Why You Cannot Build Programmable Finance on Non-Programmable Money

Three years of tokenization pilots have proven the industry can put assets on-chain. Bonds, fund shares, real estate, commodities — the asset side works. The problem is that none of it delivers the composability benefits that justified the effort, because the cash leg still settles on legacy rails. You cannot build programmable finance on non-programmable money.

This is not a refinement of the tokenization thesis. It is an argument that the industry has the sequence backwards.

What “programmable money” actually requires

Programmable money is a monetary instrument that can participate natively in conditional, multi-step financial logic — receiving, releasing, and re-routing funds based on verified on-chain states, without manual intervention at each step. The industry has used the term loosely; the technical requirement is more specific than the marketing language suggests.

The BIS defines the precondition precisely. In the 2023 Annual Economic Report, it states that “the full potential of tokenisation needs a monetary unit of account that denominates transactions, as well as the accompanying means of payment.” A unit of account and a means of payment. Both. On the same programmable platform as the assets being settled. When only the asset is on-chain, the programmable platform has nothing to work with on the cash side. Conditional execution requires both legs to be programmable, or the condition cannot fire atomically.

The BIS returned to this framework two years later. The 2025 Annual Economic Report identifies a specific sequence: “tokenised central bank reserves, tokenised commercial bank money and tokenised government bonds, all residing on a unified ledger.” Money first, then assets. The sequence is not interchangeable.

Why stablecoins do not solve this

Stablecoin settlements now run into the tens of trillions annually, according to McKinsey’s July 2025 analysis. So what is missing?

Three things, identified directly by the BIS and ECB. The BIS June 2025 press release states that stablecoins lack singleness, elasticity, and integrity. Singleness means that one unit of the instrument always exchanges at par with all other forms of the same money. ECB Executive Board member Piero Cipollone made this concrete in a March 2026 speech: “Even fiat-backed stablecoins rarely trade exactly at par, even during calm market conditions.” That position is reinforced by the ECB’s April 2026 Macroprudential Bulletin: “at-par convertibility in transactions can only be ensured if central bank money is also made available on chain.”

At-par is not a nice-to-have for institutional settlement. It is the minimum requirement. Settlement at unknown exchange rates introduces a basis risk that no treasury mandate accommodates. The BIS Bulletin 73 is direct: stablecoins as bearer instruments may “entail departures in their relative exchange values away from par in violation of the ‘singleness of money’.” Any monetary layer that fails the singleness test cannot be a reliable foundation for programmable settlement. The instrument is not interchangeable with the function it claims to fill.

There is a sharper structural problem. The Federal Reserve’s December 2025 FEDS Notes on stablecoins observe that “each dollar of deposit withdrawal can lead to a more-than-one-dollar contraction in lending.” Stablecoins as a monetary layer do not only fail the singleness test. At scale, they degrade the credit system that funds the real-world assets being tokenized in the first place.

What the asset-first sequence actually produces

Siemens issued a €60 million bond on Polygon in February 2023, bypassing Clearstream and Euroclear entirely. A genuinely tokenized asset. The announcement quoted the efficiency gains. The settlement documentation was quieter: “payments were made using classic methods as the digital euro was not yet available at the time of the transaction.” Settlement took T+2. The same as traditional rails.

The asset was on-chain; the money was not. Atomic settlement was impossible because only one side of the trade was programmable. The headline said blockchain bond. Settlement said nothing changed.

Siemens issued a €300 million digital bond in late 2024, an improved architecture. This time, settlement used the Bundesbank’s trigger solution: a bridge that detects a blockchain event and initiates a traditional central bank payment in response. The reported structure approximates atomic settlement but cannot deliver it. The trigger fires, the legacy payment begins, and there is a window between those two events. That window is not atomic. The architecture strain is legible: one side programmable, one side not, and an engineering workaround in between.

The Swift-UBS-Chainlink pilot from November 2024 makes the strongest case for bridges as a long-term answer. Swift’s framing of the pilot was that it achieved straight-through processing of the payment leg without requiring global adoption of an on-chain form of payment. This is a real capability for simple one-leg transfers. Where it breaks down is multi-step financial logic. Simultaneous collateral liquidation, position entry, and yield re-routing require all three legs to be atomic. A bridge can initiate one payment. It cannot bundle three interdependent actions into a single executable package. The composability story requires both sides of every leg to be on the same programmable platform.

Why composability collapses without the monetary layer

The word “composability” is used often in tokenization discussions and rarely defined. In a programmable financial system, composability means that complex multi-step transactions can be assembled from simpler operations and executed as a single atomic unit. Borrow against collateral and simultaneously enter a new position. Escrowed payment releases when delivery is confirmed, and yield routes automatically to beneficial owners. These are not sequential operations; they are one operation.

The BIS stated the requirement clearly in the 2023 Annual Economic Report: “By dispensing with messaging and the reliance on account managers to update records, it provides greater scope for composability, whereby several actions are bundled into one executable package.” Bundling into one executable package requires that all components of the bundle are on the same programmable platform. A bundle where one leg is on-chain and one leg is a Swift message cannot be atomic. A Swift message introduces latency and counterparty dependency that breaks the atomic guarantee.

BIS Head of Research Hyun Song Shin made the venue requirement explicit in a June 2023 speech: “a unified ledger that provides tokenised central bank digital currencies in the same venue as tokenised deposits and other tokenised claims is the most promising way to harness the potential of tokenisation.” Same venue. Not a bridge between two venues. One venue where money and assets coexist under the same programmable rules.

This is why the BIS is running Project Agorá, with seven central banks and over 40 private institutions. The project exists because the world’s central banking infrastructure recognizes that tokenizing assets without a programmable monetary layer does not solve settlement finality. The project’s stated goal is “atomic transactions — payments completed synchronously and in full.” Synchronously and in full. That is the standard a bridge cannot meet.

Where the counterargument has merit

A wholesale CBDC is not strictly required for programmable settlement. The Federal Reserve’s September 2023 FEDS Notes argues that existing reserves can serve on tokenized platforms — a technically narrow point that is probably right. The dispute is about form, not function: programmable monetary infrastructure is still required regardless of its legal instrument.

Existing reserves can serve the role, but they must be on the programmable platform where assets settle. The presence of programmable money is the requirement. Its precise legal form is a design choice.

The ECB’s position, expressed in April 2026, is unambiguous on this: “tokenised central bank money is critical to the success of an integrated European market for digital assets.” Off-ramps and bridges “introduce costs and disruptions.” The ECB is not describing a theoretical ideal. It is describing a constraint that currently limits every tokenized asset market in Europe.

What JPMorgan already understood

JPMorgan did not wait for this argument to be settled. In November 2025, JPMorgan launched JPMD, a deposit token on Base, Coinbase’s Ethereum L2. The bank positioned it as an on-chain, bank-backed alternative to stablecoins, combining banking integration, deposit-backed settlement, and on-chain programmability. JPMorgan chose not to use existing stablecoins because they lack banking integration and yield. The world’s largest bank built the monetary layer from scratch because it determined that programmable settlement cannot happen natively on-chain without it.

That decision is the thesis made concrete. JPMorgan’s engineers arrived at the same conclusion the BIS research reaches from first principles: the money layer is the constraint. Everything else in programmable finance is downstream of it. FractiFi builds the same monetary layer for institutions that are not JPMorgan, with tokenized deposit rails that enforce a 1:1 reserve invariant at the protocol level rather than through issuer trust.

The sequence that works

The BIS 2025 Annual Economic Report describes atomic settlement as “the joint execution of three previously separate steps: the debiting of the payer’s account, the crediting of the receiver’s account and settlement on the central bank balance sheet.” Three steps, executed jointly. Not sequentially. Not with bridges between them. Jointly, on the same platform, in a single operation.

That capability does not exist when the asset is on-chain and the money is not. Tokenizing bonds, fund shares, and real estate produces better record-keeping and some operational efficiency. It does not produce programmable finance. Programmable finance requires that both sides of every transaction are subject to the same programmable rules, on the same platform, at the same time.

The ECB’s Cipollone named the consequence of getting the sequence wrong: “without tokenised central bank money, a seller of a tokenised security may receive payment in an asset they are not comfortable holding.” That is not a theoretical risk. Every tokenized bond settled in traditional cash today faces exactly that architecture. The asset side is live. The monetary layer is not. The seller receives settlement through a legacy system that was never designed to participate in programmable finance.

The industry has the asset side working. That is not the hard part. The hard part is the money.

Institutions evaluating tokenized asset strategies need infrastructure that starts at the monetary layer, not the asset layer. FractiFi provides tokenized cash and deposit rails, settlement infrastructure, and programmable capital routing for banks, asset managers, private equity firms, and family offices that need institutional-grade programmable money infrastructure without building it themselves. If you are building a tokenized asset program and need the monetary layer solved, we would like to talk.