Pontes: why an obscure private chain instead of just Ethereum?

The question

On 21 September 2026 the European Central Bank (ECB) launched Pontes, a service that lets banks settle tokenised-asset trades in central bank money. Christine Lagarde summed it up as "it's a digital euro made available for banks".

The obvious objection: Ethereum already exists, is open, and hosts most tokenised assets. Why did the Eurosystem spend years building a private ledger instead of deploying a euro token there?

The short answer: the choice is about control and law far more than about technology. The technical arguments against Ethereum are weaker than the ECB's documents suggest, and the ledger is the least convincing part of the Pontes design.

What Pontes actually is

The official documents describe three modules: an interface to T2 (the Eurosystem's real-time gross settlement system), a Eurosystem ledger, and an API gateway.

• The ledger. The ECB's February 2026 technical session describes a "Permissioned DLT-based platform with Central Bank nodes". Each national central bank owns one node, all hosted by four service providers: the Bundesbank, the Banco de España, the Banque de France and the Banca d'Italia.
• The tokens. At launch, cash tokens are intraday only. They are backed by funds escrowed in T2 and swept back every evening.
• The gateway. Banks never touch the chain. According to the Business Description Document, "The API Gateway component acts as a trusted oracle" for the settlement protocol.
• The protocol. Delivery versus payment uses Hash-Link, which the Banca d'Italia paper that introduced it calls "a lightweight, API based and DLT agnostic protocol".

The software stack is not named in any official document found. The 2024 trial systems ran on Hyperledger Fabric, so Fabric is the likely answer, but that is an inference.

Performance targets at launch are modest: 10 transactions per second, 95% of cash legs within 5 seconds, open 09:00 to 16:00 CET on business days. Legal finality still happens in T2, not on the ledger: the ECB's Pontes page lists a dual settlement model and settlement finality in T2 among the features of the initial launch.

Launch versus target

That is the first release, not the end state. The User Requirements Document (URD), finalised in April 2026, specifies the target service, and it differs on three points:

The ECB says the improvements will come in stages after the launch. No date for the target service was found in the documents consulted.

The official case for a private ledger

Control and sovereignty

This is the stated reason, and the strongest one. The URD requires a consensus mechanism "fully under control of the Eurosystem", with all data inside the European Economic Area and all nodes hosted by the operator. It lists strategic autonomy on European infrastructure among its objectives.

Ethereum cannot meet that requirement by construction. Its validators are anonymous and global, fees are paid in ether, and protocol upgrades are decided by a community the ECB does not sit in. In a contested fork, the issuer of a euro token would have to pick the surviving chain.

The requirement breaks down into six dependencies. Only one of them rests on law. The other five are choices the Eurosystem made for itself in its own requirements, which makes them policy. Each is followed, in italics, by a critique.

1. Who runs the nodes (policy). The functional training material for the pilot states: "Each NCB must own a node hosted on SP infrastructure", NCB being a national central bank and SP one of the four service providers. URD requirements PONTES.UR.03.010 and PONTES.UR.03.030 say the same of the target service: a private permissioned platform owned by the Eurosystem, with every node hosted by the operator.

   This assumes that security derives from knowing the validators. Ethereum's security derives instead from explicit, publicly specified economic incentives: under proof of stake, validators commit capital that is destroyed if they misbehave, and the rules apply identically to all. Ethereum has produced blocks without interruption since 2015, whereas TARGET2 suffered an outage of almost ten hours in October 2020 that also reached its backup systems. Control of a token sits in its contract, not in the validators: the issuer keeps mint, burn, freeze and allowlist powers, and validators can delay or order a transaction but cannot create or seize euros. Four providers hosting every node form a single governance and operational point of failure, which a distributed design is meant to remove. The 2021 EIB bond, covered below, shows a Eurosystem central bank keeping control of its money on a public chain without running the chain.

2. Foreign legal reach (policy). No EU legal text was found on this point. The ECB's Pontes page gives the aim as: "Support the EU's strategic autonomy." The precedent usually cited is American: OFAC sanctioned the Tornado Cash contracts in August 2022. It is weaker than it looks, because a US appeals court overturned the designation in November 2024 and the Treasury announced in March 2025: "we have exercised our discretion to remove the economic sanctions against Tornado Cash". Discretion can be exercised the other way again.

   This concern is overstated. Sanctions bind persons and institutions, not ledgers. Euro-area banks with exposure to the United States are within OFAC's reach whatever infrastructure settles their payments, and a private ledger does not change that. The 2022 episode produced delayed inclusion of some transactions, not their exclusion, and the measure behind it was overturned and then withdrawn.

3. Who decides upgrades (policy). PONTES.UR.03.040 asks for a modular design that avoids lock-in to any one consensus mechanism. PONTES.UR.03.180 allows new smart contracts only through the Pontes change and release process, implemented and deployed by the operator.

   Ethereum's protocol changes follow the public EIP process. Proposals are debated for months or years, must be implemented by several independent client teams, and take effect only if node operators choose to adopt them. No party can alter the rules unilaterally, which is a stronger guarantee than a consensus mechanism the operator may change at will. The process is open, so the Eurosystem could take part in it like any other stakeholder. In a contested fork, the issuer of a token decides which chain its liabilities live on: that is a power held by the issuer, not a vulnerability.

4. Ordering and fees (policy). PONTES.UR.03.070 requires first in, first out processing, with no prioritisation or reordering of transactions. No official text was found on paying fees in ether. The nearest is section 1.3 of the URD, under which Pontes relies only on components operated inside its own perimeter.

   The literature on transaction reordering concerns price-sensitive trades on open markets. A transfer of central bank money between two allowlisted banks offers nothing to extract. Fees are immaterial at the value of wholesale settlement, where a single transaction may move billions of euros. A shared ledger also removes a problem Pontes has to engineer around: with cash and asset on the same chain, delivery versus payment is atomic by construction and needs no Hash-Link protocol, no timeouts and no trusted oracle holding the secrets.

5. Finality (law). The Settlement Finality Directive, Article 2(a), protects a system only if it is a formal arrangement "governed by the law of a Member State chosen by the participants", and designated and notified by that Member State. Ethereum has no chosen law and no designation. PONTES.UR.04.020 then requires settlement in tokens to have the same legal effect as settlement in existing TARGET accounts under that directive.

   The absence of legal finality on a public chain is a legislative choice, not a property of the technology. The directive is being rewritten now, with DLT explicitly in scope, and the Member States, the Commission and the Parliament could extend its protection to a euro token on a public chain if they wished. At launch, legal finality still occurs in T2 and not on the Pontes ledger, so the private chain does not yet supply what Ethereum is said to lack.

6. Data location (policy). PONTES.UR.03.120 requires all data processed and stored in the system to reside in the European Economic Area, with any exception needing explicit Eurosystem approval. A chain replicated on nodes worldwide fails this outright.

   This argument is circular. A requirement written so that any globally replicated ledger fails it restates the conclusion; it does not justify it. It is also self-imposed: the Eurosystem wrote the URD itself, under its own Market Infrastructure Board, so "Ethereum cannot meet the requirement" describes a preference, not an external constraint. The relevant question is which data must remain in the European Economic Area, and why, given that what a public chain replicates can be limited to encrypted values or commitments. A middle path was available as well: a Eurosystem-operated layer settling to Ethereum, along the lines of ZKsync Prividium discussed below, would keep sequencing and data under Eurosystem control while remaining interoperable with the chain where most tokenised assets already sit.

Settlement finality

International standards say market infrastructures should settle "in central bank money where practical and available", and require clear, final settlement. Piero Cipollone of the ECB's Executive Board puts the premise plainly: "Central bank money is the safest and most liquid settlement asset."

Ethereum's finality is of a different kind. Its consensus protocol "combines Casper FFG, a finality tool, with LMD GHOST", and a block becomes final only after about two epochs, roughly 13 minutes. That finality is economic, not legal. EU settlement-finality law protects designated systems, as point 5 above shows, and a proposal now before the European Parliament to replace the 1998 directive deals explicitly with "finality in systems that use DLT". Until such rules cover a public chain, a transfer there has no equivalent legal protection.

Access to central bank money

Reserves may only be held by eligible institutions. A freely transferable token would be a retail digital euro by the back door, and issuing one requires the regulation proposed by the European Commission in 2023, which has not yet been adopted. A token restricted by allowlist is technically possible on Ethereum, but then the "open" chain adds little.

Prudential and compliance risk

Bank supervisors are wary. A Basel Committee working paper on permissionless ledgers finds that practices for mitigating their risks "have not been tested under stress". Those risks include reliance on unknown third parties, which makes due diligence hard.

What the ECB's own powers say

The ECB has specific rights in this field. Article 127(2) of the Treaty on the Functioning of the European Union makes promoting the smooth operation of payment systems a basic task of the European System of Central Banks. Article 22 of its Statute lets the ECB and the national central banks provide facilities, and adds that the ECB "may make regulations, to ensure efficient and sound clearing and payment systems". That is the legal basis for the TARGET Services, for Pontes, and for the oversight regulation on systemically important payment systems.

These rights have three limits:

• They are permissive. "May provide facilities" lets the ECB build its own ledger. It does not oblige it to, and it does not forbid issuing on a public chain.
• They are read narrowly. In case T-496/11 the General Court held that Article 22 covers payment systems only, not securities clearing, as a Council document on the later attempt to amend the article recalls.
• They stop at the border. They bind operators in the EU, not Ethereum's validators or core developers. Finality designation, the one point that is law, belongs to the Member States, not to the ECB.

So the powers explain why the ECB can build Pontes. They do not explain why it had to.

What the academic literature adds

• Governance of the ledger. Auer, Monnet and Shin model who should update a monetary ledger and find that, when long-term incentives are strong, "a centralized ledger is always optimal". Decentralisation wins only in low-trust settings. Central banks are the textbook high-trust case. The paper is also available as BIS Working Paper 924.
• Censorship. After OFAC sanctioned Tornado Cash in 2022, Wahrstätter et al. measured compliance-driven censorship by block producers and proved "a fundamental limitation of PoS and Proof-of-Work (PoW) protocols against censorship resilience". For the ECB this cuts both ways: foreign sanctions law could reach transactions in its own currency. The sanctions themselves were lifted in March 2025, but the measured behaviour of block producers is the lasting finding.
• Transaction ordering. Daian et al., "Flash Boys 2.0" documented how block producers profit from reordering transactions, now known as maximal extractable value. Pontes instead mandates strict first-in, first-out processing.
• The ledger may not matter. The BIS Annual Economic Report 2025 envisions a unified ledger for tokenised money and assets "which may or may not use distributed ledger technology". That undermines the case for any blockchain here, public or private.

The case for Ethereum

A Eurosystem central bank already did it

In April 2021 the European Investment Bank issued a €100 million bond, settled with a Banque de France representation of central bank money. The EIB press release states: "The transaction will use Ethereum, a public blockchain protocol." So the approach is feasible and was tested inside the Eurosystem. It was not carried into production.

The private chain adds little that a database would not

All nodes sit with four service providers. Participants connect through an API. A trusted oracle holds the secrets that make settlement atomic. The ECB's own exploratory work report describes one predecessor as infrastructure that "acts as technical bridge between the T2 and market DLT platforms". A bridge needs an interface, not necessarily a ledger of its own. Given the BIS and Auer-Monnet-Shin points above, a conventional database behind the same API would arguably serve as well.

The asset side can still be public

Hash-Link only requires that a market platform can lock an asset against a secret. Nothing in the protocol prevents that platform from being a public chain, provided its operator is a regulated entity, for instance one authorised under the DLT Pilot Regime. Pontes therefore keeps the euro leg private without forcing the assets off Ethereum.

Throughput and hours

Ethereum mainnet handles roughly 100 to 240 simple transfers per second at its current gas limit, runs around the clock, and its Glamsterdam upgrade targets a much higher limit. Pontes at launch offers 10 per second, seven hours a day, with round-the-clock settlement specified only for the target service. This matters less than it seems: wholesale settlement is low-volume and high-value.

Privacy: is zero-knowledge ready?

Banks will not publish their interbank flows. The Enterprise Ethereum Alliance's 2026 privacy report agrees: "Privacy and confidentiality are not features, they are prerequisites for enterprise adoption."

Zero-knowledge proofs can in principle deliver this on a public chain. The Privacy Pools paper by Buterin and co-authors shows how users can prove funds are clean "without publicly revealing their entire transaction graph". Other approaches use homomorphic encryption (see the draft ERC-7984 confidential token standard) or trusted execution environments.

Maturity is the problem:

Of the seven enterprise solutions the EEA reviewed, only two cleared its bar for production. A central bank writing requirements in 2025 could not reasonably build base money on that. A central bank writing them in 2028 might. The Bundesbank is already reported to be testing Prividium, according to coverage listed on ZKsync's site.

What did it cost?

No official figure has been published. The following is an estimate, not a sourced number.

For scale, the retail digital euro is reported to carry a €1.3 billion development budget. Income so far is negligible: the Pontes Pricing Guide sets only "one-off connection fees" of €2,500 per market participant, and thirteen institutions had onboarded at launch.

A mainnet token would not have been free either. Deploying a contract is cheap. The link to T2, the legal framework, key custody, security and onboarding are where the money goes, and all of them would still be needed.

Verdict

The ECB did not choose a private chain because Ethereum is technically inadequate. It chose one because it will not let base money depend on infrastructure it does not govern, and because the law gives it no finality there. Those are defensible positions, but only the second is a legal constraint. The first is a decision the Eurosystem was free to take differently.

What is harder to defend is the ledger itself. With every node hosted by the operator and every participant behind an API, the "distributed" in Pontes's distributed ledger does very little. The fairer comparison is between Ethereum and a well-run database behind a good API, and what the ECB built sits much closer to the database.

Further reading

• ECB press release of 21 September 2026 on the initial launch
• ECB press release of July 2025 announcing the Pontes and Appia two-track strategy
• PDF version of the exploratory work report, with details on the three trial solutions
• Banque de France service description of DL3S, the cash-token platform Pontes descends from
• BCBS Working Paper 44, full text
• BIS press release on the unified ledger
• Ethereum Foundation Institutional Privacy Task Force map and its essay on public rails versus private ledgers
• ERC-5564 stealth addresses and EIP-8105 encrypted mempools
• Blockchain Privacy and Regulatory Compliance, journal version