The Architecture
of Money

Money is an architecture, not an object.

Modern money works because different layers reinforce one another. The unit of account anchors prices. Central-bank money anchors settlement. Commercial-bank deposits provide most everyday purchasing power. Payment systems move claims. Credit creates and destroys deposits. Regulation and insurance support trust.

What is money?

The Bank of England describes modern money as a special type of IOU and identifies currency, bank deposits and central-bank reserves as key forms of money.

From commodity to ledger.

The technology changes, but the fundamental problem remains: how can a society coordinate the transfer of value while maintaining trust in the unit of account and settlement?

Three kinds of money you should not confuse.

The Bank of England explains that commercial-bank deposits are private money, while reserves are central-bank money; the BIS describes the modern system as a two-tier architecture in which commercial banks issue deposits while central banks provide the ultimate settlement asset.

Where does new deposit money come from?

In modern banking, lending and deposit creation are linked. When a commercial bank makes a loan, it records a loan asset and a matching deposit liability. The borrower can then spend the deposit.

The Bank of England explicitly rejects the simplified idea that banks merely lend out pre-existing deposits; its research explains that lending creates deposits and that loan repayment extinguishes deposit money.

Watch the balance-sheet logic.

The bank is simultaneously a lender, money issuer and risk manager.

The central bank sits at the settlement core.

The BIS describes central-bank reserves as the ultimate risk-free settlement asset supporting the “singleness” of money, while commercial banks provide the private deposit layer.

Moving money is a chain of messages, ledgers and settlement.

Payment systems include cash, cards, account transfers and wholesale settlement infrastructure. In the United States, the Federal Reserve describes payment infrastructure as including currency, checks, ACH and large-value Fedwire services.

Credit connects today's spending to tomorrow's income.

Money can remain nominally stable while losing real value.

Inflation is a sustained increase in the general price level. When prices rise, a fixed amount of money buys fewer goods and services.

Money is stable only while trust and convertibility remain credible.

This is why the architecture of money is inseparable from the architecture of financial stability. Deposit insurance, central-bank liquidity, capital regulation, payment-system resilience and resolution rules all protect the ability of money to function as money.

Digital money changes speed, not the underlying need for trust.

The BIS 2026 Annual Economic Report argues that digital innovation can improve payment efficiency but raises new macro-financial challenges, and emphasizes that trust in money still depends on the underlying institutional architecture.

The next architecture may combine money and assets on programmable ledgers.

BIS research in 2025 proposed a next-generation architecture built around tokenised central-bank reserves, tokenised commercial-bank money and tokenised government bonds on a unified ledger, while stressing that singleness, elasticity and integrity remain essential properties.

The future question is not “cash or crypto?”

The deeper question is: who issues the claim, what backs it, where is it settled, how is it redeemed, who absorbs losses, and which institution maintains trust?

BIS's 2026 work frames these properties as central to the evolution of the monetary system in the digital age.

Twenty questions to analyze any new form of money.

One monetary system. Five layers. One trust chain.

Money is not wealth.

Every monetary claim has an issuer somewhere.

Architecture rule: never ask only “what is this money worth?” Ask “whose liability is it, and what happens if the issuer fails?”

Money creation and destruction are two sides of the same ledger.

Bank of England research explains that commercial-bank lending creates matching deposits and that the quantity of money ultimately depends on the broader monetary-policy and banking framework.

Reserves are not simply “money that banks lend out.”

Central-bank reserves are a settlement asset held by eligible financial institutions. Commercial-bank deposits are liabilities of commercial banks and are the money households and businesses normally use.

Central-bank money is backed by the central bank's balance sheet.

How a central bank rate reaches the real economy.

Payment is not the same thing as settlement.

A customer can see a payment instruction accepted while the underlying inter-institution obligations still require clearing and settlement. Settlement finality matters because the financial system needs a point at which an obligation is irrevocably discharged under the relevant legal framework.

Money becomes more complicated when it crosses borders.

BIS Project Agorá in 2026 demonstrated a prototype for tokenised commercial-bank deposits and central-bank reserves supporting multi-currency wholesale settlement and atomic cross-border transactions.

Stablecoins raise an old monetary question in new technology.

BIS's 2026 work argues that stablecoin innovation should be considered alongside the existing two-tier monetary system, with attention to settlement, integrity and interoperability.

Tokenisation could merge messaging, reconciliation and settlement.

BIS's 2025 blueprint proposed a next-generation monetary and financial system combining tokenised central-bank reserves, tokenised commercial-bank money and tokenised assets; Project Agorá's 2026 prototype demonstrates practical work toward that direction.

Five thousand years of monetary architecture.

If you invent a new money, answer these first.

What changed in the latest research?

Think of money as a stack of promises.

A bank is a machine that transforms claims.

The Bank of England emphasizes that bank capital is a loss-absorbing buffer supporting confidence in deposits, while liquidity arrangements help banks meet payment outflows.

If banks can create deposits, why can't they create unlimited money?

Money can disappear without cash being destroyed.

Central-bank balance-sheet policy changes the composition of private portfolios.

The Bank of England's recent monetary-policy analysis continues to use broad-money developments and bank lending as indicators of the transmission from balance-sheet policy to spending and inflation.

More money does not automatically mean proportionally more inflation.

The effect of additional money depends on what people do with it, the productive capacity of the economy, credit conditions, expectations, asset markets and monetary policy. A useful conceptual identity is:

This identity is an accounting relationship, not a complete causal model. The real economy determines how it translates into inflation, output and asset-price effects.

Why should ₹1, $1 or £1 in one bank equal the same unit elsewhere?

The modern system depends on par convertibility: commercial-bank deposits should be redeemable into central-bank money at the same nominal value. Without this “singleness,” money would fragment into different exchange rates depending on the issuer.

Recent Bank of England work describes reserves as the ultimate settlement asset and stresses the institutional framework needed to keep commercial-bank money equivalent to central-bank money.

A good monetary system must be elastic enough for payments and credit to keep working.

Demand for liquidity changes. Tax dates, settlement cycles, market stress, withdrawals and credit demand can all change the need for liquid assets. A system that cannot expand liquidity when needed can turn a temporary mismatch into a solvency problem.

The Bank of England's 2026 market-operations framework explicitly describes reserves as the ultimate means of settlement and its market operations as part of monetary-policy and financial-stability implementation.

Digital money creates new risks alongside new efficiency.

Use scenarios to understand whether a monetary architecture is resilient.

Score a monetary system on six properties.

Ask five questions about every form of money.

Cash, reserves, deposits, CBDCs and stablecoins are not interchangeable.

The IMF's May 2026 analysis explicitly distinguishes tokenised deposits, wholesale CBDCs and regulated stablecoins by where the settlement asset and risk ultimately sit.

Who captures the economics of money creation?

Seigniorage broadly refers to the economic benefit associated with issuing money. Its exact measurement differs by instrument and accounting system, but the basic question is simple: what does the issuer receive when it creates a monetary liability, and what does it cost to maintain that liability?

A currency is also a political and institutional system.

Monetary sovereignty is partly about the ability of a government and central bank to define the unit of account, operate monetary policy and provide public settlement infrastructure. It becomes complicated when residents increasingly use foreign currencies, stablecoins or foreign payment networks.

The global monetary system extends beyond one country's domestic banking system.

International trade and finance create cross-border claims in major currencies. Banks, corporations, funds and governments can borrow, lend and hedge in currencies different from their domestic unit of account.

“Eurodollar” does not mean euros.

The eurodollar market historically refers to U.S.-dollar deposits and dollar-denominated funding outside the United States. It illustrates an important feature of modern money: a currency's financial network can extend far beyond the country that issues the underlying central-bank money.

This architecture creates global benefits but also means dollar funding conditions can transmit U.S. monetary and financial shocks globally.

The textbook money multiplier is useful—but incomplete as a description of modern bank lending.

The simplified classroom model starts with central-bank reserves and imagines a mechanical sequence of deposit expansion through reserve requirements. Modern banking works differently: commercial-bank lending can create deposits first, while banks obtain settlement liquidity and reserves through the payments system and central-bank framework.

Bank of England research explicitly explains that the money multiplier does not accurately describe the money-creation process in the modern banking system.

Capital is not money—but it protects money.

Bank capital is loss-absorbing funding, not a payment instrument. It matters because depositors and other creditors hold claims against the bank while bank assets can lose value.

The payment experience hides a large stack underneath.

Federal Reserve payment-system principles emphasize clear settlement finality and the use of central-bank money for settlement where practical and available.

Stablecoins are moving from experiment to monetary-policy question.

The IMF published two 2026 working papers that analyze stablecoins from different angles. One studies their run risk and finds that requiring safer backing can reduce fragility but may affect issuer incentives. Another DSGE investigation finds that stablecoins can amplify shocks and weaken monetary-policy effectiveness in the modeled economy, while stronger reserve backing can reduce volatility. These are research models, not settled empirical conclusions.

Digital money is becoming economically large enough to matter.

A 2026 Federal Reserve research paper notes that the stablecoin market had exceeded $300 billion in market capitalization by April 2026, with USDT and USDC accounting for the large majority. The paper is a Federal Reserve research paper, not an official monetary aggregate.

Tokenisation has moved from blueprint to controlled real-value testing.

BIS reported in July 2026 that Project Agorá conducted real-value testing in a controlled environment. Twenty-eight financial institutions and central banks completed transactions across 17 scenarios, with total transaction value of approximately CHF 800,000 across a selection of currencies. The prototype combines tokenised commercial-bank deposits with tokenised central-bank reserves on a shared platform.

Agorá's May 2026 prototype demonstrated atomic multi-currency settlement, and the project's July 2026 update reports controlled real-value testing.

Four possible architectures for the next era of money.

Seven properties define a resilient monetary architecture.

Recent IMF work frames tokenization around infrastructure, assets and services and emphasizes that the settlement asset, legal certainty, governance and interoperability are architectural choices—not merely technical details.

Tokenized finance has three layers.

The IMF's July 2026 tokenization note explicitly uses this three-layer framework: infrastructure, assets and services.

Tokenization may move risk rather than remove it.

The IMF's July 2026 analysis warns that tokenization can shift risk toward infrastructure, platforms and code, while automated liquidity and collateral operations may accelerate transmission during stress.

A token is not automatically ownership.

A digital representation of an asset does not by itself answer the legal question of what the holder owns. The legal system must determine whether the token represents a direct property right, a claim against an issuer, a beneficial interest, or another contractual arrangement.

Instant settlement creates a new liquidity problem.

Traditional settlement delays can give institutions time to mobilize collateral and funding. Continuous tokenized settlement can compress that time.

The stablecoin problem is not only about reserves.

IMF analysis emphasizes that stablecoin par convertibility depends not just on reserve asset quality but on redemption design, issuer capacity and underlying market liquidity.

Why foreign-currency stablecoins can be different in emerging markets.

When domestic inflation, currency volatility or weak institutional credibility are high, residents may have stronger incentives to hold foreign-currency assets. Stablecoins can lower the technological barrier to accessing those assets.

The IMF's June 2026 analysis of Nigeria describes stablecoins as a growing cross-border channel, reducing payment frictions while creating new regulatory and monetary-policy trade-offs.

You cannot manage a monetary system you cannot measure.

Digital money creates an apparent paradox: transactions can become more visible on-chain while important economic information can remain hidden off-chain.

AI could turn programmable money into autonomous financial infrastructure.

As AI agents begin to transact, the architecture of money may need to support machine-to-machine payments, authorization limits, programmable spending, identity and auditability.

Who should control the architecture of money?

1. Which forms of money should remain public infrastructure?
2. Which forms can safely be privately issued?
3. Where should final settlement occur?
4. Who provides emergency liquidity?
5. Who absorbs losses during a run?
6. Who governs smart-contract code?
7. Who controls access to programmable payment rails?
8. How should privacy coexist with financial integrity?
9. How should cross-border stablecoins interact with national currencies?
10. What happens when an autonomous AI agent makes a financial mistake?

The entire architecture on one screen.

Money is continuously created, transferred, transformed and extinguished.

The same monetary unit can move through households, firms, banks, markets and governments many times. What matters is not only the quantity of money, but also who holds it, what liabilities back it, how quickly it circulates and what economic activity it finances.

Not all financial claims sit at the same distance from the monetary anchor.

Modern finance often runs on collateral as much as it runs on money.

When collateral prices fall, borrowing capacity can shrink even before an institution becomes legally insolvent. Haircuts, margin requirements and collateral eligibility can therefore transmit market volatility into liquidity conditions.

Liquidity and solvency are different problems.

Money becomes more valuable when more people accept it.

Payments are network businesses. A payment method is useful partly because other people, merchants, banks and institutions accept it.

Network effects can reinforce dominant payment standards—but they can also create concentration risk if one provider becomes systemically important.

Too many ledgers can fragment money.

BIS's 2026 Annual Economic Report notes that fragmented networks can require bridges that introduce operational risks and can undermine resilience and the singleness of money.

Digital money must solve a three-way tension.

BIS's Agorá work explicitly includes legal, privacy and AML/CFT considerations in the design of a shared programmable payment platform.

AI agents may become economic actors.

The architecture in one sentence.

The most important insight of this book is that technology changes the form of money faster than it changes the fundamental economic questions:

What this book should never oversimplify.

Money cannot be understood without government finance.

Government debt, taxation, central-bank operations and commercial-bank balance sheets interact. Government spending and taxation affect deposits and reserves; bond issuance changes the composition of private-sector assets; central-bank operations influence the settlement layer.

Anatomy of a bank run.

This creates a potential feedback loop. If depositors believe a bank is weak, withdrawals can force rapid asset sales. Falling asset prices can worsen perceived or actual solvency, creating more withdrawals.

Inflation is a system outcome, not a single-variable event.

The Bank of England describes flexible inflation targeting as balancing temporary deviations of inflation from target against output stabilization when shocks create trade-offs.

One policy rate becomes many market rates.

Market rates do not move one-for-one with the policy rate. Expectations, term premia, credit risk, liquidity and global financial conditions all influence the final price of credit.

Money affects asset markets as well as goods markets.

This channel helps explain why monetary policy can affect equities, bonds and property even before changes in consumer prices become visible.

A monetary shock rarely stays inside one border.

Exchange rates, capital flows, foreign-currency funding and trade finance can transmit changes in one major monetary system into other economies.

What a unified ledger actually tries to solve.

BIS describes a unified ledger as a venue where tokenised central-bank reserves, commercial-bank money and financial assets can coexist, potentially integrating transactions that currently span separate systems.

Project Agorá: from architecture to experiment.

Agorá is best understood as an experimental financial-market-infrastructure project, not as evidence that the world is about to replace existing payment systems overnight.

How to read claims about the future of money.

The architecture matters more than the interface.

That is why the future of money cannot be judged by transaction speed alone. The stronger question is whether the architecture can preserve singleness, finality, liquidity, safety, integrity, interoperability and accountability as technology changes.

Follow the claim, not the interface.

The fastest way to understand monetary architecture is to track the balance sheets of every participant. When one party's asset appears, another party's liability usually exists somewhere else.

What happens when the government spends?

In a simplified operational example, government spending credits a recipient's bank account. The recipient gains a deposit; the commercial bank gains a corresponding reserve balance; the government's balance sheet changes according to the financing arrangement.

Taxes reverse part of the flow—but not in a mechanical “money destruction” sense for every setup.

When a taxpayer pays a government tax from a commercial-bank deposit, the taxpayer's deposit falls and the banking system's reserve position changes. The macroeconomic effect depends on the full fiscal and monetary framework, not just one transaction.

A government bond changes the composition of financial claims.

Bond issuance can exchange one private-sector asset for another: a bank deposit or reserve-linked payment is exchanged for a government security. The consequences differ depending on who buys the bond, how the government account is operated and the monetary-policy regime.

Key insight: “government debt” is simultaneously a liability of the public sector and an asset held by someone else.

Money, credit and debt are related—but not identical.

Credit can accelerate growth—and amplify downturns.

When collateral values rise and lending expands, spending and asset prices can reinforce each other. When asset prices fall, collateral shrinks, lenders tighten standards and borrowers may repay or default—amplifying the downturn.

The monetary system includes institutions that are not banks.

Money-market funds, securities dealers, asset managers, insurers, pension funds and other non-bank financial institutions can provide funding, liquidity and credit-like services. They can also create interconnected chains of leverage and collateral.

That is why monetary stability cannot be assessed by looking only at the central bank's balance sheet.

Repurchase agreements connect collateral to short-term money.

A repo is economically a secured financing transaction: one party receives cash and provides securities as collateral, with an agreement to reverse the transaction later.

Financial systems contain assets that behave like money without being money.

Money-market funds and highly liquid securities can be close substitutes for deposits in some circumstances. But their legal claims, redemption mechanics and risk profiles differ from central-bank money and insured bank deposits.

How a financial shock becomes a monetary shock.

A practical framework for analyzing any new form of money.

1. Define the unit of account.
2. Identify the issuer and its legal liability.
3. Identify the ultimate settlement asset.
4. Map every balance sheet involved.
5. Identify who can redeem and at what price.
6. Identify liquidity providers and emergency backstops.
7. Identify capital and loss-absorption mechanisms.
8. Identify legal ownership and insolvency treatment.
9. Identify interoperability and fragmentation risks.
10. Stress-test a run, cyber outage, market crash and currency shock.
11. Separate empirical evidence from model assumptions.
12. Ask who gains power if the architecture scales.

Five lenses for understanding money.

A high-quality explanation of money has to move between economics, accounting, law, technology and institutions. Looking through only one lens produces misleading conclusions.

There are several meanings of “money.”

The same word therefore describes both an instrument and the system that makes the instrument credible.

Why “one unit is one unit” is an institutional achievement.

Suppose two banks issued deposits that traded at different values. A ₹1 deposit at Bank A might buy ₹0.98 of Bank B's deposit. The monetary unit would have fragmented into issuer-specific monies. Modern payment architecture is designed to prevent that outcome by supporting conversion at par.

Four institutional pillars help preserve par exchange.

The IMF's 2026 discussion highlights settlement finality, central-bank settlement assets, liquidity facilities and deposit insurance as part of the traditional architecture supporting par exchange.

Why a reserve-backed token can still behave differently from a bank deposit.

This comparison is conceptual. Actual legal treatment varies by jurisdiction and product design. It should not be used as a legal classification.

Tokenization does not necessarily mean blockchain.

A tokenized claim is a representation of an asset or liability on a programmable digital ledger. The underlying infrastructure can use distributed ledger technology, a permissioned database or another architecture. What matters economically is not the label “blockchain,” but the properties of the ledger and the legal system around it.

BIS's 2025 unified-ledger blueprint explicitly notes that a unified ledger need not use distributed ledger technology.

Code can execute financial contracts—but code does not eliminate judgment.

The IMF's 2026 tokenization research specifically notes that smart contracts can automate margin calls, collateral transfers and other financial logic, while also shifting risk toward code, infrastructure and governance.

Always-on settlement changes the rhythm of liquidity.

Traditional systems often use operating hours, cutoffs and netting cycles that allow participants to manage liquidity over time. A continuously available tokenized system could reduce these frictions, but it also means institutions may need sufficient liquidity at more moments of the day.

Project Agorá's architecture explicitly explores conditional and potentially around-the-clock wholesale cross-border payments.

Why correspondent banking is difficult.

Project Agorá demonstrates the potential for atomic multi-currency settlement using tokenised central-bank reserves and commercial-bank deposits, but it remains a controlled experimental architecture rather than a global replacement for correspondent banking.

Separate established facts from forward-looking claims.

The architecture can be reduced to ten questions.

1. What is the unit of account?
2. Who issues the monetary claim?
3. Who legally owns the underlying asset or claim?
4. What is the ultimate settlement asset?
5. Can the claim be redeemed at par, and by whom?
6. Who provides liquidity in stress?
7. Who absorbs losses?
8. Who controls the code and infrastructure?
9. What happens if the network fragments or fails?
10. Does the architecture strengthen or weaken trust in the monetary unit?

Money is a trust architecture built on layered balance sheets.

This is why changing the technical form of money can change the allocation of risk without changing the underlying economic questions.

Follow one ₹1,000 payment through the architecture.

One loan transaction creates a paired change in a bank's books.

Bank of England research is explicit that commercial-bank lending creates deposits, while the banking system must manage capital, liquidity and payment flows around those balance-sheet changes.

“What backs the money?” is the wrong question unless you define backing.

A bank deposit is not “backed” in the same way as a stablecoin reserve token. Its credibility arises from bank assets, capital, liquidity, regulation, supervision, deposit protection and central-bank settlement infrastructure. That distinction is central to the modern two-tier system.

Money is a chain of promises nested inside other promises.

Every monetary architecture has characteristic failure modes.

What actually changes when assets and money share a programmable ledger?

In traditional finance, these functions often occur through linked systems. Tokenization can combine them into a synchronized transaction state. That can reduce reconciliation, enable atomic delivery-versus-payment and shorten settlement cycles.

What Agorá proves—and what it does not prove.

How strong is each claim in this book?

If you remember only one framework, remember this.

Technology can make the system faster and more programmable. It cannot remove the fundamental need for credible promises, resilient institutions, clear ownership and a trusted settlement foundation.

Money is society's memory of who is owed what.

Every monetary claim records a relationship across time. A deposit says a bank owes the customer. A government bond says the state owes the holder. A loan says a borrower owes a lender. Money allows these obligations to survive the moment in which they were created.

Money coordinates millions of decisions without a central planner deciding each trade.

That is why monetary architecture is part of economic infrastructure rather than merely a payment technology.

The first job of money may be the least visible: naming economic value.

Prices become comparable when wages, goods, taxes, assets and debts share a unit of account. Without a stable accounting unit, every contract needs additional conversion and valuation rules.

A monetary contract has a nominal side and a real side.

Central banks react to the economy; markets react to central banks.

This creates a feedback system rather than a one-way pipeline. Financial conditions affect the economic outlook, which changes future policy expectations.

Public money sits inside the state's consolidated balance sheet.

Government, treasury and central-bank arrangements differ by country, so the exact accounting mechanics should not be universalized. But conceptually, public monetary liabilities coexist with public-sector assets and government debt.

Nominal interest rates contain inflation expectations.

This Fisher relation is a useful approximation, not an exact law in every period. It helps explain why a rise in expected inflation can influence nominal yields even before realized inflation changes.

The yield curve is a map of market expectations, risk and term compensation.

Long-term yields can incorporate expected future short rates, inflation, growth, term premia, fiscal risk and global demand for safe assets. The curve should therefore never be interpreted through one variable alone.

“How much money exists?” has no single answer.

Economists use different monetary aggregates because different forms of money differ in liquidity and institutional meaning.

Money and credit move together—but not perfectly.

When credit expands, deposits can expand as well. But the relationship depends on repayment, defaults, asset purchases, fiscal flows and the composition of portfolios. A sophisticated analysis therefore tracks both money and credit.

A bank loan has a price because credit has risk.

The loan rate is therefore not simply “the central-bank rate + profit.” It incorporates expected losses, funding, capital, liquidity, competition and borrower risk.

Monetary changes do not affect everyone equally.

Different households hold different portfolios. Some own financial assets; others primarily earn wages. Some borrow at floating rates; others hold fixed-rate debt. A change in rates or asset prices can therefore redistribute wealth and income.

Trust is partly psychological, partly institutional and partly mathematical.

Digital technology can strengthen the mathematical layer while leaving institutional trust unresolved. Conversely, strong institutions can preserve confidence even when the underlying technology is old.

Money has moved from objects to records to programmable state.

Each transition changes how money can be stored, transferred and monitored. None changes the fundamental need for a trusted unit of account and credible settlement.

Technology changes the rails. Institutions determine whether the money is trusted.

That means a successful architecture must answer simultaneously: Who issues? Who settles? Who provides liquidity? Who bears losses? Who governs? And what happens when everything is under pressure at the same time?

Primary sources used for this edition.