Analysis

A Debate the EU Keeps Deferring

In December 2025, the Council considered and set aside an asset-backed “reparations loan” using Russia’s roughly €210 billion in immobilised reserves, choosing a conventional €90 billion joint loan while explicitly reserving the right to return to the frozen assets later.

By May 2026, the Netherlands and four other Member States were working to reopen the file at Ecofin. Since then, the stakes have sharpened: Russia’s central bank has filed a $235 billion claim against Euroclear in a Russian court, and the Commission’s own draft guarantee scheme is already showing a documented shortfall of more than €9 billion, as several Member States decline to take up their assigned share.

Belgium’s resistance was pointed. Prime Minister Bart De Wever called the proposal “fundamentally wrong,” while seven other Member States, led by the Baltic states, jointly wrote to back it as “the most financially feasible and politically realistic solution”. Nonetheless, the debate itself has been conducted almost entirely in legal terms.

However, the windfall-profit mechanism and the proposed reparations loan should not be treated as different degrees of the same instrument. The EU treats windfall profits generated by the frozen cash as belonging to the depositories rather than to Russia, since Euroclear’s ordinary contracts do not remunerate client cash deposits. Economist Nicolas Véron has called it “misleading and inaccurate” to treat this as a smaller version of the reparations loan.

Three financial questions remain unanswered regardless of which legal theory eventually governs the revived instrument: how much exposure the EU should permit, who should bear each component of that exposure, and how the resulting liability should appear on Europe’s books.

The Extraction Rate Problem

The EU has already published a rule governing how much of the windfall profits gets transferred: 99.7% of net profits, under Regulation (EU) 2024/1469. The relevant gap is therefore not the absence of an extraction rate in the current mechanism. It is the absence of a risk-adjusted ceiling on exposure if the EU moves beyond realised windfall profits and begins using the underlying asset base to support a larger financing operation.

The 99.7% distribution rate answers what happens to profit that has already materialised in a given period. It says nothing about how much profit-generating activity should be permitted in the first place, whether the underlying cash balances should be invested more or less aggressively to generate more or less income, or what should happen if a revived reparations loan seeks to draw on the assets’ capacity far more heavily than the current windfall-profits mechanism does.

Euroclear alone generated €4.4 billion in windfall income in 2023, €6.9 billion in 2024, and €3.9 billion across the first three quarters of 2025, a decline of roughly 25% year-on-year on a pro rata basis, reflecting falling interest rates on the underlying cash balances. Cumulative transfers to Ukraine reached €8 billion by August 2026. None of this activity has ever been benchmarked against a stated ceiling; it has simply been whatever the market happened to generate in a given period, with 99.7% of it transferred.

What the EU has not done is publish an explicit rule governing how much of this return can be drawn sustainably, what happens to the mechanism’s design as yields decline, as they already have by roughly a quarter year-on-year, or how the rate would need to change if a substantially larger extraction were authorised under a revived reparations loan.

Approximately €210 billion of Russian central bank assets may be immobilised within the EU. That does not mean that €210 billion is available to finance Ukraine. Nor does it mean that the EU would be taking €210 billion of financial risk. The relevant policy variable is the instrument’s committed exposure.

This distinction is particularly important because the marginal risk of mobilisation may increase as the operation expands. The first €10 billion of exposure may be relatively easy to accommodate within existing buffers; the next €50 billion may leave less liquidity protection, increase Belgium’s legal exposure or create greater political and market risks. The EU should therefore establish a risk-adjusted exposure ceiling for any mechanism that uses the underlying reserves beyond the current windfall-profit model. The ceiling should be linked to observable conditions, including:

• the liquidity requirements of the custodians;
• realised and expected returns on the underlying cash balances;
• the size of existing risk buffers; the value and concentration of the relevant assets;
• the probability and potential magnitude of legal claims;
• the capacity of the EU and Member States to absorb losses.

The ceiling should also contain review or stop-loss triggers. A deterioration in liquidity, a material adverse legal ruling or a significant increase in the probability of claims should trigger a reassessment or suspension of additional commitments. This would transform the instrument from a binary decision, use or not use (that is the dilemma), into a prudential exposure that can be scaled within predefined limits.

Norway offers the clearest available precedent in its Government Pension Fund Global, which operates under a fiscal rule, the handlingsregel, first adopted in 2001, that limits annual withdrawals to the fund’s expected long-term real rate of return. The expected return was formally revised from 4% to 3% with effect from the 2018 National Budget.

In 2024, Norway went a step further, introducing an “asymmetric buffer” with an operational withdrawal assumption of approximately 2.7% to account for downside risks, including the possibility of macro-financial crises, the rigidity of fiscal spending commitments during downturns, and the gradual phasing out of counter-cyclical measures.

The underlying principle translates directly to the Russian case: a state, or, in this instance, a Union, should spend only the return an asset base generates, and never draw down the base itself. That is precisely the distinction the EU has drawn informally between windfall profits and principal since February 2024.

What is missing is Norway’s corresponding discipline: a published, numerically defined extraction rate, reviewed periodically against the fund’s, or, in this case, the frozen assets’ actual observed return, with an explicit and equally published buffer for downside risk given the substantially higher political and legal uncertainty Russian reserves carry. It would also directly answer the question Belgium and the ECB have both raised in different forms: not whether the extraction of returns is legally permissible, which the 2024 Regulation already settled, but how much extraction is financially sustainable, for which no EU institution has yet published a number.

Belgium’s Two Risks

The first is solvency risk: what happens if the guarantee scheme backing the loan is ever called upon. The EU has already drafted a specific answer; the Commission’s proposal uses a GNI-based, joint-and-several guarantee, with Germany covering 24.4% of any potential losses (€51.3 billion), France €34 billion, Italy €25.1 billion, and Spain €18.9 billion, structured as legally-binding, unconditional, irrevocable, on-demand commitments under Article 25 of the draft Regulation. This is a more solidary structure than the EU’s own EFSF/ESM precedent, which used several-but-not-joint guarantees specifically to limit each guarantor’s individual exposure.

But the EFSF/ESM experience is nonetheless the closest available precedent for what can go wrong with a formula-based guarantee scheme, and its documented fragilities are already reproducing themselves in the Russian-asset case in real time. EFSF guarantees required an over-guarantee of up to 165% of the amount actually needed, because only guarantors rated AA or above counted at full value; and its guarantor pool shrank precisely when stress was highest, as Greece, Ireland, Portugal and Cyprus each “stepped out” once they became bailout recipients themselves.

The Russian-asset guarantee scheme is already showing the same pattern: Hungary (€2.4 billion) and Slovakia (€1.5 billion) are expected to withhold their shares, Czechia’s Prime Minister has rejected its €3.7 billion obligation, and Bulgaria and Malta, responsible for €1.2 billion and €400 million respectively, have called for alternatives. Italy’s own governing coalition has separately expressed reservations that could widen the gap to its full €25.1 billion contribution. Taken together, the documented shortfall already exceeds €9 billion, and could exceed €34 billion if Italy’s reservations harden into a formal defection.

The European Investment Bank, which currently holds roughly €190 billion in underused balance-sheet capacity after a decade of comparatively stagnant lending growth, has been proposed as a backstop capable of covering exactly this kind of gap and formalising that role, rather than leaving it as an informal fallback, is the most direct available fix.

The second risk regards the legal retaliation against Belgium and Euroclear specifically. A GNI formula can distribute a financial loss, but it cannot redistribute the fact that Belgium and Euroclear are the parties directly exposed to litigation. Russia’s $230 billion claim against Euroclear demonstrates that this risk is no longer theoretical. Analysts have also mapped potential challenges before the European Court of Justice and under the 1989 Belgium USSR investment treaty, which could proceed to Stockholm or a UN-administered tribunal.

The EU has, in fact, already built some legal machinery for this second risk, separate from the financial guarantee.

The draft Emergency Regulation renders foreign arbitration awards against Belgium or Euroclear unenforceable within the EU, and Article 24 caps the EU’s reimbursement commitment at the loan’s face value for a maximum of 16.5 years, a figure tied to a 15-year sunset clause following the eventual termination of the underlying Belgium-Soviet investment treaty, , but Belgium argues that this window is insufficient.

A distinct proposal within policy literature, a conditional first-priority lien on the frozen assets that activates only if Russia fails to honour eventual reparations, paired with a shared EU-G7 indemnification scheme, was developed specifically to address this category of risk rather than the solvency category, and offers a more structurally appropriate answer in place of extending the existing guarantee formula.

It is also worth noting that the practical risk of an actual Euroclear default may be smaller than the political debate assumes: one analysis of Belgian insolvency law found that a Belgian entity is only legally bankrupt when two conditions are met simultaneously, persistent cessation of payment and a credit crisis, and that a temporary guarantee shortfall would more plausibly qualify Euroclear for Emergency Liquidity Assistance from the National Bank of Belgium than push it into actual default.

A Loss Waterfall Rather Than a Single Guarantee

The most robust architecture would go one step further. Rather than treating the entire potential loss as a single liability distributed among Member States, the EU could establish a loss waterfall, with each layer addressing a different category of risk.

The principle would be:

• First-loss layer: Existing windfall-profit buffers and other dedicated reserves.
• Second-loss layer: Safeguards maintained by the custodial structure and Euroclear.
• Third-loss layer: A Union-level indemnification mechanism for legal-retaliation exposure.
• Fourth-loss layer: Formula-based Member State guarantees for residual financial losses.
• Fifth-loss layer: The European Investment Bank, which holds approximately €190 billion in underused balance-sheet capacity after a decade of comparatively stagnant lending growth, formally committed to covering residual gaps, such as Member State defections already observed in the fourth layer, and to supplementing them, where necessary, with additional G7 or other international support.

It also creates an important incentive effect. If Member States are asked to guarantee the entire potential loss from the beginning, they have an incentive to focus on the maximum theoretical exposure. A layered system instead makes explicit which risk is absorbed where and in what order.

This would also help resolve the problem with a simple capital-key approach: a capital key can determine who pays if a defined financial loss occurs, but it cannot determine who should bear a Belgium-specific legal exposure.

The loss waterfall should be combined with the exposure ceiling proposed above. The EU should assess not only the total amount of exposure but also the marginal risk created by increasing it. The relevant question whenever the facility is expanded should therefore be: What additional expected loss and legal exposure does the next €10 billion of financing create, and which layer of the waterfall would absorb it? This would prevent the political debate from treating every additional euro of financing as equivalent.

A €10 billion increase in exposure may be relatively inexpensive when substantial liquidity buffers remain available. The same €10 billion may be materially more risky once those buffers have been depleted or the probability of adverse litigation has increased.

How conservatively either category of exposure should ultimately be priced depends in part on an assumption the political debate has repeatedly invoked without testing it against data: that freezing another country’s reserves makes non-aligned central banks warier of holding euros, and that any flight from the dollar following comparable US sanctions episodes should flow naturally toward the euro instead. Neither half of this assumption holds up cleanly against the record through 2025.

This is backed by the IMF’s quarterly Currency Composition of Official Foreign Exchange Reserves survey, which shows the euro’s share of allocated global reserves moving within a comparatively narrow band across all four quarters of 2025, 20.00% in the first quarter, rising to 21.13% in the second, before settling back to 20.36% in the third and 20.25% in the fourth. The dollar’s share, over the same period, moved from 57.79% to 56.32%, then recovered to 56.92% and settled at 56.77%, continuing a considerably longer, more structural secular decline that has taken the dollar’s reserve share from over 72% in 2001 to roughly its current level.

In parallel, a 2025 NBER working paper by Menzie Chinn, examining central bank reserve holdings across roughly 47 to 48 countries over the 1999 to 2022 period, finds that financial sanctions imposed by the United States do produce a measurable reduction in the sanctioned country’s own dollar holdings, but that this diversification flows disproportionately toward smaller reserve currencies and gold rather than toward the euro.

If a comparable pattern holds following the EU’s own freeze, Belgium’s exposure, on the reserve-currency dimension specifically, should probably be priced against a smaller and more narrowly distributed risk than the political debate around it currently assumes, though this remains the area where the evidence is thinnest and where standing, quarterly COFER-based monitoring would let the Council price this component rather than argue over it.

The Balance Sheet Question

NextGenerationEU is the direct precedent for what happens when classification is left unresolved. Member State loans count as national debt, but the Commission’s own underlying borrowing, up to €806.9 billion, is recorded at the EU level rather than on any individual Maastricht debt ratio.

A 2024 ZEW Mannheim study found roughly €262 billion in NGEU-linked German liabilities absent from official statistics, leading researcher Friedrich Heinemann to argue that “there is an urgent need for EU debt to be counted towards the national debt of the Member States.”

Under ESA 2010, a reparations-backed instrument could be classified as a contingent liability, a guarantee, or an outright debt, each with different implications for fiscal visibility. The lesson from NGEU is not that questions about its classification were illegitimate to raise; reasonable disagreement over how a novel joint instrument should be treated is normal, even where a Member State’s contribution share doesn’t match its share of the funding it received.

The lesson is that the EU let the question remain open after the instrument was already disbursing funds, rather than settling it beforehand, and that the sequencing choice turned a legitimate disagreement into a documented, years-long transparency dispute. However, common ground on classification needs to exist before the instrument goes live, not as a running argument conducted alongside it.

The Commission and Eurostat should therefore publish the instrument’s ESA 2010 treatment before, rather than after, disbursement. The financial architecture proposed above deliberately creates several layers of contingent exposure. The statistical treatment should make clear not only the instrument’s headline liability but also which risks sit on the balance sheets of Member States, the Union, Euroclear, and the EIB.

Towards a Priced Instrument

The three problems should not be treated as independent technical questions. They form a single prudential architecture.

Step 1: Account for the exposure

• The Commission and Eurostat should determine in advance where the resulting liabilities appear in European and national accounts and how contingent exposure is disclosed. The resulting framework can therefore be summarised as:

Step 2: Limit the exposure

• The EU should determine how much additional exposure the immobilised assets can support without compromising liquidity, financial resilience or legal safeguards. This requires a state-contingent ceiling rather than an arbitrary fixed number.

Step 3: Allocate the exposure

• The EU should distinguish between different risks; different risks should be assigned to different layers of the risk-sharing mechanism:
  • ordinary financial loss;
  • custodial concentration
  • legal-retaliation exposure;
  • and residual sovereign-repayment risk.

The EU’s next debate over Russia’s immobilised reserves should not begin with a binary question about whether the assets can be used. It should begin with a financial architecture that determines how much exposure can be created, how that exposure changes at the margin, who absorbs each component of the resulting risk, and when further mobilisation should stop. A revived instrument should therefore return to Council as a priced facility with a state-contingent ceiling, a differentiated risk waterfall, a marginal risk test and a settled fiscal classification.

None of this needs to be built all at once, and treating it as a single package would unnecessarily slow it down. The extraction ceiling and the fiscal classification can proceed now, as administrative determinations under legal bases the EU already holds. The waterfall’s final layer depends on a separate EIB decision that can run alongside them. What cannot wait is the facility covering Belgium’s exposure, since nothing resembling it exists in draft yet and the waterfall cannot be completed without it. Building that piece first, rather than leaving it for last, is not simply the faster route to a workable instrument. It is also the one step that answers the objection that stalled the reparations loan the first time around.

The objective is not to eliminate risk, which is impossible, but to ensure that risk is measured before it is committed, allocated before it materialises, and capped before it becomes politically irreversible.