the volume is small: real stablecoin payments make up about 0.02% of global payment flows. And the two are not the same thing: SWIFT is a messaging network, a stablecoin is a means of payment. The real competition is between settlement assets (private stablecoins, bank tokenized deposits, central bank money). Meanwhile SWIFT is building a blockchain ledger of its own, and in 2026 the first live transaction already ran on it. Realistically, the two will coexist. → Chapter 9

9. Who makes money on a stablecoin?
Mainly the issuer. It pays no interest to the holder but invests the reserve in interest-bearing government bonds. This is how Tether earned about $1.5 billion in a single quarter. The issuer can share part of the income with its distributors (Coinbase, for example, receives 50% of the residual income on the USDC reserve). Exchanges, on-ramp and off-ramp providers, and market makers live on trading fees and the FX spread. → Sections 7.5 and 8.8

10. Is a stablecoin safe?
Exactly as safe as its issuer and its reserve. Deposit insurance does not cover it. The regulated kind (EMT in the EU) is fully backed by reserves and redeemable at par. Depegs have still happened: in March 2023, USDC fell to $0.87 because $3.3 billion of its reserve was stuck at the failed SVB. The algorithmic UST collapsed completely in 2022. Exchange bankruptcies, hacks and lost keys are separate risks. → Section 7.6, Chapter 14

11. What is MiCA?
The Markets in Crypto-Assets Regulation, the EU regulation on markets in crypto-assets ((EU) 2023/1114). It sets uniform rules across the whole EU for who may issue crypto-assets and stablecoins and who may provide crypto services. It distinguishes three token categories (EMT, ART, other) and ten services. The stablecoin rules have applied since June 2024, the rest since December 2024, and the transitional period ended on July 1, 2026. → Sections 11.1–11.5

12. What does it take to become a crypto service provider?
In the EU you need a MiCA license (CASP), issued by the national supervisor; in Hungary that is the MNB. The conditions: an EU registered office, fit and proper management, capital of at least EUR 50,000, 125,000 or 150,000 depending on the services, segregation of client assets, IT security and an anti-money-laundering program. Anyone who transfers stablecoins on behalf of clients has also needed a payment license or a payment partner since March 2, 2026. Banks can provide crypto services based on a notification. → Sections 11.5 and 11.9, Section 12.2

13. What levels are there?
Regulation looks like an eight-rung ladder: user (0), software and information service (1), agent working under someone else’s license (2), crypto service provider (3), payment or e-money institution (4), stablecoin issuer (5), bank (6), systemic infrastructure and central bank (7). Each rung brings more capital, responsibility and supervision. The decisive jump is between rungs 2 and 3, where someone first holds other people’s money or keys. → Chapter 12

14. What is the difference between a tokenized deposit and a stablecoin?
On-chain they look the same, but something different stands behind them. A tokenized deposit is a liability of a commercial bank, just like a traditional deposit, so deposit insurance covers it (in Hungary up to EUR 100,000). A stablecoin is e-money from a non-bank issuer (or a bank), with a 100% reserve but no deposit insurance. Tokenized deposits can typically be held only by the bank’s verified customers; a stablecoin, by anyone with a wallet. → Section 10.4

15. Can real estate or shares back a stablecoin?
Not a regulated payment stablecoin. In the EU, at least 30% of an EMT reserve must be bank deposits, and the rest safe, low-risk, highly liquid assets in the same currency. In the US, the permitted reserve is cash, bank deposits, short-term Treasuries, repos and government money market funds. Real estate and shares can be tokenized too, but the result is a tokenized asset, not a stablecoin. If it is a financial instrument, MiFID II and the DLT Pilot rules apply, not MiCA. → Section 7.4, Sections 10.1 and 10.7

16. What is the digital euro?
The European Central Bank’s planned retail digital money: like cash, it is a direct liability of the central bank. Under the ECB’s plans, the pilot starts in 2027 and the first issuance could come in 2029 at the earliest, if the EU legislation is adopted. The basic service will be free, will work offline, will pay no interest, and will probably have a holding limit. Do not confuse it with Pontes, which is wholesale central bank money for banks, moving on a blockchain (live since September 21, 2026). → Sections 10.5–10.6

17. Is crypto legal in Hungary?
Yes. Holding crypto-assets, and buying or selling them through a licensed provider, is legal. The MiCA supervisor is the MNB. The first MiCA-licensed crypto service provider headquartered in Hungary is CoinCash (July 20, 2026), but any EU provider with a passport can serve Hungarian clients. The 2025 Hungarian validation regime and the related Criminal Code offenses were repealed on August 7, 2026. → Section 11.12

18. How is crypto taxed in Hungary?
Since 2022, income from crypto-assets is taxed at 15% personal income tax on the annual net gain, with no social contribution tax. Only conversion into money or into a non-crypto product is taxable; crypto-to-crypto swaps are not. The tax on a year’s crypto income can be reduced by 15% of the losses reported in that tax year and the two preceding years (loss offsetting). Small transactions have a separate exemption. From the 2026 tax year, providers report transactions to NAV (the Hungarian tax authority) under DAC8. → Section 11.12, Section 11.8

19. Can a bitcoin private key be cracked?
With today’s computers, practically no. A private key is a 256-bit random number, and no known classical method can work it back from the public key in any reasonable time. Real thefts come from stolen seed phrases, phishing, and hacked exchanges and bridges, not from weak cryptography. A large enough quantum computer could in theory break the elliptic curve signatures used today. No such machine exists yet, but developers are already working on the move to quantum-safe signatures. → Section 5.5, Sections 14.3–14.4

20. What is a repo?
A repo (repurchase agreement) is a very short-term loan secured by securities. One party sells a security (for example a US Treasury) and they agree that it will buy it back tomorrow or in a few days at a slightly higher price. From the side of the party providing the cash, this is a “reverse repo.” Stablecoin issuers often park their reserves this way, because it is safe and can be turned into cash quickly. → Section 7.4, Appendix A

21. What are on-ramps and off-ramps?
An on-ramp is any service that converts traditional money into crypto-assets or stablecoins; an off-ramp does the reverse. These are the points where the banking system meets the blockchain, so this is where customer identification happens, where a license is needed, and where most of the cost arises. They are typically run by crypto exchanges and MoonPay- or Ramp-type providers, with a MiCA license. → Section 8.8

The sources for the answers are in the source lists of the chapters referenced in each question. The most important ones:

• MiCA Articles 49–50 (springlex.eu): springlex.eu↗
• Banca d'Italia, MISP No. 86: bancaditalia.it↗
• World Bank RPW: remittanceprices.worldbank.org↗
• NAV, crypto income questions and answers: nav.gov.hu↗
• CMS, repeal of the Hungarian validation regime: cms.law↗

16.1 What does this mean for SMEs?

What is already true today, without stablecoins. The cost of a Hungarian SME’s cross-border payment has four parts: bank fees, correspondent bank fees (deducted under SHA/BEN), the FX spread and the transaction levy. Of these, the FX spread is the least visible and often the largest: a 2% spread on one million forints is about HUF 20,000. Even today, an SME should compare its bank’s rate to the mid-market rate, know whose money the fees come out of under the OUR/SHA/BEN code, and know that for euros, SEPA (and from 2027 the mandatory SEPA Instant) is the cheapest route.

Where can stablecoins make a difference?

• On exotic corridors: where correspondent banks are few (in some Asian, African and Latin American countries), the stablecoin route can be faster and cheaper. Between the major currencies (EUR, USD, GBP), where Wise-type providers and instant systems operate, the Banca d'Italia finds no systemic advantage.
• In 24/7 settlement: you can pay on weekends, on holidays and across time zones, with no cut-off times.
• In programmable payments: escrow payments, milestone-based payouts, automatic splits, payouts tied to shipping data.
• In less tied-up working capital: fewer prefunded balances are needed, and export revenue arrives faster.
• In access to dollars: for companies in high-inflation countries or countries with capital controls. (For a Hungarian SME this matters less often.)

The limits.

• The law of the destination country: the Turkish example shows that a stablecoin payment can be prohibited on the receiving side.
• Compliance: the Travel Rule, business customer identification, and since March 2, 2026 a payment license for EMT transfers.
• Accounting and tax: crypto-assets are booked and taxed differently from traditional money.
• The partner’s attitude: the supplier also has to want to, and be able to, work with stablecoins.
• Costs at the edges: on-ramp and off-ramp fees and spreads, plus the transaction levy on forints sent to an exchange.
• Issuer risk: a stablecoin is not a deposit, and deposit insurance does not cover it.

The change will probably not reach SMEs by them starting to buy stablecoins themselves. Much more likely, their bank, their fintech provider or the card network will, in the background, use tokenized money or stablecoins (the “stablecoin sandwich”), and all the SME notices is that the service got faster, cheaper and more predictable.

16.2 What does this mean for the world?

Dollar power through a new channel. About 99% of stablecoins are dollar-denominated, and issuers have become significant buyers of short-term US Treasuries (Tether alone held about $115 billion in June 2026). With the GENIUS Act, US policy explicitly treats stablecoins as a channel that extends the reach of the dollar and increases demand for Treasuries. The EU responds with several tools: the limits in Article 23 of MiCA, the digital euro, the Pontes and Appia systems, and euro stablecoins from bank consortia (Qivalis).

Emerging markets. In Argentina, Turkey, Nigeria and elsewhere there is real demand for a “digital dollar.” This also explains the low velocity of USDT: many people hold it as savings.

The banks’ response. Tokenized deposits (JPMD, Citi Token Services), SWIFT’s shared ledger, bank consortium stablecoins, card network stablecoin settlement, and bank licenses for crypto firms. The two islands are growing together.

The public sector’s response. Central banks see tokenized central bank money as the safe “money leg” (Pontes is already live, Agorá has been tested, the digital euro pilot starts in 2027). According to the BIS, the future belongs to a unified ledger, with tokenized central bank reserves, deposits and government bonds.

A new kind of payer: machines. The open x402 protocol (Coinbase, 2025) revives the HTTP 402 “Payment Required” status code. A server quotes a price for a request, an AI agent signs a USDC transfer authorization, a facilitator settles it on-chain, and the server serves the request. By mid-2026 about 205 million transactions had run on it in total, but with only about $53 million in volume, and roughly half of the early traffic was tests or bots. Stripe’s Tempo chain launched with its own Machine Payments Protocol. Traditional rails (card minimums, fixed fees) are weakest at micropayments, so this is where stablecoins may have a structural advantage.

16.3 What is driving the change?

Five forces are at work at once.

1. The traditional system is falling short of its own targets. According to the FSB, the G20’s 2027 targets (1% retail cost, 75% within one hour) will probably not be met; the global average remittance cost is 6.36% (Q3 2025), 14.55% through banks. This makes policymakers more open to new solutions too.
2. Regulation has become clear. Between 2024 and 2026 the legal framework for regulated stablecoins took shape: MiCA (EU), the GENIUS Act (US), and the Hong Kong, UK and Singapore regimes. That is why regulated players (banks, card networks) could enter.
3. The technology has matured. Ethereum fees fell to the order of cents by 2026, payment-focused chains launched (Arc, Tempo), and issuers’ cross-chain burn-and-mint solutions reduce bridge risk.
4. Tokenized assets are growing. Tokenized Treasury funds amount to about $14.7 billion (October 2026). On-chain assets need on-chain money, and stablecoins, tokenized deposits and central bank money are competing for that role.
5. The big players have moved. Visa’s stablecoin settlement accelerated to an annual run rate above $20 billion (Sep 2026), Mastercard bought BVNK, SWIFT is building a blockchain ledger, and Western Union launched its own stablecoin.

There is no shortage of brakes either. Real payment usage is still small (about 0.02%), the stablecoin supply has been stuck around $300–320 billion since the end of 2025, on-ramp and off-ramp remain expensive, central banks and the BIS are skeptical, and dollar stablecoins raise questions of monetary sovereignty.

Forecasts (each one a forecast, not a fact):

• Standard Chartered (October 2026): $2 trillion in stablecoin supply by 2030 (it previously expected this by 2028 and pushed it back).
• Citi GPS (September 25, 2025): $1.9 trillion in the base case, $4 trillion in the bull case by 2030.
• Standard Chartered: by 2028, up to $1 trillion of emerging market bank deposits could flow into dollar stablecoins.
• The skeptical view: according to the BIS (June 2025), stablecoins fail the three tests of money, and according to the Banca d'Italia (July 2026), they have no systemic cost advantage in remittances.

16.4 The whole thing on one page

1. Money is a ledger entry. The money in your bank account is a liability of your bank, cash is a liability of the central bank, and e-money and stablecoins are liabilities of their issuers. “Fiat” is state money with no commodity behind it.
2. Every payment is a ledger change. The more ledgers have to be reconciled, the slower and more expensive the payment.
3. Domestic payments are already fast and cheap, because banks settle on a shared central bank ledger (VIBER, AFR in 5 seconds; T2, SEPA Instant).
4. Internationally, there is no shared ledger. Banks hold accounts with each other (nostro/vostro) and send instructions over SWIFT. SWIFT carries messages, not money.
5. Cross-border transfers are slow and expensive because the many intermediaries, cut-off times, time zones, checks, prefunding and the FX spread all add up. gpi has sped things up, but 80% of the time passes after SWIFT.
6. The FX spread is the biggest hidden cost. Wise uses netting and local accounts, so most of the money never even crosses the border.
7. A blockchain is a shared ledger kept in sync by many machines. The private key is the ownership itself, and “crypto” refers to cryptography.
8. A stablecoin is a token pegged to a traditional currency, with a reserve behind it. The issuer lives on the interest from the reserve. Legally and technically it is crypto; economically it is closer to e-money.
9. A stablecoin transfer has five phases, with two or three conversions, because the blockchain is a separate ledger. The on-chain part is cheap (~0.4%); on-ramp and off-ramp are expensive (0.30–8.96% in total in the Banca d'Italia’s measurements).
10. Today the stablecoin is a small competitor, but it exerts big pressure. Real payments make up about 0.02% of global flows. The real competition is between settlement assets (stablecoins, tokenized deposits, central bank money), and banks and SWIFT are adopting the technology too.
11. Regulation decides who may do what. In the EU, MiCA (EMT, ART, CASP), the TFR (Travel Rule) and PSD2/EMD2 (payment and e-money institutions) are decisive, and since March 2, 2026, EMT transfers also require a payment license or a payment partner. In the US, the GENIUS Act is the basis. In Hungary, the MNB supervises, and the validation regime ended on August 7, 2026.
12. The ladder: user → software → agent → crypto service provider → payment/e-money institution → stablecoin issuer → bank → infrastructure. The decisive jump is when someone first holds client money or client keys.
13. The risks sit with the issuer, the bridges, the exchanges, key management, money laundering and monetary sovereignty.
14. SMEs will probably meet the change through their banks and fintech providers. The stablecoin route is useful on exotic corridors and outside business hours, but the destination country’s law must always be checked.
15. The change is driven by the old system falling behind, new regulation, technology and the moves of big players. The future will probably be one of coexistence and convergence: the two islands are growing together.