When Friction Disappears

What happens when moving money costs nothing and takes no time?

Friction is everywhere in the current system. Settlement takes days. Cross-border transfers take longer. Fees accumulate at every hop. Business hours constrain when value can move at all. Each of these is simultaneously someone's revenue and someone else's foregone transaction—which is why they persist long after the technical reason for them has disappeared.

The technology to remove them exists and is deployed in places. India's UPI processes more than 12 billion transactions monthly; Brazil's PIX exceeds 4 billion.¹ Neither country is wealthier than the United States, and both built instant payment infrastructure faster, precisely because neither had a mature card industry whose economics needed protecting.

This chapter is about what follows: instant settlement, money that carries its own conditions, financial services delivered by machine, and a set of second-order consequences that are considerably less comfortable than the efficiency case suggests.

Because friction is not only rent extraction. Some of it is doing structural work that nobody notices until it is gone.

2026 Snapshot — The Emerging Infrastructure

Real-Time Payments

More than sixty countries operate instant domestic payment systems, moving well over $100 trillion annually in aggregate. The leaders are not the traditional financial centers. India and Brazil demonstrated that a state-backed, zero-or-near-zero-fee payment utility achieves adoption that private networks charging interchange do not.

The remaining gap is international. Instant domestic settlement in two countries does not produce instant settlement between them, because the connection runs through correspondent banking.

Stablecoins

Circulation exceeds $150 billion, with transaction volume above $10 trillion in 2023—a figure dominated by trading but with a growing real-payments component.² Their actual function is worth stating plainly: they are dollar access for people who want dollars and cannot obtain a US bank account. In economies with unstable currencies and capital controls, that demand is enormous and largely unmet by the formal system.

Institutional variants have emerged separately, with banks operating permissioned tokens for internal settlement rather than public circulation.

Central Bank Digital Currencies

More than 130 countries are researching or piloting CBDCs, and roughly a dozen have launched.³ China's e-CNY is the largest experiment by a wide margin, covering over 200 million users and processing more than $250 billion in cumulative transaction volume.⁴

Uptake has been underwhelming nearly everywhere, including China, and the reason is instructive. In countries with functioning instant payment systems, a CBDC solves no problem the user actually has. The Bahamas and Nigeria both launched early and both struggled with adoption. The US Federal Reserve remains at research stage with no commitment to a retail CBDC, and Congressional resistance to one is substantial.

The genuine design question is not technical but political: a CBDC gives a central bank direct visibility into, and potentially direct control over, individual transactions. That is a governance decision wearing a technology costume.

Tokenization

Representing conventional assets—securities, funds, real estate, commodities—as transferable digital records. BlackRock's tokenized Treasury fund launched in 2024, and JPMorgan's Onyx platform processes billions daily in institutional settlement.⁵

This is the least publicized and most likely to matter. Tokenization does not require anyone to believe in cryptocurrency; it requires only that settling a trade in seconds is better than settling it in two days, which no one disputes.

Notable Players

Public infrastructure now competes directly with private networks. FedNow, the RTP network, UPI, and PIX represent a view of payment as utility rather than product, and where they have been deployed aggressively, private payment margins have compressed.

Stablecoin issuers divide between Tether, which dominates volume with persistently contested reserve transparency, and Circle's USDC, which took the regulated-and-audited position and trades scale for legitimacy. PayPal's entry signals that the category has become respectable enough for incumbent fintech.

Institutional settlement ventures—Fnality, Partior, and bank-consortium platforms—are attempting to build shared settlement infrastructure among competitors, which is an organizational problem more than a technical one and has historically been where such efforts fail.

Tokenization platforms include Securitize for issuance, Ondo for tokenized Treasuries, and Broadridge for enterprise infrastructure, alongside the incumbent custodians who correctly identify this as a threat to their reason for existing.

AI in finance runs from Bloomberg's domain-specific models to S&P's Kensho to the layer of startups applying general models to underwriting, advice, and compliance.⁷

Instant Settlement

Simultaneous trade and settlement eliminates the window during which one party has performed and the other has not. The direct consequences are real: no counterparty risk over the settlement period, no capital held against it, no settlement failure, and no reason for markets to close.

The requirements are demanding. Instant settlement means pre-funded positions, because there is no float period to arrange liquidity in. That trades one problem for another—counterparty risk becomes liquidity management, and liquidity management is harder for smaller participants than for large ones. Cash and asset legs must move atomically or the atomicity is illusory.

The consequence that receives least attention is that settlement delay is also error-correction time. The two days between trade and settlement are two days in which a mistake can be caught and unwound. Instant, final, irreversible settlement means a fat-fingered order or a compromised credential produces a transfer that cannot be reversed by anyone. The system becomes efficient and simultaneously unforgiving, and financial systems have historically needed to be forgiving because humans operate them.

Programmable Money

Money that carries conditions: transfers that execute when specified criteria are met, without an intermediary verifying and releasing.

The applications are genuine. Escrow without an escrow agent. Weather-indexed crop insurance that pays automatically when a rainfall threshold is recorded, eliminating claims adjustment entirely—which matters most for smallholder farmers for whom conventional insurance was never economic. Royalty splits that distribute to many parties instantly. Supply-chain payments that release against verified delivery.

The implementation approaches vary in ways that matter more than the marketing suggests: smart contracts on public ledgers, programmable interfaces over conventional bank rails, CBDC designs with conditions built in, and hybrids where off-chain logic triggers on-chain settlement. Most production systems will be the boring middle options rather than the ideologically pure ones.

The risks are underexplored relative to the enthusiasm.

Bugs become monetary events. Code that handles money handles it exactly as written, including where written wrongly, and the history of smart contract exploits is a history of nine-figure errors that could not be recalled.

Rigidity replaces judgment. Human intermediaries exercise discretion—waiving a condition for a customer in unusual circumstances, delaying a foreclosure, recognizing that a technically valid instruction is obviously an error. Encoded conditions do not, and the discretion being eliminated is frequently the discretion that made the system humane.

And the same mechanism that enables useful conditions enables harmful ones. Money that can only be spent on approved categories, expires on a schedule, or is restricted by geography is the same technology as escrow-without-an-agent, differing only in who sets the conditions. Programmable welfare payments that cannot buy alcohol are a policy proposal in several countries; programmable money makes them trivially enforceable. Programmable money is the single clearest case in this book of a capability whose character depends entirely on who holds the programming rights, and that question is being settled now, mostly by default.

AI-Powered Finance

Advice is the largest opportunity. Robo-advisors already manage over $1 trillion doing basic asset allocation.⁶ Conversational systems can do considerably more: comprehensive planning across tax, retirement, insurance, and debt, delivered to people for whom a human financial adviser was never economically available. Since financial advice has historically been rationed to those with assets large enough to justify the fee, this is a genuine distributional improvement. Its constraints are liability and suitability regulation rather than capability.

Underwriting extends credit access through alternative data and simultaneously creates the sharpest fairness problem in applied AI. A model that discovers a proxy for a protected characteristic and prices against it is illegal, effective, and hard to detect—and the affected applicant sees only a declined application with no explanation.

Trading is already substantially automated, with algorithms representing somewhere between 60 and 80 percent of US equity volume depending on definition.⁸ The systemic concern is correlation: many participants using similar models trained on similar data respond to shocks similarly, which converts diversity of opinion—the thing that makes markets stable—into synchronized behavior.

Operations—reconciliation, reporting, compliance monitoring, document processing—is where the headcount actually goes. This is the largest employment effect in the sector and attracts the least commentary.

Cross-Border Transformation

The remittance market moves over $1 trillion annually at an average cost above 6 percent, which is roughly $60 billion extracted from the people least able to spare it. Reducing that to 1 percent would transfer approximately $50 billion a year to low-income households—a larger sum than global development aid budgets, achieved through infrastructure rather than charity.

Three mechanisms are converging on it. Stablecoin rails settle dollar-denominated transfers in minutes at negligible cost, and are already used heavily in corridors where formal channels are expensive. Interlinked real-time payment systems connect domestic instant rails directly, bypassing correspondent banking. And incumbent improvement is real: SWIFT gpi now settles roughly half of cross-border payments within thirty minutes, which is dramatically better than the historical baseline.⁹

The obstacle is not technical. It is that instant cross-border settlement moves money faster than compliance regimes can evaluate it, and those regimes are national while the flows are not. Every proposal to speed settlement runs into anti-money-laundering requirements designed on the assumption that there would be time to check.

Second-Order Impacts

Bank runs are now instant, and prudential regulation has not caught up. Silicon Valley Bank's depositors attempted $42 billion in withdrawals in a single day. Every liquidity rule in the regulatory framework was calibrated to a world where withdrawal required physical presence and social coordination required a phone tree. Instant payments plus group chats removed both constraints, and the resulting failure mode is faster than any supervisory response can be.

Financial inclusion may arrive through identity rather than through banking. The binding constraint on the 1.4 billion unbanked is documentation, and mobile money has already reached over 1.75 billion registered accounts by routing around the formal banking system entirely.¹⁰ The likely path to inclusion is digital identity plus a phone, not bank branches.

Monetary sovereignty erodes for weak-currency states. Stablecoins give citizens of high-inflation economies direct access to dollars. From the individual's perspective this is protection against their government's monetary mismanagement. From the state's perspective it is the loss of monetary policy, seigniorage, and capital controls simultaneously. Both descriptions are accurate, and the second explains why the regulatory response in those countries is hostile.

Banks lose the float and must reprice everything else. A meaningful portion of bank revenue derives from holding money in transit. Instant settlement removes it, and the response will be explicit fees for services currently presented as free.

Programmability makes financial control granular. The same infrastructure enabling automatic insurance payouts enables spending restrictions, expiring balances, and transaction-level surveillance. Whether that capability is constrained is a legal question that is currently being answered by not asking it.

The Path Forward

Near-Term Likely (2026–2032)

Instant domestic payment becomes universal across major economies, and interchange-based card economics come under sustained pressure for the first time in sixty years.

Securities settlement moves toward same-day for simple instruments; the complex ones lag because the operational dependencies are deeper than the technology.

Stablecoin regulation clarifies in major jurisdictions, and regulated issuance becomes a bank product. Remittance costs fall meaningfully—perhaps by a third—without approaching zero.

AI fraud detection becomes effectively mandatory. AI-assisted underwriting expands access and produces the first serious fair-lending enforcement actions.

Plausible (2032–2040)

Major markets operate continuously, and the trading day becomes a historical artifact.

Programmable conditions become routine in business-to-business payment, where the parties are sophisticated and the conditions are contractual rather than paternalistic.

AI-assisted financial management reaches most households, and financial advice ceases to be a luxury good.

Cross-border settlement in major corridors approaches domestic cost and speed. Mobile-plus-identity infrastructure brings account access above 90 percent of adults globally.

Wild Trajectory (2040+)

Any asset, any currency, anywhere, settled immediately at negligible cost—finance as a solved logistics problem.

Or: legacy systems persist because the incumbents collecting the friction rents are politically effective, regulation moves at its accustomed pace, and 2040 looks like a somewhat faster version of today. This has been the outcome of most previous predictions of imminent financial transformation, and it should be the base case.

Risks and Guardrails

Instant bank runs. Guardrails: liquidity requirements recalibrated for instant withdrawal rather than branch queues; pre-positioned central bank facilities that can be drawn in hours; and honest acknowledgment that deposit insurance limits designed in a slower era are now the primary determinant of whether a run starts.

Irreversibility. Guardrails: reversal windows for retail transactions even where the underlying rail settles finally; distinguishing between wholesale settlement, which should be final, and consumer payment, which should not be; and preserving a mechanism to unwind obvious error.

Programmable control. Guardrails: legal limits on permissible conditions attached to money, particularly for state-issued currency; guaranteed convertibility to unconditional form; and a preserved cash option, which is the only payment instrument that cannot be programmed, surveilled, or switched off.

Correlated automation. Guardrails: circuit breakers; diversity requirements or capital charges that discourage model monoculture; supervisory visibility into automated trading; and stress tests that assume systems fail together rather than independently.

Exclusion by digitization. Guardrails: offline-capable payment; simplified identity pathways for people without conventional documentation; mandated access obligations; and retention of non-digital options for those who cannot use digital ones.

Conclusion

Finance is information technology, and most of the constraints that define it—settlement windows, business hours, national borders, per-transaction fees—are artifacts of the era in which the information moved on paper rather than properties of money itself.

Those artifacts are being removed. The rails exist, the programmable layer works, and AI handles the risk assessment that previously justified the delay. The direction is not seriously in question.

What the efficiency framing obscures is that some of the friction was load-bearing. Settlement delay was also error-correction time. The slowness of bank runs was also the survivability of banks. Human intermediation was also human discretion, exercised on behalf of people whose circumstances did not fit the rule. Remove the friction without replacing the function it was quietly performing, and the system becomes faster, cheaper, and more brittle at the same time.

The distributional question is sharper still. Roughly $60 billion a year is currently taken from remittance senders in fees, and the technology to reduce that to almost nothing already exists. Whether it happens is not a technical matter—it depends on whether the institutions collecting those fees can slow the transition long enough to keep collecting them, which historically they have been very good at.

That is the whole shape of this chapter. The engineering is largely done. What remains is an argument about who keeps the money that friction was hiding, and that argument will be settled by regulation and market power rather than by anything in a whitepaper.

Endnotes — Chapter 54

1. India's UPI processed more than 12 billion transactions monthly (2024); Brazil's PIX exceeded 4 billion monthly. Both are state-backed near-zero-fee systems that achieved adoption faster than private alternatives in wealthier markets.
2. Stablecoin transaction volume exceeded $10 trillion (2023). The majority is trading-related, though payment, payroll, and remittance use is growing, particularly in economies with unstable local currencies.
3. CBDC research or pilots are underway in more than 130 countries per the Atlantic Council tracker; roughly a dozen have launched, with adoption generally below expectations.
4. China's e-CNY pilot covers more than 200 million users with cumulative transaction volume over $250 billion—the largest CBDC experiment to date, and still a small fraction of Chinese payment volume relative to Alipay and WeChat Pay.
5. Tokenization: BlackRock launched a tokenized Treasury fund (BUIDL) in 2024; JPMorgan's Onyx platform processes billions daily in institutional settlement.
6. Robo-advisor assets under management exceed $1 trillion globally, with Betterment, Wealthfront, and Schwab Intelligent Portfolios among the largest. Current offerings perform basic allocation rather than comprehensive planning.
7. LLM-powered financial tools include Bloomberg's domain-specific model (2023) and a range of startup products; regulatory treatment of AI-generated financial advice remains unsettled.
8. Algorithmic trading represents roughly 60–80 percent of US equity market volume depending on definition and venue.
9. SWIFT gpi provides end-to-end tracking for cross-border payments; approximately half settle within 30 minutes on the gpi network, a substantial improvement over historical correspondent banking timelines.
10. Mobile money has reached more than 1.75 billion registered accounts globally, extending financial access by routing around conventional banking rather than through it.