Why Payment Innovation Should Not Be Judged by the Architecture It Is Attempting to Replace
At the 2026 Jackson Hole Economic Policy Symposium, Professor Wenxin Du offered an intelligent and deceptively simple challenge to some of the more exuberant claims surrounding financial innovation. The paper was tweeted or X’ed by one of her Harvard colleagues, triggering much debate.
Her paper, Three Myths About Payment Innovation, questions three increasingly popular propositions: that 24/7 atomic real-time gross settlement represents the future of wholesale payments; that retail payment friction is principally a legacy-technology problem; and that stablecoins are inherently efficient for cross-border payments.
There is much in Du’s paper that one agrees with.
A payment system cannot be evaluated merely by asking how quickly a token moves from A to B. Liquidity matters. Netting matters. Compliance matters. Counterparty risk, dispute resolution, market depth, customer support and institutional trust matter.
Her central warning is therefore sound: do not mistake the visible payment rail for the economic architecture of payments.
But it is here that the argument begins to fall short.
Du evaluates technological innovation within a static conception of the financial and economic architecture. Du treats the financial and economic architecture surrounding today’s payment system as a constraint to which innovation must adapt.
As we rapidly transfer from the industrial economy to the intelligence economy, the more important question contextual to the future economy — what if the architecture itself is limiting, warranting a redesign for its relevance in the intelligence economy, an economy that is agentic?
The Bureaucracy of Finance
Financial infrastructure did not emerge in a vacuum. Correspondent banking, clearing houses, central-bank settlement systems, capital requirements, operating hours, compliance processes, foreign-exchange markets and settlement conventions are the products of institutional evolution. Collectively, they constitute what is now being described as the bureaucracy of finance.
Bureaucracy here is not meant pejoratively. Much of existing financial architecture exists for good reason. It provides trust, stability, accountability and risk management.
However, technology evolves faster than institutional architecture; commerce evolves faster than institutional architecture; geopolitics changes faster than institutional architecture. Unfortunately for the incumbent institutional architecture, emerging technology is not of the general-purpose variety.
Significant change is beginning to challenge the status quo, create opportunities and, most importantly, reset expectations; the result is growing institutional latency: the distance between change and the speed at which governing institutions are capable of adapting to change. It is understandable that the gatekeepers of the incumbent architecture — well-meaning intellectuals like Wenxin Du, and bureaucrats — fiercely defend it.
Many financial innovations, and the push-backs to these innovations including Du’s paper, should be understood in this context.
Stablecoins did not emerge because the world lacked electronic databases. Fintech did not emerge because banks lacked computers.
Digital assets did not emerge because modern financial institutions were technologically primitive. They emerged, at least partly, because existing institutions were unable or unwilling to recognise change or reorganise themselves around technological possibilities quickly enough.
Innovation Is Always Small Before It Is Large
Du repeatedly compares emerging technologies in payments with incumbent infrastructure of enormous scale. That is analytically useful but arguably an unfair comparison. Incumbent payment systems possess extraordinary liquidity, regulatory recognition, institutional underwriting, established customer relationships and decades of accumulated network effects. The new systems Du compares them with possess almost none of these things. Naturally, the incumbent appears larger, cheaper and more liquid.
As we step into the intelligence age, it is important to recognise that innovation is almost always marginal when measured against the system it may eventually displace.
The early internet was insignificant beside telecommunications networks. Renewable energy once appeared economically trivial beside fossil fuels. Electric vehicles were dismissed for years against the scale, infrastructure and engineering advantages of the internal-combustion automobile.
China’s electric automobile industry, which continues to challenge traditional American, European and Japanese hegemony on automobile manufacturing, offers a particularly interesting analogy.
Chinese new-energy-vehicle companies have gone from the fringe to challenging an otherwise established manufacturing hegemony. Apart from continuously refining the quality of the automobile and improving battery performance — BYD’s charging technology can add approximately 400 kilometres of range in five minutes, a performance metric that would have appeared technologically implausible only a few years ago — the more than 150 Chinese new-energy car manufacturers appear to have a hidden act up their sleeve.
Their collective manufacturing heft appears to enable Chinese car manufacturers to implement an innovative strategy. It enables these manufacturers to dramatically compress development cycles.
Leading manufacturers now operate on design cycles as short as 18 months, compared with more than 48 months for many legacy manufacturers. This means the Japanese, German or American model is out-featured within months by the 150-odd competing offerings.
The current size and efficiency of an incumbent are poor measures of the future potential of a technology whose economics are still being engineered.
Did Du Frame the Right Question?
Du’s first argument provides the clearest example. She demonstrates the extraordinary liquidity efficiency created by netting. CLS, for instance, settles approximately $7.9 trillion of FX transactions on an average day while requiring roughly $300 billion in actual funding — a netting efficiency of around 96 per cent. CHIPS and securities clearing systems generate similarly substantial liquidity savings.
Her conclusion is intuitively compelling: replace this architecture with trade-by-trade real-time gross settlement and the liquidity requirement explodes. But there is an analytical problem.
Twenty-four-hour availability, real-time settlement, atomicity and gross settlement are not the same technological property. They are different design variables.
- Atomicity determines whether linked legs of a transaction complete together or fail together.
- Gross versus net determines whether obligations are individually funded or offset before settlement.
- Real-time describes timing.
- Twenty-four-seven describes availability.
There is no technological law requiring a future financial system to choose all four simultaneously.
A programmable financial architecture could preserve netting while introducing atomic finality. It could continuously optimise liquidity before settlement. It could dynamically mobilise tokenised collateral. Intelligent treasury systems could offset obligations, source intraday liquidity and determine settlement pathways algorithmically.
The relevant technological frontier is therefore not simply gross settlement versus netting. It is, perhaps, how much of today’s liquidity-management architecture can itself become programmable.
Du asks what happens if atomic settlement is inserted into today’s economics of liquidity. Innovation replies by asking whether technology can change the economics of liquidity.
Her Second “Myth” May Actually Prove the Opposite Point
Du’s argument concerning retail payments is even more revealing.
She notes that SWIFT is considerably faster than consumers often experience: approximately 75 per cent of payments reach the beneficiary institution within ten minutes, with much of the delay occurring in the domestic “last mile.”
Fintech money-transfer operators such as Wise and Remitly solved much of this problem differently. They prefunded local accounts, netted transactions internally, digitised compliance and transformed what looked like an international transfer into two locally executed payments connected through an intelligent treasury operation.
Du concludes that technology was therefore not the principal constraint. Large banks already possessed sophisticated technology but lacked sufficient economic incentives to deploy it for smaller customers.
But that observation arguably proves the institutional-latency thesis.
The technology existed. The incumbent architecture failed to deploy it. Innovation emerged outside the incumbent and reorganised the economics of the transaction.
That is not evidence against technological disruption. It is almost a textbook description of technological disruption.
What it also proves is that innovation forced the incumbent to work better — enabling the incumbent to revert back to its latency argument.
The Stablecoin Sandwich May Eventually Lose Its Bread
The same problem arises with stablecoins. Du describes fiat-to-fiat stablecoin payments as a “stablecoin sandwich”: Fiat → Stablecoin → Blockchain → Stablecoin → Fiat.
The blockchain is merely the filling. The expensive parts may be the fiat conversion, foreign exchange, banking access and on/off ramps surrounding it. This is correct today. But it assumes that the sandwich remains structurally unchanged.
What happens when an increasing proportion of economic activity begins and ends digitally?
- Tokenised deposits are likely to interact directly with stablecoins.
- Stablecoins are likely to settle tokenised securities.
- Corporates are likely to retain digital-dollar balances rather than immediately converting them to fiat.
- Machine-to-machine transactions are likely to require programmable money.
- Digital assets are likely to themselves become collateral.
- Treasury agents may dynamically choose between bank money, stablecoins, CBDCs and tokenised deposits.
Given this likelihood, which is not in the distant future, the two slices of bread begin to shrink — the use of fiat begins to shrink.
There is an even more revealing point in Du’s own analysis. She observes that stablecoins sometimes command significant premiums in economies characterised by inflation, exchange-rate volatility or capital controls.
Here the premium is itself an indicator: it indicates that economic actors are prepared to pay for access, convertibility, optionality and monetary credibility.
Payments Are Now Geopolitical Infrastructure
There is another dimension that conventional transaction-cost analysis risks understating. Payments have become geopolitical infrastructure.
SWIFT access, correspondent banking, dollar clearing, central-bank swap lines, sanctions, capital controls, reserve currencies and cross-border financial networks increasingly intersect with national strategy.
This means that payment infrastructure cannot be evaluated solely in basis points. Resilience has value; sovereignty has value; optionality has value; programmability has value; access itself has value.
A payment architecture that is marginally more expensive may nevertheless be enormously valuable if it provides an institution, business or country with a capability that the incumbent system can no longer guarantee.
China appears to understand this particularly well.
Its experiments with the digital Renminbi and cross-border CBDC infrastructure illustrate an institutional willingness to treat financial architecture as something capable of being continuously engineered. Project mBridge was expressly designed to explore instant cross-border settlement through a multi-CBDC architecture.
By 2026, the platform had progressed from experiment towards meaningful transaction volumes, while China was simultaneously expanding cross-border e-CNY infrastructure.
Why This Mattered Particularly at Jackson Hole
This is ultimately why Du’s argument deserves closer examination.
Jackson Hole is not an ordinary academic conference. The Kansas City Fed describes it as a gathering of central bankers, policymakers, economists and financial-market participants confronting long-term policy questions.
Ideas expressed there influence the intellectual priors through which institutions assess change. And therein lies the danger of architecture-preservation bias.
Central bankers — particularly those governing smaller or less technologically sophisticated financial systems — could easily take from arguments such as these the proposition that today’s institutions have already internalised the relevant economics and that technological alternatives remain peripheral curiosities.
That would be precisely the wrong lesson. The responsibility of a central bank is not to preserve today’s financial architecture. It is to preserve the objectives that architecture exists to achieve: monetary stability, finality, liquidity, resilience, financial integrity, competition and trust.
The mechanisms through which those objectives are achieved must remain open to radical redesign.
The Question Central Bankers Should Be Asking
Du concludes that central banks should prioritise better netting, better access, better interoperability, better resilience and better competition rather than building new ledgers merely to accommodate the crypto ecosystem.
Few would disagree with those objectives. But that is not the end of the inquiry. It is the beginning.
Artificial intelligence, programmable money, tokenisation, autonomous economic agents and geopolitical fragmentation are not merely productivity tools operating at the margins of finance. Collectively they are capable of producing changes in commerce, capital and institutional behaviour at a pace that existing bureaucracies may find increasingly difficult to absorb or comprehend.
Central bankers therefore need to ask a more uncomfortable question: how should the economic architecture change because technology has changed? And, more importantly: is the institutional architecture of money evolving at the same speed as the technological, commercial and geopolitical environment in which money now operates?
Financial innovation should certainly be required to understand the economics of finance. But finance must equally be required to understand the economics of innovation.
The future of money will not be determined by asking whether new technologies fit comfortably inside the financial architecture inherited from the past. It will be determined by whether that architecture can evolve quickly enough to remain relevant to the world being created around it.
The author: Yadunath Bhargavan is a Lawyer & Litigator and the Founder of Access2Justice Technologies. The views expressed in this column are personal. Originally published on CNBC-TV18, 14 October 2026.