Short reads from the people who ran the mandate.

Where it starts: clearing and settlement

Every operations career in capital markets starts the same way, learning that a trade isn't done when it's executed, it's done when it settles. T+2 became T+1, netting at NSCC compresses thousands of offsetting trades into a single position, and the daily discipline is simple to state and hard to live: check that everything actually settles, and if it doesn't, find out why before it becomes someone else's problem tomorrow.

This is where you learn the instinct that defines the whole career: the platform is built for the trade that behaves. The one that doesn't, the mismatched instruction, the late confirmation, the broker who reports a position differently than the custodian, becomes a manual exception the moment it deviates from the expected path. Nobody designed it that way on purpose. It is just where the product roadmap stopped and a person started.

The first real gap: domestic proxy

Move from settlement into shareholder communications and the same pattern reappears at a larger scale. A single US proxy vote travels from the issuer's DEF 14A filing to the transfer agent, to DTC's systems, to Broadridge, which ingests and redistributes the agenda to roughly ninety-seven percent of beneficial holders, to the broker-dealer, which transforms it again for its own client interface, before it finally reaches the investor. Four transformation steps for one shareholder communication, each one a fresh opportunity for the instruction to drift from what the issuer actually filed.

This is where NOBO and OBO classification, non objecting and objecting beneficial owners, first taught me that the system wasn't built to see the actual investor at all. The issuer doesn't know who its shareholders are. It knows who its broker-dealers are, and trusts that the cascade downstream reconstructs the rest correctly. Most of the time it does. When it doesn't, the vote is wrong, and nobody upstream necessarily finds out.

The same gap, multiplied: global proxy

Then the scope widens again, and this is where it stops being one chokepoint and becomes more than a hundred of them. Managing shareholder communications and proxy voting internationally means Switzerland's filing windows, Japan's record date logic, Brazil's notarised power of attorney requirements, and India's NSDL and CDSL e voting timelines, all live simultaneously, all subject to change without a central notice, none of them written down in a single place a new team member could simply read.

Almost none of this lives in any product. The platforms were built to handle the common case, the standard deadline, the expected format. Everything outside that, which in a hundred plus markets is most of it, gets absorbed by operations. We built the workarounds. We tracked which countries needed physical signatures. We remembered which custodian in which market still required a fax, an actual fax, well into this decade. None of it showed up on a roadmap, because from the product team's perspective, the job looked done.

The throughline

What connects clearing and settlement, domestic proxy, and global proxy isn't three different problems. It is the same one, appearing at a larger scale each time the aperture widens: the infrastructure handles the expected case, and a person becomes the seam for everything else.

The reason that seam matters is that it is holding up something real, actual capital, actual ownership, actual votes on actual corporate decisions. When it fails, a shareholder gets disenfranchised or a position settles wrong, not because anyone was careless, but because the person holding the exception in their head made a mistake a machine would not have made, or left the firm before writing it down.

What would actually change this is not less complexity. Every jurisdiction will keep doing its own thing regardless of what infrastructure sits underneath it. What changes it is encoding the deadline, the notarisation rule, and the settlement logic directly into the transaction itself, so the system checks the current rule at execution instead of relying on someone remembering it correctly at seven in the morning. That is not a smaller job. It is a different one, and it is overdue.

Clearing and settlement, compensating from day one

The first thing you learn in capital markets operations is that a trade isn't finished when it executes, it is finished when it settles, and the daily job is catching everything that will not settle cleanly before it becomes a failure. What I didn't understand for years is that this daily catch isn't a skill worth admiring. It is a symptom.

Every mismatched instruction and late confirmation that a person has to manually chase down is a gap the underlying infrastructure left open, one that a batch processing, overnight reconciliation model was never designed to close in real time. I got good at catching those gaps. Getting good at it was never the point. The gaps should not have been there.

Here is what removes an entire class of these failures, not theoretically, but in production today: atomic settlement, where the security and the cash move in one indivisible transaction instead of two separate legs on two separate clocks. There is no timing window for one side to settle and the other to lag, so that particular failure mode disappears instead of being chased by a person. Not a faster version of my old job. A version of that job that goes away for the risks this structure eliminates.

Domestic proxy, a chokepoint dressed up as a process

US shareholder communications route from the issuer's filing through the transfer agent, through DTC's proxy infrastructure, through Broadridge, which redistributes to roughly ninety seven percent of beneficial holders, through the broker dealer, which transforms it again, before a beneficial holder ever sees it. That is four structural transformation points for a single instruction, and an entire commercial layer built on top of the fact that nobody fixed the underlying fragmentation problem.

I spent years getting skilled at catching where that chain introduced errors. I was rewarded for absorbing a structural cost that should have been engineered away decades ago, and I mistook the reward for evidence that the work was valuable rather than evidence that the industry had found it cheaper to keep paying people than to fix the architecture.

Here is what removes that specific pattern: the issuer publishes the agenda once, as an on chain event at filing, and every holder receives it directly with a cryptographic signature recorded immutably. One filing. One source of truth. Zero downstream re keying, because there is no chain of reformatting left for the instruction to drift across.

Global proxy, the same failure, just harder to see

Internationally, it gets worse, not because the problem changes, but because there is no shared rulebook to hide the gap behind. A hundred plus markets, each with its own deadline, its own notarisation requirement, its own record date logic, none of it centrally documented, all of it changing without notice. I held it in my head. I built spreadsheets nobody asked for. I remembered which custodian in which country still needed an actual fax machine.

None of that made the system structurally better. It made me a single point of failure that happened to be reliable, until the day I wasn't, or the day I left, and whoever came next inherited a job with no instructions, because the instructions were never the point, my memory was the product.

Here is what removes that: the deadline, the notarisation rule, and the record date logic for every market live inside the transaction itself, checked automatically at the moment of execution, not inside one person's head, checked only when that person happens to remember. The rule still gets enforced correctly the day after I leave, which is the entire test my old job never had to pass, because there was always someone there catching it manually, until, eventually, there wasn't.

What this actually is

I don't tell this story anymore as a tribute to operational grit, because it isn't one. It is a description of a tax the industry has been paying for decades, in headcount, in error rates, in shareholders who lose a vote because a notarisation deadline got missed three hops upstream of anyone who could have caught it.

The skill I built wasn't infrastructure. It was a workaround for the absence of infrastructure, and workarounds do not scale, they just get more expensive and more fragile the longer nobody replaces them.

The case for tokenization is not abstract, and it is not that it makes my old job slightly easier. Every one of the three gaps above already has a specific, working answer: atomic settlement, single source on chain filings, embedded jurisdictional logic. None of them require a person to hold the exception in their head. That is not a smaller job. It is the job disappearing because the failure mode it existed to catch disappeared first. I would rather see that happen than get better at absorbing it.

Ask anyone in capital markets what post trade infrastructure costs the industry annually and you will get a confident number back, fifteen to twenty billion dollars. It is the figure that shows up in conference keynotes, vendor pitch decks, and regulatory testimony. It is also, by a conservative reconstruction of where the money actually goes, somewhere between roughly fifteen and twenty five times too small.

What the headline number actually measures

The fifteen to twenty billion figure covers exactly one thing: direct settlement, custody, clearing, and transfer agency fees for exchange traded securities, as they are typically reported. That is it. It is a real number, and it is not wrong. It is just answering a much narrower question than the one people think it is answering.

The layers nobody puts in one stack

Rebuild the cost stack layer by layer and a different picture appears.

1. Layer one is the direct infrastructure fees everyone already counts, plus custody and safekeeping fully loaded, which pushes the figure into the low hundreds of billions when you treat global custody and related charges as part of post trade, not as a separate service.

2. Layer two is the global operations workforce doing reconciliation, exception handling, and corporate actions processing, multiplied across thousands of institutions. Easily tens of billions, and plausibly into the low hundreds.

3. Layer three is the proprietary technology every institution builds and maintains separately to bridge gaps in the underlying infrastructure, from in house matching engines to bespoke corporate action workflows.

4. Layer four is error and litigation exposure from settlement fails and processing mistakes. Not just direct losses, but legal fees and increased capital or margin held against operational risk.

5. Layer five is regulatory penalties and compliance screening, much of it duplicated across institutions checking the same transactions independently to satisfy similar rules.

6. Layer six is the entire over the counter and non exchange traded market, which never touches the infrastructure the headline number describes at all, but carries equivalent or higher operational complexity.

7. Layer seven is the data vendor subscriptions every institution pays independently for information that is frequently identical, just piped differently into internal systems.

Add it up illustratively, not to claim false precision but to show scale, and the total lands in a range between roughly $345 and $541 billion a year when you include all seven layers, not just the first one. Even the low end is well over fifteen times the headline. That is not a rounding error. It is a fundamentally different scale of problem.

Why the gap is the point, not a footnote

None of this is a claim that the industry is lying about its costs. It is a claim that everyone has been measuring the visible top slice of an iceberg and reporting it as the whole thing, because the rest was absorbed quietly by operations teams, buried in technology budgets, or scattered across markets that do not touch the standard reporting categories.

The real number is not just bigger. It is the largest single structural cost reduction opportunity in financial services, and it has been hiding in plain sight because nobody added the layers together in one place before.

What matters for anyone reading this number

If your institution is sizing the opportunity in tokenization, agentic automation, or infrastructure modernisation using the fifteen to twenty billion figure as your total addressable cost, you are underestimating by at least an order of magnitude. The number that matters for a business case is not the one that is easiest to cite. It is the one that actually adds up when you count everything operations and technology currently do to compensate for the gaps.

What tokenization actually is, briefly

Strip away the noise and tokenization is a simple idea: represent ownership of a real asset, a bond, a share, a fund unit, as a digital token on a shared ledger, with the legal and economic rights of that asset attached to the token itself. Pair it with a regulated stablecoin or tokenised deposit as the cash leg and a trade can settle atomically, the asset and the payment move in one indivisible transaction instead of two separate legs that can fail independently. That is the mechanism. Everything else is implementation detail.

The adaptation nobody predicted

The popular narrative around blockchain in finance has always been disruption from outside. New entrants building parallel rails, incumbents caught flat footed, market share migrating to whoever moves fastest. That is not what is happening.

What is happening is that the incumbents are building the infrastructure themselves, in parallel, at the same time, without waiting for a single winner to emerge.

DTCC, custodian of over $114 trillion in assets, is running a phased rollout of its own tokenization service, backed by SEC staff no action relief and a working group of more than fifty firms. Broadridge's Distributed Ledger Repo platform is processing $357 billion a day in tokenized repo transactions, up 68 percent year over year, with data flowing into Bloomberg. Goldman Sachs has had its own tokenization platform, GS DAP, in production since 2022, settling bond issuances at same day cycles. Nasdaq and NYSE are each building tokenized equity frameworks. And in July 2026, SWIFT, the messaging network at the center of over eleven thousand financial institutions, moved a blockchain based shared ledger for tokenized deposits into an initial live pilot with seventeen banks.

None of these institutions waited for someone else to build the winning platform. Each built its own, on its own timeline, for its own part of the market. That is not disruption. That is every major incumbent independently concluding the same thing at once.

Why no single chain wins

The natural next question is which of these platforms ultimately wins. It is the wrong question.

Look at the global regulatory landscape and the end state is not one blockchain. It is many, permanently. The United States, the European Union, the UK, India, the BRICS bloc, and the Gulf states are each building infrastructure aligned with their own regulatory and geopolitical interests, and there is no credible path to consolidating them onto a single chain. Data sovereignty laws alone guarantee permanent fragmentation.

What resolves this is not consolidation. It is a common language above the chains. ISO 20022, the messaging standard SWIFT and central banks globally are converging on, functions like TCP/IP for the internet. It does not care which network carries the packets, it only makes sure every network can talk to every other one. Cross chain interoperability protocols, including the one SWIFT and DTCC have selected for joint experiments, are emerging as the connective layers that let value move between chains that will never merge into one.

The takeaway

If you are evaluating this space by asking which platform to bet on, you are playing the wrong game. The winning move is not picking the chain. It is building for a world of permanent multiplicity, where the institutions that matter are the ones that can move value across every chain, not the ones that own the single chain everyone else has to use.

It is tempting, looking back, to ask why the institutions with the deepest pockets, the most talent, and thirty years of watching the same reconciliation failures repeat every morning did not simply build the structural fix themselves. Today's self styled fintech disruptors are, in most cases, younger, smaller, and less resourced than the incumbents whose inefficiencies they are now addressing. The obvious question is why the incumbents did not get there first.

The answer is not that they failed to see the problem. It is that they were responding rationally to the incentives in front of them, and those incentives did not reward the fix.

The logic that made sense at the time

Building the structural solution to reconciliation, settlement friction, and duplicated infrastructure would have meant cannibalising the fee income that friction generates. No management team accountable to quarterly earnings invests in eliminating its own revenue line, not because the people involved lacked vision, but because fiduciary duty to current shareholders does not reward an investment that pays off over a decade at the direct expense of this quarter's numbers.

In 2010, and even in 2015, treating this as a distant, low priority bet was a reasonable read of the landscape. The technology was not mature, the regulatory path was not clear, and the near term cost of building it dwarfed the near term benefit.

Where the calculation stopped being rational

What changed is that the timeline kept slipping while the technology kept maturing. A reasonable bet in 2010 became a harder case to defend by 2020, once the building blocks, distributed ledgers, smart contracts, regulated stablecoins, were demonstrably production ready elsewhere in finance. By 2023, the absence of a serious internal response was less a strategic choice and more a missed window.

Today, firms that spent a decade treating this as someone else's problem now find themselves funding, and in some cases co governing, infrastructure built by companies that did not exist when the decision not to build was first made.

The pivot that is already happening

This is not, in the end, a story about institutions being replaced. It is a story about institutions changing role.

Once it became clear that foundational infrastructure was being built with or without them, incumbents did not compete against it or ignore it. They joined it: working groups, funding rounds, board seats, node operation. DTCC's own tokenization working group includes more than fifty firms spanning both sides of this divide. That is not capitulation. It is the same institutions finding a different way to stay at the center of a market they no longer fully control the architecture of.

What this means going forward

The lesson here is not about any specific firm's judgment ten years ago. It is about what the same incentive structure will do to the next transition, and the one after that.

The firms reading this today face an identical version of the same choice: build now, at a cost that shows up this quarter, or wait, and eventually engage on someone else's terms. History suggests which of those two paths gets chosen by default. It does not have to be this time.

The misconception, stated plainly

Scan enough tokenization commentary and you will see a recurring claim: full tokenization of public securities is waiting on Congress to amend the relevant law. It is a clean, quotable line. It is also wrong, and the error matters, because it points everyone toward watching the wrong body entirely.

What Article 8 actually is

UCC Article 8 is state law. It is part of the Uniform Commercial Code, a model statute drafted jointly by the Uniform Law Commission and the American Law Institute, then separately enacted, state by state, by each individual state legislature. Congress has never had jurisdiction over it and does not now.

Article 8 is the law that currently defines how investment securities are legally held, and it is why, today, the vast majority of publicly traded US securities are registered not in the name of the investor who owns them, but in the name of Cede & Co., a nominee entity, with individual investors holding what the law calls a security entitlement rather than direct legal title.

What already changed, and what did not

Here is where the picture has genuinely moved, and where a lot of coverage gets confused.

In 2022, the Uniform Law Commission and American Law Institute approved a significant package of UCC amendments, including an entirely new Article 12 covering digital assets, cryptocurrencies, NFTs, and similar instruments. That package has since been enacted in a large number of states, including New York, effective June 2026. That is real, substantial legislative movement, and it is reasonable to assume it resolves the tokenization question.

It does not, and the text of Article 12 says so explicitly. Article 12 carves out investment property, meaning securities, security entitlements, and securities accounts, from its own scope. It was built to give legal clarity to disintermediated, crypto native assets. It was not built to touch the intermediated holding system that public equities and bonds still operate under.

The SEC staff's no action letter authorizing DTC's tokenization pilot confirms this directly: the pilot does not alter the existing indirect holding model, and registered ownership stays with Cede & Co. as nominee, unchanged.

What is actually still needed

The specific reform that would let a public company's shareholders hold direct legal title on chain, with no nominee in between, is a change to Part 5 of Article 8, the provisions governing security entitlements and indirect holding. That is a different, narrower, and more consequential change than anything in the 2022 amendments, and as of today it has not been formally proposed in any state.

Why this matters for anyone tracking the timeline

If you are modeling when full tokenization arrives, the question to track is not a federal bill. There is not one. The question is whether any state begins the Uniform Law Commission process for Article 8 Part 5 specifically. That process, historically, takes years even once it starts, and it has not started.

Everything currently live, DTCC's pilot included, operates entirely within the existing legal structure, tokens as a new way to instruct the same old ownership record, not a new record itself. That is not a criticism of the pace of progress. It is just where the actual bottleneck sits, and it is not in Washington.

The four phases, honestly stated

Tokenization's path to full maturity runs in four phases.

Phase one, tokenized money market funds and repo, is live today, and it is real: BlackRock's BUIDL, JPMorgan's Kinexys, and Broadridge's Distributed Ledger Repo are all in production, moving real volume. Phase two, tokenized equities and fixed income at the infrastructure level, is beginning now with DTCC's phased rollout through 2026, backed by SEC staff no action relief. Phases three and four, migrating the legal register on chain and ultimately retiring the nominee structure entirely, require a change to state law that, as things stand, has not been proposed anywhere.

That is not pessimism. It is just the honest shape of the timeline. And it means something specific and important: phases one and two will run in parallel with legacy settlement infrastructure for years, potentially into the next decade. Every institution operating through that period maintains two full processes, reconciliation, corporate actions, compliance reporting, on both rails simultaneously.

Why the parallel running period is the real cost, not a footnote

This is the part that gets underweighted in most tokenization commentary, which tends to talk about the future state and skip the transition.

Running two infrastructures in parallel is not a minor overhead. It is arguably the most expensive phase of the entire transition, because it requires the full cost of the legacy system plus the incremental cost of standing up and reconciling against the new one, for as long as the parallel period lasts. Almost nobody's business case models this properly, because everyone is modeling the destination, not the multi year corridor to get there.

The business that exists regardless of the timeline

Here is the reframe that matters: the reconciliation break between an on chain record and a legacy record does not wait for state legislatures to act. It happens this week, and someone has to resolve it, regardless of whether Phase 3 arrives in 2029 or 2035.

A managed services layer built specifically to sit between incumbent batch infrastructure and the tokenized end state, cloud native, augmented with agentic AI to absorb the exception handling volume that currently consumes operations headcount, is not a bet on a future regulatory outcome. It is infrastructure the industry needs today, funded by a cost that already exists, independent of when or whether the final legal destination gets confirmed.

Who this is actually for

This is not a pitch aimed at institutions with the capital to build proprietary tokenization infrastructure from scratch. Firms like that are already in DTCC's working group or running their own pilots.

It is aimed at the much larger population of mid sized broker dealers and asset managers who have neither the capital to build it themselves nor the internal scale to absorb the transition through added headcount. That space is large, underserved, and, critically, it does not require a single piece of state legislation to pass before it starts paying off.

The point worth sitting with

Every conversation about tokenization eventually arrives at the same question: when does the legal reform happen, when does the parallel running period end. Nobody knows, and treating the uncertainty with real humility is the intellectually honest position.

But that question, the one everyone is stuck on, is not the one that determines whether this business is worth building today. The operational cost of running two systems at once is being paid right now, by every institution in this transition, regardless of when the end state arrives. The service that resolves it does not need the destination confirmed. It needs to exist. It already does.

The wall, not the person

Every institution has a last mile. It is the point where a rule, once written down, has to be applied to a specific fact, in real time, by someone who can only see part of the picture. For thirty years, operations has been the shock absorber for that gap. When the architecture underneath does not hold, a person holds it instead, manually, under pressure, at volume, until they can not.

A recent OFSI penalty against a major bank's London branch is what it looks like when that stops working. Not one failure. Eleven of them, each one a version of the same thing. A fact existed somewhere in the bank, and the person who needed it at the last mile could not see it. That is the wall. Not a metaphor for effort or diligence, operations in this case did plenty of both. The wall is structural. It is the point in the process where visibility simply ends, and everything past that point runs on inference, memory, or hope.

Two walls, not one

The first wall is the vendor wall, a KYC record and a sanctions list holding two different spellings of the same entity, with nobody positioned to see they matched. The second wall is the structural wall, an ownership determination made in good faith on the information available, later judged unreasonable, because true beneficial ownership was never visible anywhere the bank could subscribe to or query. Both walls produced the same downstream symptom, a person made a call with a partial view and the call was wrong. But they are not the same problem, and treating them as the same problem is exactly how "process failure" becomes the label that lets the real cause hide for another cycle.

Why this is not rare

The pattern underneath this case is close to the default state of most operational software built under commercial pressure. A product ships the clean case, and the edge cases, the offshore structure, the transliterated name, the licence that only applies under specific facts, get handed to a person with a login. That person's accumulated judgment becomes the real system. Nobody planned this on purpose. Building the edge case into the architecture at design time is expensive, and a manual workaround costs nothing today. It only costs something later, when the person who held it moves on, or the edge case turns out to be the one a regulator cares about most.

The error was never the last mile

Eleven failures added up to a number, and that number carries the weight to end an institution that gets it wrong twice. A monetary exposure of this size belongs on a board agenda, mapped, measured, and assigned to an owner, not absorbed into a headcount conversation nobody wants to have twice. The fix is a system that makes every gap visible as a metric, tracked against a name, reviewed on a schedule that does not wait for a regulator to set it. The technology to do this exists today. What is missing is not capability, it is the willingness of leadership to trade a familiar architecture for one that admits, in writing, exactly where the old one falls short.

One agenda, two pipelines

Every proxy season starts with a single filing, an issuer's agenda submitted to the SEC. From there the path a shareholder's vote takes depends entirely on how their shares are held, not on the company, not on position size. A registered shareholder has a direct line to the transfer agent. A beneficial owner's vote instead travels through DTC and its nominee Cede & Co., a broker dealer, often a sub custodian or global custodian running in parallel, and a proxy service or solicitation firm such as ProxyEdge, before it ever reaches a ballot.

Where the friction actually lives

Talk to anyone who has run a proxy season from the operations side and the same issues surface every time. Agenda data gets re keyed by hand rather than consumed as structured data. Every additional intermediary adds a handoff and a delay. Issuers frequently cannot see who actually holds their shares. Multiple notice and ballot formats proliferate across the chain. None of this is any one participant's failure, it is an infrastructure that has been incrementally adapted for decades rather than rebuilt as a single interoperable workflow.

A more machine-readable future state

A canonical notice layer, structured agenda data that every participant references instead of copying and re keying at each stop, is the credible near term direction. ISIN and CUSIP remain the reference identifiers underneath it, and ISO 20022 and related market infrastructure messaging standards could inform cross border interoperability, though proxy specific data models and governance would still be required. Tokenization, used well, is a verification layer on top of that structure, not a replacement for the transfer agents, brokers and custodians who keep their recordkeeping, compliance and custody functions in this model.

What changes for each participant

The transfer agent's role transforms rather than disappears, shifting from keying the agenda by hand toward overseeing entitlement accuracy and exceptions. Brokers stay essential for servicing, custody and compliance, even as their re keying workload shrinks and commercial models evolve accordingly. A new role opens for token and entitlement management, and for the human reviewers who validate extraction, resolve mismatches and audit the model, reskilling toward work like blockchain contract review rather than manual data entry.

The honest caveat

None of this is a near term full replacement of today's system. The credible path is hybrid and staged, structured data first, entitlement validation next, tokenized voting only where regulation permits it, with paper and assisted channels available at every stage. The transfer agent does not disappear under tokenization. It moves up a layer.

The bigger lever

Proxy voting is the visible example here, not the whole story. The same canonical ownership record that fixes it also fixes dividend and distribution notices, corporate actions like splits and tender offers, and tax and cost basis reporting, since all four read the same underlying record instead of solving the same reconciliation problem four times over. That is a security level fix, not a proxy specific one, and it stops there. Broader investor communications, earnings, analyst relations, roadshows, sits outside it, none of that depends on who currently owns the security, so this argument does not reach that far.