Most of what is claimed for tokenisation is claimed for the wrong thing. The token is not the gain, the conditionality is, and almost everything that has failed in this field failed for the same reason.
A tokenised bond is a bond.
That sentence sounds like a deflation and it is meant as a foundation. Tokenisation does not create an asset. It creates a second way of holding a claim that already exists, with the same issuer, the same coupon, the same maturity and the same legal standing as it had in the register.
If that is true, and we think it plainly is, then most of what gets promised for tokenisation is being promised for the wrong thing.
It changes where the record lives and what the record can be made to do.
A holding in a register is a line in a book that somebody updates when instructed. A holding on a programmable ledger is a line that can carry conditions: it can be made to move only when something else moves, only to counterparties on a permitted list, only within limits set by whoever is entitled to set them.
That is the entire technical difference. Everything else claimed for it is a consequence of that, or is not true.
The rights do not change. The obligor does not change. The security's position in an insolvency does not change, and nor should it, because the thing that gives a bond its value is a promise from an issuer, not the medium in which the holding is recorded.
Treating a tokenised instrument as a distinct asset class produces two errors at once. It invites a separate risk framework for something whose risk has not moved, and it invites a separate market for something whose liquidity should not be split.
A representation is worth only as much as the places that accept it. Build a market that only accepts your token and you have built a demonstration.
Here is the gain, stated as narrowly as we can make it.
In a conventional chain, delivery and payment are coordinated by institutions that hold both legs and release them together. It works, it has worked for decades, and it is one of the reasons the settlement layer is as robust as it is. What it costs is time and intermediation: the coordination has to be performed by somebody, in a process, in a window.
On a programmable ledger the condition can be a property of the instrument rather than a task performed on its behalf.
The gain is not the token. The gain is that the instrument can be made to move only when the money moves.
Which is why tokenisation without a settlement asset on the same terms is a half-built idea. An instrument that can enforce its own delivery condition, paired with a payment leg that still has to be chased through a correspondent chain, has automated one side of a two-sided problem.
That phrase appears on our own site and it is not hedging.
Atomicity is a property of software: both legs occur or neither does. Whether a given transfer may be executed that way is a question of permissions, of which settlement asset is available to the participants, and of the rules of the system in which finality is conferred. Those are legal and policy constraints. They are not solved by better engineering, and a venue that implies otherwise is misdescribing the problem.
So the honest formulation is that atomic settlement is available where the law and the permissions allow it, and coordinated settlement is what happens everywhere else. Both need to work. Designing only for the first is designing for a market that does not exist yet.
If we had to defend tokenisation on one use rather than in general, it would be this one.
Collateral spends most of its life immobile: pledged in one place, needed in another, moved by instruction through a chain that takes long enough that desks hold buffers to cover the gap. The buffer is the cost, and it is paid every day whether or not anything moves.
An instrument that can be transferred under conditions, with eligibility rules travelling attached to it rather than checked separately at each stop, reduces that. Not because the token is fast, but because the eligibility question and the movement question stop being two processes.
The qualifier matters as much as the claim: eligibility criteria are set by risk functions, central counterparties and regulators, and encoding them does not mean choosing them. Any design that lets a movement escape its guardrail has made things worse, not faster.
This is the part of the argument that constrains us rather than anyone else.
A tokenised venue with its own instruments, its own participants and its own settlement arrangement is a closed system. It can demonstrate every property in this piece and still be irrelevant, because the value of a representation is entirely a function of where it is accepted. If a desk cannot move between the tokenised form and the book entry form without leaving the market, the tokenised form is a silo with better software.
That is why our position is that this has to be an extension of existing market infrastructure rather than a parallel one: one engine, two formats, both settling into the same place. It is a harder thing to build than a separate venue and a slower thing to launch, and we think the alternative is a pilot that never carries volume.
Three things, and we would rather name them than be asked.
The legal treatment is not uniform. Whether a tokenised holding is recognised as the holding, or as a record of one, varies by jurisdiction, and cross-border cases are where that bites hardest.
Settlement in central bank money on tokenised terms is being actively worked on by central banks and is not generally available today. Until it is, a tokenised delivery leg will often be pairing with a payment leg that carries commercial bank credit, which is a real improvement over the current arrangement and is not the endpoint.
And the operational chain is unchanged by any of this if the surrounding processes stay the same. A token that arrives instantly into a workflow that still reconciles overnight has moved the delay rather than removed it. Reconciliation is the tell, and it will still be the tell after the instrument is tokenised.
None of that argues against doing it. It argues against describing it as finished.