When Rearmament Is Financed Off Budget, Sovereign Bond Markets Read the Numbers Differently

Between 11 February and 10 April 2026, the Council of the European Union approved defence loan packages for 18 member states under an instrument that had entered into force less than a year earlier. Over roughly the same window, euro area general government debt climbed to 88.9% of GDP in the first quarter of 2026, up from 87.2% a year before, on Eurostat's release of 21 July 2026. And yet spreads between the bloc's higher-debt sovereigns and German Bunds spent the period near their narrowest levels in close to two decades.

Those observations are usually presented as being in tension. They are not. European rearmament is being financed through arrangements that separate the volume of new borrowing from the price charged for it, and the separation runs through accounting perimeters, issuance calendars and rating methodologies rather than through any single yield. The useful question is not whether defence spending pushes yields up, but which channel a given country's rearmament is travelling through, and how long that channel takes to produce an observable number.

Where the Numbers Actually Stand in Mid-2026

Start with the commitment. At the Hague summit in June 2025, allies adopted a target of 5% of GDP by 2035, split into 3.5% for core defence requirements and NATO capability targets and 1.5% for broader defence-related investment, with a review scheduled for 2029. That structure matters more than the headline, because the two components are measured differently and one has no settled definition yet.

On NATO's own accounting, European allies and Canada spent 2.27% of combined GDP on defence in 2025, against 1.40% in 2014. NATO's estimates published on 28 August 2025, with an information cut-off of 3 June 2025, put Poland first at 4.48%, followed by Lithuania at 4.00%, Latvia at 3.73% and Estonia at 3.38%. The United Kingdom stood at 2.40%, France at 2.05%, Italy at 2.01% and Spain at 2.00%. Poland's equipment share reached 54.44% in the 2025 estimate, up from 42.90% a year earlier — a mix implying cash outflows stretching years beyond the budget year in which contracts are signed.

One year on, compiled 2026 figures place Lithuania at 5.33%, Estonia at 5.11%, Latvia at 4.92%, Poland at 4.68% and Greece at 3.65%, with most allies clustered between 2.0% and 3.0%. Ahead of the Ankara summit on 6–7 July 2026, NATO's Secretary General described European allies and Canada as already investing around 4% of GDP in defence and security — a figure that reconciles with the 2.27% core measure only if the broader category is included. Two accurate numbers describing different things.

Sovereign	2026 defence, % GDP (NATO basis)	Gov. debt, % GDP (Q1 2026)	Deficit, % GDP (2025)
Poland	4.68	61.6	-7.3
Lithuania	5.33	42.3	—
Estonia	5.11	25.2	—
Latvia	4.92	46.9	—
Germany	2.0–3.0 band	64.4	-2.7
France	2.0–3.0 band	117.6	-5.1
Italy	2.0–3.0 band	138.9	-3.1
Euro area	—	88.9	-2.9

Read the table across rather than down. The sovereigns spending most on defence carry the lightest debt stocks; the heaviest debt stocks belong to countries not driving the rearmament numbers. Any model mapping defence spending directly onto sovereign risk has to survive that pattern first.

Three Definitions of the Same Spending

The common error here is treating one measurement as the measurement. At least three are in active use, and they disagree by percentage points of GDP.

• The alliance basis. NATO counts payments by a national government to meet the needs of its armed forces and those of allies. Personnel expenditure explicitly includes pensions paid to retirees, and equipment expenditure includes R&D on major equipment. NATO's documentation notes this diverges from national definitions.
• The national accounts basis. When Fitch affirmed Poland at A- with a negative outlook on 27 February 2026, it put Polish military spending at about 3% of GDP in 2025, up from 1.6% in 2021. The same year, on the alliance basis, Poland registered 4.48%. Neither figure is wrong. They count different things over different perimeters.
• The debt perimeter. Poland's Ministry of Finance reported on 10 June 2026 that national-definition public debt stood at PLN 2,008.2bn, or 50.6% of GDP, at the end of the first quarter, while EDP debt — the EU-harmonised measure — stood at PLN 2,444.3bn, or 61.6% of GDP. The PLN 436.1bn gap exists because special-purpose vehicles, among them the Armed Forces Support Fund, sit inside the European measure and outside the domestic one.

A sovereign can be simultaneously below its own constitutional debt threshold and among the fastest-indebting economies in the European Union. Both statements can be verified from official releases published in the same month.

Four Channels, Four Different Clocks

Channel one: auction volume, measured in days

This is the channel most commentary assumes is the only one. Germany's finance agency announced on 18 December 2025 a 2026 issuance plan of roughly €512bn, against €425bn issued in 2025. The composition is instructive: about €309bn refinances maturing federal securities, €98bn covers the core budget, €60bn funds the infrastructure and climate special fund, and €25.5bn the special fund for defence. Defence is roughly 5% of the announced programme; the refinancing wall is over 60%.

Auction evidence arrives fast: bid-to-cover ratios, tail sizes and the concession at each tap are visible within hours. Shortest observation lag, least persistence.

Channel two: duration mix, measured in quarters

Volume understates the effect if new supply concentrates at the long end, where price sensitivity is highest and the natural buyer base thinnest. Germany's 2026 plan allocates roughly €92bn to two-year Schatz, €73bn to five-year Bobl, €82bn to ten-year Bund, €20bn to the 15- and 20-year basket and €29bn to the 30-year basket, alongside €176bn of Bubills. Procurement cycles running to the mid-2030s argue for long-dated funding; absorption capacity argues for the belly. Where an issuer resolves that tension sets how much term premium the programme generates.

Central bank balance sheet reduction compounds this. One sell-side estimate put 2026 net supply that private investors must absorb at roughly €930bn — about €550bn of government issuance plus around €380bn from quantitative tightening. That is a bank projection, not an official statistic. The direction is not in dispute: the price-insensitive buyer has stepped back while gross programmes have grown.

Channel three: the accounting perimeter, measured in years

This is the channel that distinguishes the current episode from a conventional fiscal expansion. Three mechanisms move defence borrowing off the ordinary budget line.

1. Constitutional carve-outs. Germany's March 2025 amendment exempted defence spending above 1% of GDP from the debt brake and created a €500bn special fund for infrastructure and climate neutrality. The constraint it would otherwise have hit no longer binds.
2. Fiscal-rule flexibility. The EU's national escape clause permits an annual deviation of up to 1.5% of GDP from agreed net expenditure paths, for four years covering 2025 through 2028. The Council granted it to 15 member states on 8 July 2025, to Germany on 10 October 2025, to Austria on 17 February 2026 and to Spain on 12 June 2026 — 18 in total.
3. Supranational intermediation. The SAFE regulation, in force since 29 May 2025, provides up to €150bn in loans funded by EU-bond issuance; Council approvals for 18 member states ran from 11 February to 10 April 2026. The Commission announced €90bn of EU-bond issuance for the first half of 2026, raised that target in May 2026, and added €80bn for the second half on 19 June 2026, taking planned 2026 EU-bond issuance to €180bn.

Each mechanism relocates borrowing rather than removing it. Headline national debt ratios therefore lag the underlying commitment, sometimes by years, and the lag differs by country.

Channel four: the rating channel, arriving discretely

Rating actions do not smooth. They arrive as steps. Fitch's 27 February 2026 action on Poland is the clearest instance in the current cycle: rating affirmed at A- with a negative outlook, general government deficit at 7.0% of GDP in 2025 against an A-category median of 2.9%, public expenditure near 50% of GDP, and debt projected at around 70% of GDP in 2027 from about 59% in 2025. Eurostat's 22 April 2026 release recorded Poland's 2025 deficit at 7.3% of GDP, the largest in the Union that year, against a euro area average of 2.9%.

Note what the rationale did not say. It cited the deficit level, the debt trajectory and the absence of a credible consolidation plan — not defence spending as an independent negative. Agencies price the fiscal path, not the purpose of the outlay.

What the Market Actually Did

The supply-pushes-yields-up intuition has one strong data point in its favour. On 5 March 2025, following the announcement of the German fiscal package, the ten-year Bund yield rose 30 basis points in a single session — the largest daily increase since the fall of the Berlin Wall. Spillover analysis of that episode found Italian yields moved roughly 83% and Spanish yields roughly 84% of what statistical models predicted, with Dutch and Belgian correlations to the Bund reaching about 95% and the French correlation about 91%.

That is a repricing of the risk-free curve, not a credit event. Everything moved together, which is what a duration shock looks like; a credit shock would have widened spreads.

The following period reinforced the point. Commentary in January 2026 put the Italian ten-year yield roughly 130–150 basis points above Bunds, the narrowest in nearly two decades, with the EU bond spread narrowing from around 70 basis points in 2022 to about 40 basis points on average in 2025. Record issuance and two-decade-tight spreads coexisted.

Historical Parallel: Unification Financing and the Special Funds

The closest structural precedent is not the post-Cold War drawdown but German unification financing between 1990 and 1996, which used the same off-budget architecture now deployed for defence.

Bundesbank data show combined central, regional and local authority debt at DM 929bn, just under 42% of GDP, in 1989, rising to DM 2,135bn, almost 60.5% of GDP, by end-1996 — about 18.5 percentage points in seven years. Much arrived through vehicles outside the ordinary federal budget. The German Unity Fund borrowed DM 95bn for transfers to the eastern Länder through 1994 and still carried DM 84bn at end-1996. The Redemption Fund for Inherited Liabilities held DM 332bn, having absorbed DM 205bn from the Treuhand agency and DM 102bn from the Debt-Processing Fund.

The instructive number is the servicing cost. Federal interest expenditure rose from 11% of total federal spending in 1989 to 16.5% in 1996. The obligation was not recorded in the headline budget when it was incurred; it appeared later as a claim on every subsequent budget, competing with every other line.

Three features carry across: off-budget vehicles were used because the ordinary constraint would have bound; the debt was eventually consolidated into the general government perimeter; and the fiscal consequence outlived its political trigger by over a decade. That does not indicate where yields go. It indicates which variable to track.

Where This Stops Being Diagnostic

Several conditions would break the framework above, and they should be stated plainly.

• Unification was a domestic transfer; rearmament is partly an import. Poland's equipment share above 54% implies heavy foreign procurement, with a current account and exchange rate footprint unification transfers lacked.
• The monetary regime is different. Germany in 1990 had its own central bank responding to its own fiscal shock. Euro area members share a monetary authority holding anti-fragmentation instruments that have not been tested in a defence-driven widening.
• The tight-spread evidence is regime-dependent. Compression through 2025 and early 2026 occurred with inflation off its peak. That is evidence supply did not bind under those conditions, not that supply does not matter.
• Bank supply estimates are not official statistics. The €930bn figure is sell-side work, wrong in either direction, and not equivalent to Eurostat or debt agency data.
• Commitment is not outlay. A 2035 target reviewed in 2029 is a stated intention, and published defence trajectories have historically been revised.
• Low starting debt is a buffer, not immunity. Estonia at 25.2% and Latvia at 46.9% have room, but small issuers can face liquidity-driven pricing unrelated to solvency metrics.

One reasonable reading: the fiscal path dominates and the defence label is close to irrelevant to pricing. An equally defensible opposing reading: spreads compressed because the supply had not yet arrived — the escape clause runs only to 2028, SAFE approvals ran through April 2026, and the 2035 trajectory implies acceleration rather than a plateau. Neither can be dismissed on present evidence.

What to Watch Next Week

• Long-end auction results. Bid-to-cover and tails at 15-year and longer taps, where the duration channel shows first. Germany's 2026 plan concentrates about €49bn in those baskets before syndications.
• Any revision to second-half funding targets. The Commission raised its first-half 2026 EU-bond target in May; a further revision would signal SAFE drawdown running ahead of schedule.
• Rating commentary separating defence from consolidation. Whether any agency begins treating defence outlays as a distinct category rather than folding them into the deficit path.
• Monthly debt releases from high-deficit issuers. Poland's Ministry of Finance publishes State Treasury debt monthly; the national-to-EDP gap is the cleanest proxy for off-budget accumulation.
• Spread behaviour on defence-headline days. Spreads widening on procurement news while the risk-free curve is flat would mean the credit channel is activating. Both moving together keeps it a duration story.

Concrete Framework — The Order That Matters

A monitoring checklist for the rest of the cycle.

1. Fix the definition before comparing. Record whether a figure is on the NATO basis, national accounts basis, or budget-appropriation basis. A 1.5 percentage point gap for the same country and year is normal, not an error.
2. Track the perimeter gap, not just the ratio. For any sovereign using special funds, monitor the national-versus-EDP difference. Poland's was PLN 436.1bn at Q1 2026. A widening gap is off-budget accumulation in progress.
3. Separate refinancing from new money. Roughly €309bn of Germany's €512bn 2026 plan is rollover. Headline issuance overstates the fiscal impulse whenever the maturity wall is large.
4. Weight by duration, not volume. A programme skewed to Bubills and two-year paper is a different market event from one skewed to 30-year baskets, at identical total size.
5. Use the deficit, not the defence share, as the credit variable. Poland at -7.3% with an outlook cut, against Estonia above 5% of GDP on defence with debt at 25.2%, supports the deficit as the operative variable.
6. Diarise the two hard dates. The escape clause lapses after 2028; the alliance target is reviewed in 2029. Both are scheduled points where the arrangement either extends or does not.
7. Watch the interest bill as a share of spending. The unification precedent moved from 11% to 16.5% of federal expenditure in seven years. Published in ordinary budget documents, it is the slowest and most reliable signal that a financing decision has become a structural claim.

The honest summary: transmission from a defence commitment to a sovereign yield is real, multi-channel, and slower than most commentary assumes. As of mid-2026 the fastest channels have priced and the slowest have not. That is a statement about observation lags, not a forecast.

Disclaimer: This article is analysis of publicly available macroeconomic and policy data. It is not investment advice and contains no recommendation regarding any security or issuer.