This morning we started a conversation about how digital-asset technologies could reshape Financial Market Infrastructures (FMIs).
It sounded at first like a technology question. It did not remain one for very long.
We wanted to look properly at the economics: what parts of today's trading, clearing, settlement and custody chain were at risk if securities, cash and ownership records became programmable; which functions would remain economically valuable; where current FMI revenue pools might migrate; and, ultimately, what a credible market structure could look like around 2030–35.
If digital-asset technology removes parts of today's post-trade process, how much of the FMI business does it actually remove?
The answer came from where we think any FMI analysis should start: not with blockchain, but with what the infrastructure actually does and what clients really pay for.
Take Clearstream. In 2025 its Securities Services business generated €722 million from custody, but only €154 million from settlement. It also generated a €559 million treasury result from net interest income and margin fees. (1) Settlement matters, but the economics around custody, asset control, cash and financing are much larger.
If new technology makes a settlement instruction disappear, it does not follow that it makes Clearstream disappear. It may simply remove one relatively mechanical part of what Clearstream does.
The more we looked at the existing value chain, the more vulnerable the processes appeared, while many of the underlying economic functions looked remarkably resilient.
The Bank for International Settlements (BIS) goes as far as envisaging infrastructure in which messaging, reconciliation and asset transfer are combined rather than remaining separate operations. Importantly, the BIS also says such a “unified ledger” need not necessarily use Distributed Ledger Technology (DLT) at all. (2)
That raised the next question: if cash and securities can move together automatically, does central clearing begin to lose its purpose?
Atomic settlement was the obvious place to start. If the buyer has the cash and the seller has the security, programmable infrastructure can make one transfer conditional on the other. Cash and security move together or neither moves. Much of the settlement exposure between execution and delivery can disappear.
Does atomic settlement put central clearing under threat?
Evidence from the Depository Trust & Clearing Corporation (DTCC) after the move to T+1 complicates that conclusion. Shortening settlement by a day cut the National Securities Clearing Corporation (NSCC) Clearing Fund by about 23%, yet NSCC was still taking roughly $2.2 trillion of gross daily activity and reducing settlement obligations by approximately 98%. (3)
A market maker that buys €100 million and sells €98 million does not necessarily want €198 million of gross trades to settle immediately. It may much prefer to fund a €2 million net obligation.
So the question is not whether technology allows instant settlement. Of course it does. The question is whether instant gross settlement is economically better than allowing trades to accumulate briefly, netting them, calculating the resulting risk and then settling the net obligation atomically.
We came away thinking that the institutional end-state may not be everything settles instantly at all. It may be netted programmable settlement, perhaps several times through the day: shorter periods of settlement exposure without throwing away the funding and liquidity benefits of netting.
Even if atomic settlement can remove exposure between trade and delivery, what does it do to a risk that still exists years into the future?
The distinction became clearer still with derivatives. No blockchain can atomically settle today the economic exposure of a five-year interest-rate swap with four years and 364 days left to run. Initial Margin, Variation Margin, portfolio offsets, liquidity resources, default funds and the ability to manage the failure of a large clearing member remain necessary. Technology can automate those functions. It does not remove the risk they exist to manage.
We began to see two broad categories. Some functions derive their value largely from processing and reconciling information between institutions. Others derive it from something genuinely scarce: liquidity, risk-bearing capacity, netting, balance sheet, legal authority, collateral or money.
The first group looked exposed. The second did not.
If that distinction is right, where should the next major FMI value pool appear? The answer increasingly seems to point towards collateral rather than settlement itself.
Euroclear's Collateral Highway is mobilising around €2.5 trillion of collateral every day. (4) Clearstream had €22.6 trillion of assets under custody in June and more than €1 trillion outstanding in Global Securities Financing. (5)
A large dealer's problem is rarely as simple as not enough assets. It may own billions of euros of government bonds, corporate debt, equities and cash distributed across several custodians, central securities depositories (CSDs) and jurisdictions. The difficulty is knowing which assets are free, which qualify for a particular obligation, what haircut applies, where each asset sits, what it costs economically to pledge it and whether it can get to the right place in time.
That is not principally a securities-recording problem. It is a balance-sheet problem.
So the conversation moved again: what if the real breakthrough is not putting an asset on a ledger, but making that asset immediately visible, eligible and usable wherever liquidity is required?
Now imagine those variables being readable continuously. A margin call arrives. A system checks available assets, eligibility and haircuts, compares funding costs, identifies the cheapest asset to deliver, considers the value of keeping alternatives free and then mobilises the collateral.
DTCC is explicitly moving in that direction. Its Collateral AppChain, expected to go live in the fourth quarter of 2026, is designed as shared infrastructure linking collateral providers, receivers, managers, custodians and triparty agents. Research produced by DTCC with Finadium estimates that intraday repo could cut some intraday funding costs roughly in half. (6) That is a projection rather than realised industry savings, but it shows where the commercial proposition is heading: away from simply processing collateral movements and towards optimising liquidity and capital.
Was the important innovation really that the security had been tokenised, or that a digitally issued instrument could subsequently be mobilised as collateral and turned into central-bank funding?
Clearstream provided an even more tangible example in June. The European Investment Bank (EIB) issued €77.5 million of ten-business-day commercial paper natively on Clearstream's D7 DLT platform. DekaBank and Eurex Clearing were subsequently able to mobilise the same instrument through Clearstream's triparty collateral system and the Eurosystem Collateral Management System for financing with the Bundesbank. (7)
That is where tokenisation starts to become economically interesting.
The asset was not merely issued digitally. It became usable for funding.
The future value of custody may therefore lie less simply in holding an asset and increasingly in knowing that it is available, understanding what it can be used for and being able to mobilise it when required. Custody starts to shade into asset intelligence, collateral optimisation and liquidity provision.
If tokenised assets become most useful when they connect into custody, collateral and funding networks, are incumbent FMIs really being disintermediated, or are they becoming the natural operators of the new infrastructure?
DTCC, whose Depository Trust Company (DTC) custodies more than $114 trillion of assets, used tokenised DTC-held securities in live production transactions in July. About 40 firms participated, across transactions including Treasury repo, Treasury and equity delivery-versus-payment, collateral pledges and cross-chain transfers. Its Tokenization Service is due to launch in October. (8)
Clearstream is building what it calls a digital-first, fully hybrid infrastructure spanning issuance, distribution, settlement, custody, asset servicing, liquidity and financing across traditional and digital securities. (9)
SIX has taken an equally revealing step. Rather than preserve a separate digital CSD alongside its conventional one, it obtained approval from the Swiss Financial Market Supervisory Authority (FINMA) in May to merge SIX Digital Exchange's CSD into SIX SIS and to provide crypto custody through the consolidated CSD. Traditional and digital asset services are being brought together inside the same regulated legal entity. (10)
That is hardly the behaviour of institutions preparing for the disappearance of the CSD.
It suggests that the database itself may matter less than the regulated authority around it: who recognises the asset, controls its transfer, provides legal certainty, services it and connects it to the wider financial system.
If the incumbents remain central, does the strategic advantage then belong to whoever controls the fullest vertical stack?
At first the answer on vertical integration looked straightforward. Deutsche Börse combines Eurex trading and clearing with Clearstream custody, settlement, collateral and financing. (11) SIX combines exchange and post-trade infrastructure with payments and digital-asset capabilities. (12) Euronext has been drawing trading, clearing and settlement closer together. (13) If an institution can move seamlessly from execution to risk calculation, collateral selection and settlement inside one group, surely owning the whole stack becomes more valuable.
But no large dealer lives inside one stack.
Its assets may sit at Euroclear, Clearstream, DTC, BNY and State Street. Its obligations may sit at LCH, CME, Eurex and the Fixed Income Clearing Corporation (FICC). Its cash is spread across banks and central-bank accounts, while funding prices move across repo, foreign-exchange swaps and internal treasury pricing.
Can a closed vertical really optimise a client's balance sheet when so much of that balance sheet sits outside it?
The most useful view of the client's balance sheet is therefore wider than any single vertically integrated FMI.
We moved away from the idea of a closed vertical winner. What emerged instead was something closer to an interoperable vertical: own the functions in which you genuinely have an economic advantage but make those functions capable of interacting with the rest of the market.
And even if the security leg becomes programmable, what happens to the money?
A tokenised security is of limited institutional value if the cash leg still has to navigate a separate process later. Yet central banks are not simply placing reserves onto every blockchain.
The architecture emerging from the BIS and from central-bank programmes is more pragmatic. Central-bank money remains the settlement anchor while commercial-bank money becomes increasingly programmable. In some cases central-bank money may itself be tokenised; in others, existing central-bank settlement systems can be synchronised with external ledgers so that cash and securities still move on an all-or-nothing basis. The BIS envisages a unified ledger bringing together tokenised central-bank reserves, commercial-bank money and financial assets. (2)
Australia is useful precisely because Project Acacia tested rather than merely promoted the idea. Fifty-seven industry participants explored 20 use cases across different assets and forms of public and private money. (14) The Reserve Bank of Australia (RBA) subsequently concluded that some benefits of tokenised finance could be achieved by synchronising new platforms with its existing Reserve Bank Information and Transfer System (RITS) and Fast Settlement Service infrastructure. (15)
Put those pieces together and what does a credible FMI actually look like around 2030–35?
By this point we had a fairly convincing picture.
The familiar institutions could still exist, but increasingly expose their capabilities to one another in machine-readable form. A trade could trigger a risk calculation; obligations could be netted; eligible collateral identified; the cheapest funding source selected; cash and securities reserved; and settlement executed when all the necessary conditions were satisfied.
That became our working version of a programmable FMI world.
We do not mean that every security becomes a blockchain token, or that smart contracts replace exchanges, central counterparties (CCPs), CSDs, custodians and banks. We mean that rules which today require a sequence of separate instructions between those institutions increasingly become machine-readable and executable. Infrastructure begins to behave less like a chain of organisations passing a transaction from one to the next and more like a network of functions able to react automatically to the changing state of that transaction.
If that happens, the direction of the value pool looks persuasive. Less value attaches to moving messages, matching records, repairing discrepancies and issuing settlement instructions. More attaches to liquidity, risk aggregation and netting, collateral mobilisation, funding, authoritative asset control, trusted money and the ability to coordinate the whole process.
We could have stopped there.
It would have made for an impressively bullish piece on the role of digital-asset technologies in the future of financial infrastructure.
Instead, we went looking for the evidence that challenged the argument.
How much of this new architecture has actually escaped the pilot stage?
The effect was sobering and the first number reset the mood.
The European Central Bank (ECB) estimates that tokenised assets on public blockchains reached about €38 billion globally in February 2026. Global traditional financial assets were around €241 trillion. (16)
That puts the former at roughly 0.016% of the latter.
Not 1.6%. Not 0.16%. Only 0.016%.
A market can grow very quickly from a very small base and still remain a very small market.
Europe's experience hardly points to an unstoppable migration either. The European Union DLT Pilot Regime has operated since March 2023. European Securities and Markets Authority (ESMA) reporting records six DLT market infrastructures authorised since the regime's launch. (17)
The United Kingdom supplies another reality check. Sixteen firms are preparing activity through the Bank of England and Financial Conduct Authority (FCA) Digital Securities Sandbox, including Euroclear, HSBC, JPMorgan and London Stock Exchange Group (LSEG). But the Bank's dashboard showed only HSBC as having passed Gate 2 by 13 July 2026. (18)
That does not mean the initiatives are failing. It means that moving from a convincing proof of concept to live, regulated and scalable financial infrastructure is slow.
Even where tokenisation is live, has it actually reduced the cost of running the market?
The ECB has compared tokenised bonds with matched conventional bonds. It found encouraging results, including lower borrowing costs and improved liquidity indicators. (19)
But it found no reduction in underwriting fees. Those fees were actually 0.04 percentage points higher on average for tokenised bonds, although the difference was not statistically significant. (19) The wider ECB analysis also stresses that tokenised markets remain small and still depend heavily on interaction with existing financial infrastructure. (16)(19)
That matters because so much of the tokenisation case rests on an attractive assumption: remove reconciliation and intermediaries, therefore remove cost.
Perhaps eventually.
But an institution cannot switch off its existing systems because a parallel digital market has appeared. Enough clients, assets, liquidity and counterparties have to migrate with it. Until then the firm often has to operate the new infrastructure and the old one.
Migration can create more plumbing before it removes any.
And if a token moves on a blockchain, has legal ownership of the security necessarily moved with it?
A token on a blockchain does not necessarily become the legal security or the authoritative ownership record. (20)
The U.S. Securities and Exchange Commission's (SEC) January 2026 statement distinguishes several models. In one, the crypto network can form part of the issuer's master securityholder record. In another, the security remains recorded off-chain and movement of the token merely causes that conventional master record to be updated. In the latter case, the blockchain has not replaced the authoritative database at all. It has been added to it. (20)
That punctures one of the more casual assumptions around tokenisation: that once an asset is “on chain”, ownership, custody and securities law somehow collapse into the token.
They do not.
Financial finality is not simply a computer saying that a transaction can no longer be reversed. It is also a legal determination of who owns what, whose claim has priority, what happens in insolvency and which record courts and regulators recognise.
If the future contains several ledgers and networks rather than one, does interoperability remove fragmentation, or simply move the fragmentation somewhere else?
Interoperability supplied another dose of reality.
It sounds reassuring to say that there will be several ledgers and that they will simply be made interoperable. In practice, a connection has to preserve not only data but the economic and legal meaning of whatever moves across it. And if assets and money become stranded in different networks, interoperability can end up recreating the fragmentation the technology was supposed to remove. The ECB has likewise highlighted the risk of liquidity fragmentation between traditional and tokenised environments and the importance of interoperability. (21)
What, then, does the new infrastructure actually have to beat?
The alternative against which all of this has to compete is not primitive.
DTCC's NSCC already achieves netting rates of around 98% or more across U.S. equities. (3) Modern CCPs calculate risk across enormous portfolios. CSDs process vast volumes at extremely high levels of resilience. Central banks already settle systemically important payments in real time. The legacy system is fragmented and expensive in places, but large parts of it are extraordinarily efficient.
A blockchain therefore does not win because it can perform a function.
It has to perform that function sufficiently better to justify rebuilding legal arrangements, technology, operations, liquidity pools and participant connectivity around it.
The RBA has now taken that logic a step further. Its September consultation is deliberately technology-neutral. Rather than assume the answer is a new tokenised central-bank settlement platform, it is asking how far tokenised markets can be supported through the existing RITS and Fast Settlement Service, including synchronisation between external asset platforms and central-bank reserves. The RBA says synchronisation may provide the more incremental route while decisions on tokenised reserves take longer. (15)
If synchronisation, common data and better interfaces can deliver some of the same economic outcomes, is DLT really the objective, or merely one possible technology for achieving them?
We had started by asking what tokenisation and DLT might do to financial-market infrastructure.
We ended up asking something else: what happens when financial markets become more programmable, regardless of whether DLT supplies all of the underlying technology?
If the objective is to move collateral when it is needed, perhaps a token is the best instrument. Perhaps an application programming interface into a conventional CSD achieves the same economic result.
If the objective is atomic settlement, perhaps the security and money belong on the same ledger. Or perhaps synchronisation achieves almost the same economic outcome while leaving central-bank money exactly where it is. The RBA's current work explicitly examines that latter route. (15)
If the objective is to reduce reconciliation, perhaps a shared ledger is the answer. Or perhaps common data, authoritative interfaces and better interoperability remove enough duplication that wholesale migration never pays for itself.
This is why we still think FMIs will become substantially more programmable, but we now use the word carefully.
It means a market in which the rules governing assets, risk, collateral, funding and settlement become progressively machine-readable and executable; where infrastructure can react automatically to changes in positions and obligations; and where institutions can interact without today's volume of sequential messages, duplicated records and manual intervention.
That does not mean CSDs and CCPs become redundant, banks lose control of money or tokenisation inevitably becomes the dominant form of securities ownership.
In fact, after trying quite hard to disprove the bullish case, we find the survival of the major incumbent functions more convincing than we did when we started.
Liquidity remains scarce. Netting remains valuable. Future counterparty risk still has to be managed. Collateral still has an opportunity cost. Someone must provide credit and funding. Someone must stand behind the money. Someone must determine which asset record is authoritative. Someone has to manage default when software reaches the limit of what software can resolve.
Those are economic and institutional functions, not database functions.
Our prospective 2030–35 view is therefore less spectacular than the digital-asset narrative, but probably more consequential for FMIs.
We expect the transformation to advance selectively where programmability releases something genuinely scarce: capital, liquidity, collateral, balance sheet, funding capacity or risk capacity. It should advance much more slowly where it merely replaces one already-efficient electronic process with another.
That leaves us with a different view of the value pool from the one we expected at the start.
The exposed part is not post-trade as a whole. It is the part whose economics depend on processing and reconciling information between institutions.
The stronger franchises are those built around things the market cannot manufacture cheaply: liquidity, risk mutualisation, collateral optimisation, credit, trusted money, legal control of assets, authoritative data and the ability to coordinate those resources across institutions.
If those hard-to-replicate functions survive, where might the genuinely new layer of strategic value emerge?
There may even be a new prize emerging above them: not ownership of the ledger, but the ability to see enough of the system to decide how those scarce resources should be used.
Where are the assets? Which are available? Which satisfy the obligation? Which positions offset? What is the cheapest funding source? Which form of money is available? Which settlement route delivers finality at the lowest balance-sheet cost?
An infrastructure capable of answering those questions and acting on the answers starts to look less like a settlement utility and more like a capital-allocation and orchestration layer.
That possibility looks genuinely important.
But the sobering reality check mattered. Tokenised assets remain minute beside traditional markets. (16) Regulated live adoption remains slow. (17)(18) Operational savings have not yet been demonstrated convincingly. (19) Legal records still matter. (20) Liquidity can fragment. (21) Running parallel infrastructures costs money. And some of the benefits attributed to DLT may ultimately be achievable without DLT. (2)(15)
We would therefore not describe the next decade as a transition from traditional FMI to digital-asset FMI.
We would describe it as a contest to determine which parts of capital markets are worth making programmable, which technology does that job best, and who captures the economics when they are.
We started with digital-asset technologies and expected to find a battle over ledgers.
The winners may not be the institutions that put the most assets on chain. They may be those that make liquidity, collateral, balance sheet, trusted money and risk capacity easier to use programmatically without compromising the trust that makes financial market infrastructure valuable in the first place.
References
(1) Deutsche Börse Group, Annual Report 2025, Securities Services segment. Custody €722m, settlement €154m, Treasury result €559m. Source.
(2) Bank for International Settlements, Annual Economic Report 2025, Chapter III, The next-generation monetary and financial system. Source.
(3) DTCC, 2024 Annual Report; DTCC/SIFMA/ICI, T+1 After Action Report; NSCC corporate information. Source.
(4) Euroclear, H1 2026 results. Collateral Highway mobilising €2.5tn daily. Source.
(5) Clearstream, June 2026 monthly figures. Assets under custody €22.606tn; Global Securities Financing volume outstanding €1.048tn. Source.
(6) DTCC/Finadium, Collateral Infrastructure for Tokenized Capital Markets; DTCC, Tokenized Collateral Could Unlock Billions in Capital and Transform Liquidity Management. Source.
(7) Clearstream/Deutsche Börse, EIB Issues First DLT-native Commercial Paper on Clearstream's D7 Platform, 29 June 2026. Source.
(8) DTCC, DTCC Turns Tokenization into Reality and Advances Development of New Tokenization Service. Source.
(9) Clearstream, Clearstream Unveils Next-Generation Digital Securities Infrastructure, 11 June 2026. Source.
(10) SIX, SIX Receives FINMA Approval to Merge Digital CSD into SIX SIS AG and to Provide Crypto Custody Through the Consolidated CSD, 5 May 2026. Source.
(11) Deutsche Börse Group, Markets & Services / Post-trading. Source.
(12) SIX, Annual Report 2025. Source.
(13) Euronext, Our Business and fixed-income value-chain materials. Source.
(14) Reserve Bank of Australia / Digital Finance Cooperative Research Centre, Project Acacia Final Report, May 2026. Source.
(15) Reserve Bank of Australia, The Role of RITS in Supporting Settlement in a Tokenised Ecosystem, Consultation Paper, September 2026. Source.
(16) European Central Bank, Towards an efficient and integrated digital capital market in Europe: the role of tokenisation and the Eurosystem's policy response, Macroprudential Bulletin 33, 2026. Source.
(17) European Securities and Markets Authority, DLT Pilot Regime reporting and Trends, Risks and Vulnerabilities No. 1, 2026. Source.
(18) Bank of England / Financial Conduct Authority, Digital Securities Sandbox materials and dashboard. Source.
(19) European Central Bank, Tokenised bonds: assessing efficiency and liquidity in a nascent market, Macroprudential Bulletin 33, 2026. Source.
(20) U.S. Securities and Exchange Commission, Statement on Tokenized Securities, 28 January 2026. Source.
(21) European Central Bank, Central banks on-chain, August 2026. Source.