AGP Picks
View all

Financial Stability Review, November 2026

News originally published by

ECB - European Central Bank

The euro area economy has proved resilient so far this year, in the face of energy supply disruptions, trade frictions and extreme weather events. This resilience in part reflects drawdowns of petroleum reserves, which have kept energy markets more stable than they would otherwise have been, and optimism about the productivity-enhancing potential of artificial intelligence (AI), which has, despite recent volatility, supported financial market sentiment, investment and global growth.

AI promises not only to drive up long-term economic growth, however; it also raises risks. Operational risk for private firms and public authorities, for instance, is amplified by rising cyber threats and heavy reliance on a small cluster of third-party providers. While the technology may deliver large productivity gains, current valuations depend on very ambitious earnings expectations for AI-related firms. In parallel, those firms are continuing to increase the volume of debt on their balance sheets, while circular financing is making intrinsic demand hard to gauge.

Sovereign vulnerabilities are also building up, and debt levels are high in many advanced economies, including some euro area countries. At the same time, sovereign funding costs have risen. Fiscal consolidation is overdue and increasingly important in an environment where the presence of news-sensitive, leveraged investors in sovereign bond markets is growing.

Rising exposures to AI and sovereign debt warrant attention, with vulnerabilities being more pronounced in some segments of the non-bank financial sector. Even if banks are not exposed to a potential correction in AI-related assets in material terms, they could face spillovers via strong links with non-banks. Moreover, a prospective deterioration in macro-financial conditions due to the energy supply disruptions could lead to rising credit risk. Banks have weathered past shocks well overall, thanks in part to the regulatory reforms introduced since the global financial crisis. They remain strongly positioned to absorb shocks, with resilience underpinned by ample liquidity buffers, sizeable capital headroom and robust profitability.

This edition of the ECB’s Financial Stability Review also includes three special features. The first of these discusses the financial stability implications of AI adoption in the financial industry and the real economy. The second explores the complex interplay between AI and cybersecurity, while the third assesses the financial stability implications of large initial public offerings.

The Financial Stability Review is prepared with the involvement of the ESCB Financial Stability Committee, which assists the decision-making bodies of the ECB in the fulfilment of their tasks. It is intended to foster awareness of systemic risks among policymakers, the financial industry and the public at large, with the ultimate goal of promoting financial stability.

Boris Vujčić
Vice-President of the European Central Bank

The financial stability outlook is clouded by risks from AI valuations, fiscal imbalances and geoeconomic tensions

The euro area economy has proved robust to recent shocks, but prolonged geoeconomic headwinds could yet test this resilience. The euro area economy has demonstrated remarkable resilience in the face of multiple challenges, including an energy shock, trade frictions and extreme climate-related events, with incoming economic data consistently exceeding market expectations between July and October this year (Chart 1, panel a). At the same time, rising energy prices have fuelled higher inflation. In response, major central banks worldwide have been tightening their monetary policy stances, with market participants expecting further interest rate hikes ahead (Chart 1, panel b). Higher inflation in combination with growing fiscal concerns has also translated into a considerable rise in long-term bond yields. Looking ahead, the resilience of the euro area economy to shocks could yet be tested as the factors cushioning the impact of energy supply disruptions, including the ability to draw on strategic oil reserves, fade, while trade policy uncertainty lingers at elevated levels despite coming down from earlier peaks (Chart 1, panel c). Any associated supply chain disruptions could pose upside risks to inflation and downside risks to economic growth, while they could also increase market volatility and challenge debt servicing capacity as financing costs rise.

Chart 1

The euro area economy has shown resilience to recent shocks, but tighter financial conditions and continued geoeconomic uncertainty are weighing on the outlook

a) Citigroup Economic Surprise Index

b) Policy rate expectations for end-2026

c) Oil geopolitical risk index and US trade policy uncertainty

(1 Jan. 2025-10 Nov. 2026, indices)

(1 Jan.-10 Nov. 2026, percentages)

(1 Jan. 2000-31 Oct. 2026, indices)

Sources: Bloomberg Finance L.P., Iacoviello and Tong*, Baker, Bloom and Davis** and ECB calculations.
Notes: Panel a: the Citigroup Economic Surprise Index measures the extent to which economic data releases exceed or fall short of market expectations. Values above (below) zero represent positive (negative) economic surprises in sum. Panel b: dashed lines represent actual policy rates. Panel c: the oil geopolitical risk index is a sub-index of the AI-GPR Index, identifying articles that discuss oil or energy supply disruptions driven by geopolitical events. Indices are shown as seven-day moving averages.
*) Iacoviello, M. and Tong, J., “The AI-GPR Index: Measuring Geopolitical Risk using Artificial Intelligence”, EmergingTech Economic Research Network (EERN) Research & Insights, Federal Reserve Research: Board of Governors, March 2026.

**) Baker, S., Bloom, N. and Davis, S., “Measuring Economic Policy Uncertainty”, The Quarterly Journal of Economics, Vol. 131, No 4, November 2016, pp. 1593-1636.

The financial stability outlook is also being shaped by an optimistic narrative surrounding artificial intelligence (AI) that is partly offsetting the adverse impact of the energy shock. In fact, optimism over AI’s potential to boost productivity has bolstered both financial market sentiment and economic growth (see Section 1.1). Such a transformative and general-purpose technological change does, however, also introduce potential systemic risks for financial stability. Markets remain vulnerable to broader sentiment shifts should news on the rate and form of AI adoption disappoint. Such disappointment could have adverse implications for corporate earnings expectations, with risks of spillovers across asset classes and geographies. Furthermore, AI financing remains opaque due to interlinked ownership structures and mutual purchasing and finance commitments inflating future revenues and heightening spillover risks from circular financing. At the same time, improved model capabilities have increased cyber risks and operational vulnerabilities more generally. These include those relating to the widespread reliance of financial and non-financial firms on a highly concentrated cluster of third-party providers (see Special Feature B).

In this context, three closely linked sources of risk and vulnerabilities for euro area financial stability stand out. First, financial market pricing rests on expectations of extraordinary earnings growth for AI firms, while the outlook for these firms and the global economy remains surrounded by significant risks. This leaves markets exposed to the possibility of an abrupt repricing which might be amplified by a highly uncertain geopolitical environment where AI-powered cyber threats are rising. Second, a challenging fiscal outlook in some major advanced economies could test investor confidence and lead to a further repricing of sovereign risk. Third, risks stemming from AI, geopolitics and sovereigns are exposing banks and non-banks alike to market, credit and funding risks which could be amplified by interconnection across sectors. In the current highly uncertain geoeconomic setting, the potential for these risks to materialise simultaneously, possibly amplifying each other further, increases the risks to financial stability.

A shift in the AI narrative could trigger an abrupt market repricing in a highly uncertain geopolitical environment

AI remains the central theme shaping global equity markets. Despite some volatility over recent months, equity markets have posted strong gains in the year to date. These gains were based on market optimism about the profitability of AI-linked firms, despite growing calls for slowing AI development and tighter regulation to address rising cybersecurity concerns posed by frontier models (see Special Feature B). That said, investor focus has become more selective in the AI space, with a stronger distinction between firms producing hardware, model developers and hyperscalers and growing signs of a broader risk reversal for AI valuations (see Chapter 2). Equity market developments are increasingly driven by extraordinarily strong earnings expectations (Chart 2, panel a). While major US tech firms have largely met these expectations in the recent past, equity prices depend largely on their ability to keep doing so going forward. There are concerns that these expectations may be exaggerated due to different forms of interlinkages – including ownership structures and circular financing – between AI-related firms (see Box 3). While the technology may turn out to be just as transformative as expected, it is also possible that cheaper, open-weight AI models will gain traction, jeopardising the profit trajectory for firms which are currently at the frontier. The AI investment boom is being funded in part by equity issuance, driving a significant surge of IPO volumes. This jump in issuance is now on track to surpass the dotcom bubble peak as a share of US GDP. Historically, such trends have often coincided with market peaks, signalling late-cycle exuberance (see Special Feature C).

Chart 2

Optimistic AI narratives have fuelled equity markets in recent years, but growing reliance on debt for infrastructure buildout raises the risk of repricing in both markets

a) Decomposition of S&P 500 returns

b) Debt issuance of large cloud and AI infrastructure providers

c) Global equity and euro area bond market valuations

(1 Jan. 2010-6 Nov. 2026, cumulative changes, percentage point contributions)

(2011-26, USD billions)

(1 Jan. 2005-10 Nov. 2026; left-hand scale: one-year forward P/E ratios, right-hand scale: z-spreads in basis points)

Sources: LSEG, Bloomberg Finance L.P. and ECB calculations.
Notes: Panel a: the decomposition is based on a dividend discount model. The model includes share buybacks, discounts future cash flows with interest rates of appropriate maturity and covers three expected dividend growth horizons. The residual from decomposing the returns into earnings expectations and discount factors, shown in green, can be interpreted as the contribution from the equity risk premium. A higher contribution means a lower equity risk premium. Positive contributions can be interpreted as risk-on sentiment, supporting equity prices beyond the contributions from earnings expectations and discount rates. “Start of AI boom” refers to end-October 2022. Panel b: data as at 10 November 2026. Debt issuance by Microsoft Corp., Meta Platforms Inc., Alphabet Inc., Amazon.com Inc. and Oracle Corp. Other currencies include the Swiss franc, pound sterling, Japanese yen and Australian dollar. Panel c: for equities (bonds), the boxplot shows the interquartile range of one-year forward price/earnings ratios (the latest z-spread) relative to the distribution since 2005. The previous edition of the ECB’s Financial Stability Review was published on 27 May 2026. JP stands for Japan; EA stands for euro area; IG stands for investment grade; HY stands for high yield.

Very large volumes of AI-related debt are being issued in global bond markets despite rising long-term yields. Much of the internal funding gap faced by AI firms is now increasingly being met in debt markets, with issuance for AI infrastructure buildout hitting new record highs this year (Chart 2, panel b). Further issuance is expected to be needed in the future to accommodate the cumulative capital expenditure of US big tech firms, projected to total USD 5.5 trillion by 2030. A significant share of recent AI-related issuance has been raised in euro, and euro area investors have increased their direct exposure (see Box 4). The scale and maturity profile of this debt has been raising concerns that it could drive long-term yields further upwards, potentially crowding out other corporate and sovereign borrowers, although evidence of such effects remains limited to date. Already, inflationary pressures stemming from the energy shock triggered by the conflict in the Middle East and fiscal expansion in major advanced economies have pushed yields higher globally in recent months, driven by higher real rates and term premia (see Section 2.2). Higher yields can pose risks to asset valuations by raising the rate at which future cash flows are discounted, weighing most on AI-related equities where earnings only start to be reaped further into the future. Additionally, as investment is increasingly debt-financed, higher yields raise debt service costs, particularly for more leveraged segments of the AI ecosystem that rely on floating-rate or private credit financing.

The AI narrative is vulnerable to geopolitical and macroeconomic risks which could trigger an abrupt repricing in both bond and equity markets. Equity valuation measures, notably of US tech firms, have moderated considerably on the basis of ambitious earnings expectations (Chart 2, panel c). At the same time, euro area corporate bond spreads have recently widened (see Box 2), as issuers increasingly compete for investor demand and refinancing at lower rates is no longer available. These vulnerabilities are interlinked and could be exposed by a range of triggers. A re-escalation of trade tensions or sharply deteriorating energy market outlooks would push up yields and inflation, weighing on both AI valuations and the cost of financing. The resultant repricing risks triggering abrupt sentiment shifts, with spillovers across asset classes and regions. A sell-off in financial markets triggered by a stagflationary shock would likely be broad based, as stock-bond correlations tend to turn positive under such conditions, which can amplify losses in diversified portfolios, even if sovereign bonds still serve a hedging role in risk-off episodes driven by other types of shock (see Chapter 2). Furthermore, growing market concentration among AI-related firms and sectors leaves not only global equity markets but also bond and private markets sensitive to shocks emanating from individual firms. A sustained rise in energy prices might also weigh on the optimistic AI narrative, as AI infrastructure and data centres are highly energy-intensive and sensitive to energy cost shocks.

A challenging fiscal outlook could test investor confidence and lead to a repricing of sovereign risk

Rising concerns about the sustainability of sovereign debt in major advanced economies are among the factors driving up long-term yields and increase the risk of global spillovers. Sovereign debt levels are elevated in several euro area countries and many major advanced economies. They are projected to rise further in the coming years, nearing the peaks observed during the pandemic (Chart 3, panel a). Signs of investor unease have already emerged in some advanced economy sovereign bond markets. Longer-maturity sovereign yields have risen sharply, driven by expectations of higher financing needs, higher term premia and renewed inflation concerns, reaching their highest levels in over two decades in several places (Chart 3, panel b). While markets have continued to function in an orderly manner, strains have prompted official action, including a US Treasury decision to increase buybacks of longer-dated bonds, although these only had short-lived impacts. Higher interest payments and structural headwinds to potential growth from weak productivity could further strain fiscal positions and rekindle market concerns about sovereign debt sustainability in highly indebted euro area countries. Stress in global benchmark bond markets or a highly indebted euro area country could prompt a broader reassessment of sovereign risk in the euro area as well. At the same time, and taking a longer-term perspective, market discipline can help contain vulnerabilities by providing price signals which encourage prudent fiscal policies and timely corrective policy adjustment.

Chart 3

Elevated funding needs, rising borrowing costs and fiscal slippage risks could strain already vulnerable governments and spark sovereign risk repricing

a) Sovereign debt-to-GDP ratios and projections in selected advanced economies

b) Frequency distributions of 10-year benchmark sovereign bond yields

c) Euro area sovereign CISS and economic policy uncertainty

(2000-30, percentages of GDP)

(1 Jan. 2002-10 Nov. 2026, percentages)

(Jan. 2014-Oct. 2026, z-scores)

Sources: IMF, LSEG, Garcia-de-Andoain and Kremer*, Baker, Bloom and Davis** and ECB calculations.
Notes: Panel b: latest refers to the average over the latest five business days. “Euro area” shows the GDP-weighted average of individual euro area countries’ yields. Panel c: both indices are computed as a GDP-weighted average for Germany, Spain, France and Italy. The sovereign Composite Indicator of Systemic Stress (CISS) combines data from the short end and the long end of the yield curve (two-year and ten-year bonds) for each country, i.e. two spreads between the sovereign yield and the euro swap interest rate, two realised yield volatilities and two bid-ask bond price spreads. The aggregation into country-specific sovereign CISS is based on time-varying cross-correlations between all homogenised individual stress indicators pertaining to each sovereign CISS variant following the CISS methodology developed by Holló et al.*** Z-scores are calculated from January 2000 onwards, Q2 2026 GDP data are used as a proxy for July to October 2026.
*) Garcia-de-Andoain, C. and Kremer, M., “Beyond spreads: measuring sovereign market stress in the euro area”, Working Paper Series, No 2185, ECB, October 2018.
**) Baker, S.R., Bloom, N. and Davis, S.J., “Measuring Economic Policy Uncertainty”, The Quarterly Journal of Economics, Vol. 131, Issue 4, November 2016, pp. 1593-1636.
***) Holló, D., Kremer, M. and Lo Duca, M., “CISS – a composite indicator of systemic stress in the financial system”, Working Paper Series, No 1426, ECB, March 2012.

Sovereign risks could emerge from higher issuance needs and shifts in the investor base. High defence spending and major infrastructure investment plans in some large euro area countries, notably including Germany, have driven sovereign issuance in 2026 to its highest levels since the pandemic (see Section 1.2). With the Eurosystem reducing its presence in euro area bond markets and a risk of future crowding-out effects from rising corporate bond supply by AI firms, the ability of the investor base to absorb this additional supply will be key for ensuring the orderly functioning of sovereign bond markets. Meanwhile, significant shifts in the investor base are under way, with foreign investors playing a growing role (see Section 1.2). Historical patterns suggest, however, that foreign investors tend to be flightier during periods of stress. Additionally, while supporting sovereign debt absorption and market liquidity in tranquil periods, the growing presence of more procyclical and often highly leveraged investors, such as hedge funds, in euro area sovereign bond markets may also amplify any sudden repricing of sovereign risk (see Box 7). Institutional investors’ demand for longer-dated debt is changing too, including in the context of the recent reform of the Dutch pension system or more recently due to the increased home bias of Japanese investors as rising yields make their domestic assets more attractive. Together with rising yields at the long or ultra-long end of the maturity spectrum, this shift in investor demand makes issuance of shorter-dated debt more attractive for sovereigns while simultaneously increasing rollover risk.

Heightened policy and geopolitical uncertainty, alongside longer-term structural challenges, could exacerbate sovereign vulnerabilities. Fiscal fundamentals remain susceptible to slippage with upcoming elections in major euro area countries later in 2026 and 2027 (Chart 3, panel c). Fiscal consolidation is overdue and increasingly important in an environment of growing presence of news-sensitive, leveraged investors in sovereign bond markets. Also, high energy prices in the context of heightened geopolitical tensions could amplify demands for fiscal support measures to cushion the impact on vulnerable households and firms. More severe and prolonged energy supply disruptions and weaker growth prospects could trigger a reassessment of sovereign risk by market participants. This could further constrain fiscal space, limiting governments’ ability to shield the economy from future shocks and to address structural challenges from digitalisation, low productivity, population ageing and climate change.

Uncertainties from AI, geopolitics and sovereigns expose banks and non-banks to market, credit and funding risks

The AI exposure of euro area banks and non-banks has increased, with non-bank vulnerabilities amplified by the interplay of rising portfolio concentration and leverage. Euro area non-banks have increased their exposures to AI significantly in recent years. This trend has started to extend beyond equities, with non-banks also purchasing debt issued by AI firms (see Section 4.1). A repricing in markets due to a possible reassessment of the AI narrative may trigger sudden valuation losses and spikes in investment fund share redemptions, increasing the risk of forced asset sales. High exposure concentration can be of particular concern when it meets growing leverage, as shown by the near collapse of Situational Awareness, a US hedge fund, in July 2026. Larger euro area equity funds have a higher share of their portfolios allocated to AI-related firms and are also increasingly exposed to risks from synthetic leverage in equity derivatives (Chart 4, panel a). AI exposures also remain heavily skewed towards US dollar-denominated assets, making the sector susceptible to valuation losses from any depreciation of that currency. Total direct AI-related lending by euro area banks has risen in recent years too but, it remains rather limited. However, there are signs of deteriorating credit quality in exposures to software firms and semiconductor producers (see Section 3.3). In the medium term, the indirect effects of an AI-driven transformation in the real economy may pose greater challenges for euro area banks, particularly those with significant exposures to firms whose financial performance or competitive position could weaken as AI adoption spreads (see Special Feature A), possibly increasing corporate default risk. In the medium to longer term, the increased use of agentic AI tools may also challenge the funding structure of banks by reducing the stickiness of deposits.

Chart 4

Growing AI and sovereign exposures could pose risks, especially when paired with leverage, with cross-sectoral spillover potential given rising bank-non-bank linkages

a) Synthetic leverage and AI-related securities holdings of euro area equity funds

b) Euro area sovereign debt transactions, by institutional sector

c) Euro area banks’ credit exposures to leveraged non-banks

(2022, H1 2026; x-axis: gross notional value over initial margin, y-axis: percentages)

(H1 2014-H1 2026, € trillions)

(Q4 2021, Q4 2025; left-hand scale: € billions, right-hand scale: percentages)

Sources: ECB (EMIR, SHS, AnaCredit, RIAD), LSEG Lipper and ECB calculations.
Notes: Panel a: synthetic leverage is measured as the gross notional value of equity derivative positions divided by initial margin for each tercile, weighted by the group’s assets under management. The vertical axis shows holdings of securities issued by firms involved in AI model development, chip production, cloud provision, data centres and related infrastructure, or utilities. Bubble size indicates the absolute gross notional amount of equity derivatives for each group. Panel b: transactions are defined as the sum of purchases minus sales of euro area sovereign debt securities, recorded at transaction value in euro including accrued interest. ICPFs stands for insurance corporations and pension funds. Panel c: intragroup exposures are excluded from leveraged NBFI exposures. CRE stands for commercial real estate; REIT stands for real estate investment trust.

Sovereign exposures of banks and certain non-banks have increased, exposing them to valuation losses from rising interest rates. In fact, non-banks have continued to be strong absorbers of new sovereign debt issuances (Chart 4, panel b). They have focused primarily on the short end of the yield curve while showing less appetite for longer maturities, given expectations of rising interest rates and fiscal concerns. Additionally, global hedge funds are highly active in European sovereign bond markets, using euro area sovereign bonds as collateral in repo markets and other leveraged trading strategies (see Box 7). They could amplify price swings in times of sudden and substantial bond price movements, increasing the likelihood of sharp non-linear yield spikes. Banks’ sovereign exposures have also increased in recent years, driven by a shift in the composition of their liquidity buffers (see Section 3.2). Rising sovereign exposures coupled with shifts in sovereign risk perceptions have reignited market concerns about the re-emergence of the sovereign-bank nexus. Much of this increase in sovereign holdings has, however, been driven by investments in non-domestic government bonds, which has reduced home bias. While this diversification somewhat mitigates concerns about the sovereign-bank nexus, higher sovereign exposures also expose euro area banks to valuation losses on their sovereign portfolios subject to mark-to-market accounting in the context of rising sovereign yields.

Banks and non-banks may see their credit risks increase if macro-financial conditions deteriorate markedly as a result of the Middle East war. The quality of euro area bank assets has remained sound, with the aggregate non-performing loan ratio touching a record low of 2.2% in September 2026, albeit with some variation at country and sector levels (see Section 3.3). Asset quality risks could increase as the full impact of the oil supply shock unfolds, however, especially for euro area firms operating in sectors that are concurrently energy, interest rate and trade sensitive (see Section 1.3). Moreover, cost pressures from rising energy prices and higher interest rates may be compounded by continued tariff-related uncertainty and the risk of broader supply chain disruptions. The resilience of euro area households could also be tested if stress in vulnerable parts of the corporate sector drives up unemployment. Also, cost-of-living pressures from higher energy prices and inflation may weaken households’ debt servicing capacity, especially among lower-income groups. Banks could face higher provisioning costs as a result. Increasing credit risk as a result of a weaker economic outlook and renewed energy price pressures may also weigh on valuations in non-bank portfolios. Non-banks’ holdings of highly valued equities and long-duration bonds are particularly vulnerable to rising interest rates and risk premia, and could suffer simultaneous losses.

Rising bank-non-bank interlinkages could facilitate cross-sectoral spillovers if market conditions turn volatile. The importance of non-bank funding for banks has continued to grow, exposing banks to rollover and redemption risks. These risks are especially pronounced for global systemically important banks which have substantial short-term liabilities towards non-banks. A growing share of these liabilities stems from cross-border activities, rendering banks more vulnerable to sudden funding withdrawals triggered by shocks affecting non-banks domiciled abroad. On the assets side, euro area banks’ credit exposures to non-banks have also risen markedly, driven primarily by lending to investment funds in general and hedge funds in particular (Chart 4, panel c). Asset-price and liquidity shocks affecting non-banks may lead them to reduce funding to banks. At the same time, banks may curtail lending to such entities under stressed market conditions, forcing them to liquidate trading positions and amplifying volatile market movements.

Banks are well positioned to absorb potential adverse shocks, supported by strong profitability coupled with ample capital and liquidity buffers. Banks’ return on equity has reached its highest level in a decade, with a positive earnings outlook in the context of rising interest rates also translating into higher bank valuations (see Section 3.1). Bank resilience has also been underpinned by capital and liquidity ratios that are well above regulatory requirements, also thanks to the regulatory reforms implemented since the global financial crisis. Preserving bank resilience remains a key priority for macroprudential policy in the context of elevated geopolitical and trade policy uncertainty. Releasable capital buffers should be kept to ensure that relevant authorities are able to respond to severe shocks or bank capital constraints. Furthermore, existing borrower-based measures should be maintained to uphold lending standards and mitigate the build-up of vulnerabilities among borrowers (see Section 3.5). Completing the banking union and reducing the complexities of the existing regulatory and supervisory framework, without compromising bank resilience, will be key to enhancing the competitiveness of EU banks.

Persistent liquidity and leverage vulnerabilities in parts of the NBFI sector require a macroprudential policy response to safeguard market liquidity during periods of stress. Previous stress episodes, such as the March 2020 market turmoil or the liquidity stress in the UK gilt market in September 2022, have shown how NBFI vulnerabilities can be amplified if the NBFI sector holds similar concentrated positions and draws on the same sources of liquidity. This underscores the importance of appropriately calibrated measures to address vulnerabilities related to leverage, liquidity mismatches and margin preparedness as well as risks arising from crowded trades and concentrated positions. Progress will also depend on improving data availability and cross-border information sharing, which are essential for assessing sources of systemic risk arising from cross-border hedge funds in European sovereign bond markets. Also, accelerating progress on the EU’s savings and investments union – by deepening equity markets, mobilising savings, reducing insolvency and tax barriers to cross-border investment as well as strengthening integrated supervision – will be essential to support EU-wide growth and competitiveness while safeguarding financial stability.

Financial stability vulnerabilities remain high, with risks from AI valuations, sovereigns and geopolitics looming

All in all, risks from AI valuations, sovereigns and geoeconomic tensions and are weighing on the outlook for euro area financial stability. The AI investment boom raises concerns around earnings expectations which might be disappointed, high market concentration, late-cycle exuberance reminiscent of past investment booms and growing cyber vulnerabilities in a fragile geopolitical landscape. AI-related debt issuance has surged, potentially contributing to rising global bond yields, which could crowd out other borrowers. Meanwhile, sovereign risks are growing as elevated debt levels, high issuance needs and rising yields challenge fiscal sustainability, putting global sovereign bond markets under pressure. Banks and non-banks alike have increased their AI and sovereign exposures, with risks of valuation losses being compounded by rising leverage and interlinkages across sectors. These vulnerabilities are strongly connected and at risk of materialising simultaneously. More generally, vulnerability constellations appear to be more elevated for certain non-banks than they are for banks due to greater risk exposures and a less developed prudential framework for building resilience against shocks.

In addition, several cross-cutting structural challenges remain critical for financial stability, with the potential to amplify existing cyclical vulnerabilities. These include rising vulnerabilities associated with climate change and nature degradation, notably the materialisation of physical risks as evidenced by a series of extreme weather events this summer like wildfires, heatwaves and droughts. While these events likely had only a limited and temporary negative impact on overall economic activity in the euro area, the increasing frequency of physical climate risks is a growing concern. Additionally, further concerns include the risk that global regulatory fragmentation and deregulation could erode the achievements of the regulatory reforms since the global financial crisis, as well as challenges linked to ageing populations. The potential for cyclical and structural vulnerabilities to crystalise simultaneously and amplify each other heightens the materiality of risks to euro area financial stability.

1.1 Euro area economy balances resilience and headwinds

A fragile geoeconomic environment is posing headwinds for the euro area economy, with disruptions to energy supplies especially concerning. A re-escalation of the conflict in the Middle East – after it had shown some signs of easing during the summer of 2026 – and the ongoing war between Russia and Ukraine are weighing on global energy markets. Supply disruptions have pushed up the prices of crude oil and natural gas. This is hitting Europe at a sensitive moment, as gas storage levels are historically low ahead of the winter, and European buyers need to compete with Asia for deliveries of liquefied natural gas. Crude oil inventories have also declined. The impact of both conflicts on global refining capacity has resulted in a significant increase in the crack spread, which measures the differences between crude oil prices and wholesale petroleum product prices (Chart 1.1, panel a). In addition, climate change-related extreme weather events, such as the heatwaves and wildfires experienced in Europe over the summer, have dampened economic activity, with low water levels in key waterways disrupting shipping, among other things. Smaller crop yields have also resulted in higher food prices. Together with the El Niño weather phenomenon, which usually causes global agricultural goods prices to rise over the following 12 months, this could extend price pressures into 2027 (Chart 1.1, panel b).

Chart 1.1

Geopolitical conflicts and extreme weather events are driving energy and food price surges, adding to inflationary pressures and dragging on euro area economic activity

a) Crude oil and natural gas prices, and crack spread

b) Growth in global agricultural prices

(1 Jan. 2021-10 Nov. 2026; USD/bbl, €/MWh)

(7 Jan. 2000-6 Nov. 2026, percentages)

Sources: Bloomberg Finance L.P. and ECB calculations.
Notes: Panel a: the crack spread shown is the 3-2-1 spread based on European benchmarks. It assumes that three barrels of Brent crude oil are refined into two barrels of Eurobob gasoline and one barrel of gasoil/heating oil. Natural gas prices show the Dutch TTF natural gas forward day ahead. Crude oil prices and the crack spread are measured in US dollars per barrel, whereas natural gas prices are measured in euro per MWh. Panel b: the chart shows year-on-year growth rates in the Bloomberg Agriculture Spot Index, which measures the price movements of agricultural commodities included in the Bloomberg Commodity Index.

Investment in artificial intelligence (AI) infrastructure is supporting global growth, but trade frictions are persisting and debt sustainability concerns are on the rise. The buildout of AI infrastructure has become a key driver behind global growth and is acting as a counterweight to the adverse impact of energy supply disruptions. In particular, demand for AI-related technology products is bolstering global commerce. Nevertheless, euro area businesses continue to face challenges. First, tariff barriers have hampered trade in goods with the United States. Second, continued weak domestic demand in China is pushing Chinese products into euro area markets while constraining imports from the euro area (Chart 1.2, panel a). Third, the combination of mounting sovereign debt and persistent fiscal deficits across major advanced economies has heightened investor concerns over the long-term sustainability of global sovereign debt (Chart 1.2, panel b). Among other factors, this has led to substantial increases in long-term sovereign bond yields (see Chapter 2). If sustained, these issues could spill over into broader financing conditions for the real economy, with dampening effects on investment and activity.

Chart 1.2

AI investment is boosting global growth but trade frictions and debt sustainability concerns are weighing on the euro area outlook

a) Euro area goods trade balance

b) Sovereign debt across advanced economies

c) Share of AI adoption and hours saved per week, by employment sector

(Q1 2005-Q2 2026, € billions)

(Q1 2005-Q2 2026, USD trillions)

(Q2 2026; percentages, hours)

Sources: Eurostat, ECB (CES), IIF and ECB calculations.
Notes: Panel a: the chart uses data from external trade in goods statistics, which are based on customs and Intrastat data, and differ conceptually from balance of payments statistics. “Others” refers to the difference between countries shown in the chart and the total external euro area goods trade balance. Panel c: based on responses to the annual labour market module of the ECB’s Consumer Expectations Survey. Net sentiment measures the difference between positive and negative views. Larger bubbles indicate a more positive net sentiment.

The euro area economy has proven robust to shocks, but prolonged headwinds could yet test this resilience. Despite the challenges posed by energy price shocks, trade frictions and extreme weather events, euro area activity has maintained its moderate pace, with quarterly GDP growth standing at 0.4% in the third quarter of 2026.[1] This resilience was supported by domestic consumer demand and public investment, linked partly to defence and infrastructure spending. Nonetheless, there are severe downside risks as disruptions in energy markets are continuing. Even if there were to be a resolution of the Middle East conflict, normalisation would still take a substantial amount of time given the damage that has been inflicted on infrastructure and refineries. This, together with climate-related surges in global food prices and rising prices for AI-related technology products, is sustaining inflationary pressures, both in the euro area and globally. A prolonged period of above-target inflation, especially if energy prices feed through to broader goods and services, will maintain upward pressure on global policy rates and, eventually, on financing costs for firms, households and sovereigns. Over the medium term, increasing AI adoption and the associated productivity gains could lead to positive growth surprises (Chart 1.2, panel c and Special Feature A). However, the impact that this will have on productivity, labour markets and investment is still surrounded by high uncertainty, and it is not yet clear whether the consequences for employment and inflation will be net positive or negative.

1.2 Risks of global spillovers are adding to fiscal challenges

Rising debt levels and persistent deficits in some euro area countries remain key vulnerabilities and are limiting fiscal space to react to shocks. After a phase of fiscal consolidation following the COVID-19 pandemic, debt levels are consistently increasing again in many euro area countries. Persistent budget deficits well in excess of the limit set by the EU fiscal rules are underpinning the rising debt trajectory of some already highly indebted countries (Chart 1.3, panel a). These dynamics are limiting the ability of governments to address shocks that come on top of long-standing fiscal challenges such as low productivity, ageing populations and the green and digital transitions. At the same time, some countries do not have a clear fiscal consolidation path going forward, and any fiscal slippage in certain major euro area countries could test investor confidence. In this context, a rigorous interpretation and application of the fiscal rules by the European Commission will be essential to safeguard the credibility of the EU fiscal framework and help place public finances on a sustainable path.

Chart 1.3

Fiscal positions are under pressure in parts of the euro area and face the challenges of defence spending, energy supply disruptions and extreme weather events

a) Change in debt-to-GDP ratio and general government budget balance

b) Number of energy support measures

(Q2 2025-Q2 2026; percentages, percentage points)

(Sep. 2026, total)

Sources: ECB and Eurostat (GFS), OECD and ECB calculations.
Notes: Panel a: the budget balance is shown as annualised quarterly data based on four-quarter trailing sums. Panel b: the number of energy support measures is taken from the OECD energy support measures tracker, which was last updated on 23 September 2026.

Defence spending and energy security are now key objectives owing to ongoing geopolitical conflicts, including in the Middle East. Investments in defence and infrastructure, driven in particular by Germany, have been supporting GDP growth and are expected to continue to do so over the medium term.[2] Once the extraordinary spending measures have expired, however, the positive impact on growth will start to wane. In addition, the end of disbursements under the Next Generation EU programme at the end of 2026 may weigh on public investment. In light of the protracted conflict in the Middle East, governments may face pressure to prolong energy support measures, including reductions in taxes on fuel and other energy products (Chart 1.3, panel b). Such measures have accounted for a moderate 0.1% of GDP in 2026, but they need to remain temporary, targeted and tailored so as not to risk derailing fiscal positions. Heatwaves and wildfires across euro area countries during the summer have underscored the need to adapt to climate change. As the insurance protection gap widens, there is a risk that without functioning public-private risk sharing, governments could need to provide fiscal backstops, leading to several challenges becoming increasingly intertwined. For instance, investments in renewable energy would benefit both energy security and affordability, as well as climate protection. In this regard, the European Commission has proposed broadening the scope of the national escape clause, which provides exemption from certain fiscal rules for defence expenditure, to include energy-related measures.[3]

High issuance needs and rising yields across the globe are increasing debt service costs and putting pressure on government budgets. Debt issuance by euro area governments has reached its highest level since the pandemic, driven predominantly by the German defence and infrastructure package (Chart 1.4, panel a). At the same time, sovereign yields, especially for longer maturities, have risen substantially in 2026, owing in part to concerns over the long-term debt sustainability of several major advanced economies, including the United States (see Section 2.2). In the euro area, many countries face the highest financing costs since the sovereign debt crisis of 2012 as sovereign bond yields have increased. While euro area-wide and global factors, including rising risk-free rates, have played a major role in pushing up yields, the underlying country-specific drivers can differ. For some sovereigns, the higher yields may partly reflect a stronger macroeconomic outlook, while for others, they may stem more from fiscal sustainability concerns and the associated risk premia. A widening of French sovereign credit spreads points towards investor perceptions of country-specific risks, although markets have continued to function in an orderly manner. A higher-for-longer interest rate environment could therefore lead to greater differentiation in financing conditions across sovereigns. Against this background, interest burdens are continuing to rise as maturing debt is rolled over at significantly higher interest rates (Chart 1.4, panel b). To mitigate the impact of rising long-term yields, governments have gradually shifted towards issuance of shorter maturities. However, this increases sensitivity to shocks as debt needs to be rolled over more frequently.

Chart 1.4

High issuance needs and rising yields are pushing up interest burdens as shifts in the investor base are increasing sensitivity to global market sentiment

a) Cumulative debt issuance by euro area governments

b) Government interest payments

c) Holdings of euro area sovereign debt securities, by institutional sector

(Jan. 2017-Sep. 2026, € billions)

(Q4 2021, Q2 2026; percentages of GDP)

(2022, Q2 2026; percentages)

Sources: ECB and Eurostat (GFS, QSA) and ECB calculations.
Notes: Panel a: “pre-pandemic” refers to the average over the period 2017‑19; “pandemic” refers to the average over the period 2020‑21; “post-pandemic” refers to the average over the period 2022‑24. Panel b: government interest payments are shown as annualised quarterly data based on four-quarter trailing sums. Interest payments are consolidated, meaning net of payments within the government sector. Panel c: MMF stands for money market fund; ICPFs stands for insurance corporations and pension funds.

The possibility of global spillovers is compounding domestic risks and could lead to a further repricing of sovereign risk. Rising interest burdens are adding to acute and more medium to long-term fiscal challenges in the euro area. Without credible consolidation efforts, there is a risk that rising interest payments contribute to government spending outpacing economic growth in some euro area countries, which could ultimately result in unsustainable debt trajectories. Investor sentiment could turn negative if such dynamics become increasingly likely or if adverse rating actions point towards a rise in sovereign risks. In addition, shifts in the investor base are making euro area sovereign bond markets more sensitive to global market conditions. The Eurosystem continues to passively reduce its footprint in sovereign bond markets, and institutional investors’ demand for longer-dated sovereign debt has declined owing to structural shifts such as the Dutch pension fund reform. In this environment, more volatile and price-sensitive actors, such as foreign investors and hedge funds, account for an increasing share of the investor base (Chart 1.4, panel c). Accordingly, even without changes in the underlying fundamentals, the pricing of euro area sovereign risk could become more sensitive to global bond market sentiment and portfolio rebalancing, as well as the financing conditions of leveraged actors. Any repricing of sovereign risk could result in a rapid tightening of credit conditions for corporates and banks. This would have an adverse impact on growth and add to the downside risks to macroeconomic activity.

1.3 AI-related exposures add to corporate risks among firms sensitive to interest rates, energy costs and trade

Financing conditions for the euro area corporate sector have tightened, while credit volumes have remained resilient but lost momentum. The overall cost of financing for euro area non-financial corporations (NFCs) rose, driven mainly by market-based funding, while increases in the cost of bank borrowing remained more contained. Survey responses covering the third quarter of 2026 point towards widespread risk aversion limiting the supply of credit, alongside subdued demand for long-term credit. Credit standards for loans to firms have tightened further, with banks citing higher perceived risk related to the economic outlook and borrowers’ exposure to geopolitical tensions and energy supply shocks.[4] On the demand side, survey evidence from earlier in the year indicated that firms were postponing or scaling back fixed investment while increasing inventories and, in some cases, building strategic stockpiles of energy and critical inputs as a hedge against possible energy and supply disruptions.[5] In the third quarter of 2026, higher interest rates and weaker demand for fixed investment among SMEs dampened demand for bank borrowing, as firms increasingly relied on internal funds while financing needs for working capital remained elevated. However, business investment is expected to grow in the second half of 2026, supported by spending on artificial intelligence (AI) technology, digitalisation and defence, although tighter financing conditions could weigh on the recovery.[6]

Corporate vulnerability indicators have deteriorated, although the aggregate balance sheet of the sector does not point towards stress. Corporate vulnerabilities have edged up over recent quarters, driven mainly by higher trade liabilities and, to a lesser extent, by weaker profit margins and liquidity positions (Chart 1.5, panel a). Debt servicing capacity has held up better than in previous tightening cycles. This reflects a still relatively low share of debt at floating rates in several jurisdictions, the lengthening of maturities that took place during the period of low interest rates and the accumulation of liquidity buffers. Overall, the composite indicator remains far below the levels observed in the aftermath of the global financial crisis or the sovereign debt crisis.

Rising bankruptcies point to growing pressure among vulnerable firms, even as market-based indicators of credit risk for listed companies have improved. Corporate bankruptcies continued to rise across sectors, with the sharpest increases recorded in education and health care, transport as well as finance, real estate and professional services (Chart 1.5, panel b). While rising bankruptcies can partly reflect firm turnover and the reallocation of economic activity rather than a general deterioration in firms’ financial soundness,[7] declining business registrations in education and health care, as well as in finance, real estate and professional services, point to net firm exit rather than turnover. By contrast, the distance to insolvency for listed firms has recovered after deteriorating at the start of the Middle East conflict (Chart 1.5, panel c). These two developments are not contradictory: listed firms are typically larger and enjoy better access to financing than smaller firms, which account for the majority of bankruptcies. Furthermore, market-based indicators of credit risk are derived from equity valuations and volatility, so recent improvements partly reflect buoyant market sentiment and could reverse abruptly in case of a market correction. Overall, the data indicate uneven rather than broad-based corporate distress, with pressure on vulnerable firms likely to intensify if weak demand, high energy costs and tighter financing conditions persist.

Chart 1.5

Corporate vulnerabilities are rising but remain below long-term averages, while default risk for listed firms has eased since the outbreak of the Middle East conflict

a) Corporate vulnerability indicator

b) Corporate bankruptcies

c) Distance to insolvency for NFCs in the STOXX Europe 600

(Q1 2019-Q2 2026, z-scores)

(Q1 2022-Q2 2026, index: Q4 2019 = 100)

(Jan.-Oct. 2026; standard deviations, percentages)

Sources: ECB, Eurostat, LSEG and ECB calculations.
Notes: Panel a: for details of the construction of the composite indicator of corporate vulnerabilities, see Gardó et al.* Positive (negative) values indicate higher (lower) vulnerability. Panel c: for details on the calculations of the distance to insolvency, see Atkeson et al.** The share of vulnerable firms is the percentage of firms below the vulnerability threshold, i.e. a distance to insolvency of 1.5.
*) Gardó, S., Klaus, B., Tujula, M. and Wendelborn, J., “Assessing corporate vulnerabilities in the euro area”, Financial Stability Review, ECB, November 2020.
**) Atkeson, A.G., Eisfeldt, A.L. and Weill, P.-O., “Measuring the financial soundness of U.S. fir6s, 1926−2012”, Research in Economics, Vol. 71, Issue 3, September 2017, pp. 613-635.

Firms in AI-intensive sectors have so far faced more benign financing conditions for both the price and the volume of external funding. These firms have generally retained easier access to market-based funding, with debt issuance and equity stocks well above the levels reported for firms in sectors with medium and low AI intensity (Chart 1.6, panels a and b). Moreover, AI-exposed firms earn higher equity returns and enjoy more compressed corporate bond spreads than otherwise comparable firms.[8] However, the recent global sell-off in AI-related assets, especially among highly leveraged infrastructure providers, shows how financing conditions can tighten when confidence in the returns on AI investment weakens. Conversely, firms in AI-intensive sectors rely less on bank lending (Chart 1.6, panel c). This reflects both their cheaper access to market-based funding and the fact that AI-related investment is focused on intangible assets, which are harder to collateralise and therefore less suited to bank financing.[9]

Chart 1.6

Firms in AI-intensive sectors tap debt and equity markets, while other firms rely mostly on bank loans

a) Euro area NFCs’ gross debt issuance, by sectoral AI intensity and maturity bucket

b) Equity stock by euro area NFCs, by sectoral AI intensity

c) New loans to euro area NFCs, by sectoral AI intensity

(Jan. 2023-July 2026; € billions, percentages)

(Jan. 2023-July 2026; € billions, percentages)

(Jan. 2023-July 2026; € billions, percentages)

Sources: ECB (AnaCredit, CSDB), Moody’s, Calvino et al.* and ECB calculations.
Notes: The classification of firms by AI intensity is based on Calvino et al.* The green dots represent the ratio of the total volume to firm size, proxied by total assets in the previous fiscal year. Panel a: the ratio for 2026 is estimated assuming that gross debt issuance over the period from August to December 2026 is equal its 2025 level. Panel b: the chart uses equity stock at the end of the time period. Panel c: new loans to euro area NFCs are measured by total outstanding nominal amount excluding syndicated loans. The ratio for 2026 is estimated assuming that new lending over the period from August to December 2026 is equal to its 2025 level.
*) Calvino, F., Dernis, H., Samek, L. and Ughi, A., “A sectoral taxonomy of AI intensity”, OECD Artificial Intelligence Papers, No 30, OECD Publishing, December 2024.

Reliance on global supply chains and energy-intensive infrastructure leaves AI-intensive firms vulnerable in the present geopolitical environment. Their market valuations appear to be less sensitive to conventional shocks than those of comparable firms in less AI-intensive sectors (Chart 1.7, panel a). However, these firms are deeply embedded in global supply chains and their bond and equity valuations are affected by global geopolitical risk shocks at least as much as those of other firms. Their dependence on the availability and price of energy for data centres and related infrastructure also exposes them to energy supply disruptions, a risk that has become more significant since the start of the Middle East conflict. Escalating trade tensions, export restrictions on critical technologies or a renewed energy shock could thus weigh on their earnings prospects. In addition, their valuations rest on expectations of uncertain returns on AI investments, and AI exposure is priced in equity and corporate bonds well beyond the technology sector. Such shocks could therefore also trigger an abrupt and broad-based repricing, affecting firms outside the core of the AI investment cycle. This risk is amplified by the increased reliance of global growth on AI-related investment and trade, particularly between the US and Asian economies. Disruptions to growth drivers could propagate through global value chains, weakening external demand and heightening uncertainty for euro area firms.

Chart 1.7

Build-up of risks from AI exposure, interest rate hikes and higher energy prices

a) Relative sensitivity to shocks of bonds and equities for high AI-intensity sectors

b) Estimated volumes of euro area NFC loans due for repricing from 2026 to 2030

(Jan. 2019-Aug. 2026, percentages)

(2026-30, percentages)

Sources: LSEG, Moody’s, Bloomberg Finance L.P., Calvino et al.*, Jarociński and Karadi**, Bondarenko et al.***, Iacoviello and Tong****, ECB (AnaCredit, MIR) and ECB calculations.
Notes: Panel a: non-financial corporations (NFCs) in the STOXX Europe 600 in high AI-intensity sectors are compared with firms in medium and low AI-intensity sectors based on the sectoral taxonomy set out by Calvino et al.* The chart shows the relative sensitivity of 1 standard deviation in selected shocks on option-adjusted corporate bond spreads and equity returns, for firms in high AI-intensity sectors. The relative sensitivity is calculated as 100*(βhigh/βmed, low-1). Positive (negative) values indicate that firms in high AI-intensity sectors are more (less) sensitive to the shock than firms in medium and low AI-intensity sectors. The whiskers indicate 95% confidence intervals. The estimates are obtained from regressions with security and date-by-sector fixed effects, controlling for leverage and firm size; bond regressions are weighted by bond outstanding amount, whereas equity regressions are weighted by market capitalisation. Entropy balancing re-weights observations in order to align pre-shock characteristics across the two comparison groups. Characteristics are residual maturity, expected default frequency and total assets for bonds; EBITDA, interest coverage ratio and total assets for equities. “MP shock” refers to the pure monetary policy shock of Jarociński and Karadi**; “EA GPR” refers to the euro area geopolitical risk index of Bondarenko et al.***; “World GPR – oil” and “World GPR – non-oil” refer to the global geopolitical risk indices of Iacoviello and Tong****. Panel b: loans in fixation buckets above one year are assigned a single assumed reset date: three years for one to five years, 7.5 years for five to ten years and 15 years for over ten years. Loans are counted at origination value, without amortisation or prepayment. 2026 covers October to December.
*) Calvino, F., Dernis, H., Samek, L. and Ughi, A., “A sectoral taxonomy of AI intensity”, OECD Artificial Intelligence Papers, No 30, OECD Publishing, December 2024.
**) Jarociński, M. and Karadi, P., “Deconstructing Monetary Policy Surprises – The Role of Information Shocks”, American Economic Journal: Macroeconomics, Vol. 12, No 2, April 2020, pp. 1-43.
***) Bondarenko, Y., Kang, N., Lewis, V., Rottner, M. and Schüler, Y., “Geopolitical risk in the euro area: Measurement and transmission”, Discussion Papers, No 14/2026, Deutsche Bundesbank, May 2026.
****) Iacoviello, M. and Tong, J., “The AI-GPR Index: Measuring Geopolitical Risk using Artificial Intelligence”, Technology & Growth Center – Research & Insights, Federal Reserve Bank of San Francisco, March 2026.

Debt rollover and the repricing of loans originated in a low interest rate environment imply rising interest burdens for euro area firms. The pass-through of higher interest rates to firms’ debt servicing costs is still incomplete, with over 10% of the outstanding loan volume taken on by euro area NFCs expected to be repriced each year between 2027 and 2029 (Chart 1.7, panel b). Roughly half of these loans were originated before 2022, at interest rates well below current levels. As these loans reprice at higher rates, interest expenses will rise. This implies higher financing needs and cost pressures, particularly for more leveraged and interest-sensitive firms, leaving less scope for internal funds to absorb additional shocks.

Extreme climate events and renewed energy price pressures are compounding these vulnerabilities. The droughts and wildfires recorded during the summer of 2026 disrupted economic activity in several euro area countries by damaging physical assets and interrupting production. With only around one-quarter of losses from natural catastrophes being insured in Europe over recent decades, a substantial share of the resulting losses is borne directly by firms, eroding their liquidity buffers and debt servicing capacity.[10] Small firms, which tend to be less insured and have more limited access to external finance, are particularly exposed. In turn, losses that are not absorbed by firms or insurers can spill over to banks through higher credit risk and to sovereigns through contingent liabilities. Looking ahead, the developing El Niño event is raising the likelihood of droughts and heat extremes, and hence of stronger and more frequent destructive events. Meanwhile, cost pressures are intensifying amid rising energy prices, low gas storage levels ahead of the winter and renewed disruptions to energy deliveries through the Strait of Hormuz, where traffic has partly recovered but shipping remains under attack. Such shocks are transmitted with a lag to the financial sector: corporate default expectations react first, with bank asset quality and lending volumes affected with a delay of up to two years. This implies that the financial stability effects of the Middle East conflict may only start to materialise in the course of 2027 (see Box 5).

1.4 Weaker labour demand and higher energy costs may challenge vulnerable households

Household vulnerabilities remain contained, underpinned by resilient employment and broadly stable incomes. The household vulnerability indicator does not show signs of a broad deterioration in debt servicing capacity (Chart 1.8, panel a), while the unemployment rate remains low and stable. Real disposable income continued to increase in the first half of 2026, supported by higher employment levels, nominal wages and other sources of income. However, these gains have been partly offset by direct taxes and inflation, which slowed growth in real disposable income in the second quarter of 2026 (Chart 1.8, panel b). These developments continue to bolster households’ ability to service their debt, mitigating immediate risks to banks’ mortgage portfolios. By contrast, consumer confidence weakened in September and October, and it remains below the levels seen prior to the Middle East conflict (Chart 1.8, panel c). Moreover, the overall resilience suggested by aggregate indicators may be masking potential vulnerabilities among households with lower incomes, limited liquidity buffers or adjustable-rate mortgages, as these households are more exposed to higher energy prices and rising borrowing costs.

Chart 1.8

Euro area households’ resilience is underpinned by low unemployment and stable real disposable income growth

a) Household vulnerability indicator

b) Real disposable income growth and contributions

c) Labour market indicators and consumer confidence

(Q1 2005-Q2 2026, z-scores)

(Q1 2020-Q2 2026; percentages, percentage points)

(Jan. 2020-Oct. 2026; percentage points, index)

Sources: S&P Global Market Intelligence, European Commission, Eurostat and ECB (QSA, MNA) and ECB calculations.
Notes: Panel a: the composite indicator is based on a broad set of indicators along five dimensions: (i) debt servicing capacity (measured by gross interest payments/income ratio, saving ratio and expectation of personal financial situation); (ii) leverage (gross debt/income and gross debt/total assets ratios); (iii) financing (bank lending rate, short-term debt/long-term debt ratio, quick ratio (defined as current financial assets/current liabilities) and credit impulse (defined as the change in new credit issued as a share of GDP)); (iv) income (real income growth and income/GDP ratio); and (v) activity (labour participation rate and unemployment expectations). Positive (negative) values indicate higher (lower) vulnerability. Panel b: “Other income” includes gross operating surplus and mixed income, property income, social contributions, social benefits and other current transfers. Panel c: the change in the unemployment rate is the difference in year-on-year growth rates from the previous year. The latest observations are for June 2026 for the job vacancy rate, September 2026 for the change in the unemployment rate and October 2026 for the consumer confidence indicator.

Repayment capacity could erode gradually if labour market conditions or real incomes were to weaken. Employment and labour force growth have continued to slow, while labour demand appears to be cooling, as shown by declining vacancy rates (Chart 1.8, panel c). The broadly stable unemployment rate is thus more likely a reflection of firms’ reluctance to lay off employees than of a buoyant labour market. A softer labour market could limit household income growth, while higher energy prices and their indirect effects on other consumption items would further compress real purchasing power.

Euro area households remain resilient on aggregate but have not fully recovered from the confidence shock caused by the Middle East conflict. Employment prospects have improved in the manufacturing sector, although those in the construction sector remain weak (Chart 1.9, panel a). Consumers’ expectations for their financial and general economic situation have also recovered somewhat in recent months but remain subdued (Chart 1.9, panel b). Low consumer confidence is also dampening demand for credit for house purchases. At the same time, the continued weakness in expectations suggests that households may sustain high levels of precautionary savings and delay major purchases, thereby bolstering their liquidity buffers. In a stress scenario, however, any renewed deterioration in confidence combined with tighter credit supply could weigh on house prices and on the ability of households to refinance or service their debt. This would apply in particular to those more reliant on borrowing to finance housing or durable goods purchases.

Chart 1.9

Business and consumer sentiment is resilient despite global shocks

a) PMI for employment in the euro area

b) Consumers’ expectations for the economy

c) Household expectations on being late on payments

(Jan. 2021-Oct. 2026, indices)

(Jan. 2020-Oct. 2026, indices)

(Q1 2021-Q3 2026, percentages)

Sources: S&P Global Market Intelligence, European Commission, ECB (CES) and ECB calculations.
Notes: Panel a: PMI values above (below) 50 imply an improvement (deterioration) in employment. Panel b: values above (below) 0 imply an improvement (deterioration) in expectations. Panel c: the chart shows the share of respondents in the first and second income quantiles who expect to be late on rent and mortgage payments over the next three months, as well as the share of respondents across all income quartiles who expect to be late on utility bills over the next three months.

Financial difficulties remain concentrated among lower-income households, indicating pockets of vulnerability rather than widespread household distress. The share of households expecting to fall behind on rent or mortgage payments remains substantially higher in the lower-income groups, while reported delays in utility payments remain limited despite elevated energy prices (Chart 1.9, panel c). Delayed mortgage payments could directly affect banks’ asset quality, while rent arrears could reduce landlords’ rental income and indirectly affect real estate exposures. The risk would become more material if weaker labour demand, higher costs of living and tighter credit conditions occurred simultaneously, since lower-income households generally have less liquid wealth and a higher marginal propensity to reduce consumption when credit conditions tighten.

1.5 Real estate markets remain broadly stable, but pockets of vulnerabilities persist

The overvaluation of house prices across the euro area as a whole remains contained, but some countries are showing signs of exuberance. The average valuation estimate for the euro area remains above the levels seen before the COVID-19 pandemic (Chart 1.10, panel a), but this masks significant cross-country variation, with estimates ranging from undervaluation in some countries to pronounced overvaluation in others. The dispersion has widened recently as valuations in the most stretched markets have picked up further. The main driver has been strong house price dynamics in a subset of countries, coupled in some cases with robust mortgage lending growth (see Box 1). In addition, housing supply is continuing to fall short of demand, sustaining growth in house prices.

Chart 1.10

House price overvaluation remains contained at the euro area level but is increasing in some countries, while mortgage markets seem to be slowing

a) Estimated over/undervaluation of residential property prices

b) Bank lending survey indicators for loans for house purchase

(Q1 2018-Q2 2026, percentages)

(Q1 2023-Q4 2026, weighted net percentages)

Sources: ECB (BLS, RESV) and ECB calculations.
Notes: Panel a: the blue line shows the average valuation estimate for the euro area (euro area 21, fixed composition as of 1 January 2026); the shaded area shows the maximum-minimum range for the same measure across euro area countries. A positive value indicates overvaluation and a negative value undervaluation. The average valuation estimate is the simple average of the price/income ratio and an estimated Bayesian vector autoregression model. For details of the methodology, see European Central Bank* and **. Panel b: net percentages are weighted by the share of each country in the total outstanding loan amounts of the euro area aggregate. For credit standards, the net percentage is the difference between the share of banks reporting a tightening and the share of banks reporting an easing. For net demand, the net percentage is the difference between the share of banks reporting an increase and the share of banks reporting a decrease. Actual values refer to the backward-looking three-month question and expected values to the forward-looking three-month question. The latest observations for actual values are for Q3 2026; the values for Q4 2026 refer to banks’ expectations.
*) European Central Bank, “Tools for detecting a possible misalignment of residential property prices from fundamentals”, Financial Stability Review, ECB, June 2011.
**) European Central Bank, “A model-based valuation metric for residential property markets”, Financial Stability Review, ECB, November 2015.

Going forward, the recovery in residential real estate (RRE) markets could face headwinds. Net demand for loans for house purchase moved back into contractionary territory in the second quarter of 2026, reflecting higher interest rates and lower consumer confidence. Banks tightened credit standards over the same period and expect both net demand and credit standards to decline further in the near term (Chart 1.10, panel b). Consequently, rising measures of overvaluation, together with tightening financial conditions, are contributing to tail risks for euro area house prices. Moreover, households’ debt servicing capacity could come under pressure in countries with a high share of variable-rate mortgages if financial conditions were to tighten further alongside a cooling labour market and with rising energy prices weighing on household expenses.

Commercial real estate (CRE) markets have stabilised at subdued levels, but vulnerabilities persist and downside risks have increased recently. Investor sentiment has remained broadly stable since the end of 2025, although slightly more investors now consider the market to be in a downturn (Chart 1.11, panel a). At the same time, financing conditions have become less supportive, with the share of investors reporting deteriorating credit conditions increasing markedly in the current context of higher interest rates and continued geopolitical uncertainty. Upcoming refinancing needs could also weigh on borrowers’ debt servicing capacity as loans originated during the period of low interest rates mature. Transaction activity remains subdued but stable, while CRE valuation growth is positive, although it is slowing and is also uneven across countries and segments. CRE prime yield spreads have narrowed markedly as long-term interest rates have risen faster than property yields and are now close to levels last seen in 2022 (Chart 1.11, panel b). Unless risk-free rates decline, compressed spreads could put property yields under renewed pressure and, in turn, exert downward pressure on CRE valuations.

Chart 1.11

Euro area CRE markets have stabilised, but tighter financing conditions are increasing downside risks

a) CRE market sentiment

b) CRE prime yield spread

(Q1 2015-Q2 2026, percentages of investors surveyed)

(Q1 2015-Q3 2026, percentages)

Sources: RICS, JLL and ECB calculations.
Note: Panel b: the shaded area shows the range across the office, retail and industrial sectors.

Structural vulnerabilities remain uneven across CRE segments and countries. Prime office markets are continuing to perform relatively well, while non-prime offices face weaker demand, higher vacancy rates and more subdued rent expectations. In prime segments, demand is continuing to outstrip new supply in some markets, driving up rental growth as high construction and financing costs constrain development. Logistics and parts of the retail market have remained comparatively resilient. Strong investment in data centres, partly related to the buildout of capacity related to artificial intelligence (AI), is also supporting demand in parts of the CRE market, although the broader impact of AI is still limited. CRE conditions have improved more noticeably in some southern euro area countries, while several larger markets remain weaker. Developers remain a particular vulnerability in some countries as weak activity, high input costs and tighter financing conditions weigh on profitability and constrain new development.

Real estate markets remain broadly stable, with vulnerabilities concentrated in selected segments and countries, and risks are still below previous peaks. High energy prices could add to risks in both RRE and CRE markets as they squeeze real disposable incomes and push up operating and development costs. For RRE, a substantial deterioration in labour market conditions would be the main trigger for broader household credit risk, but favourable labour markets and borrower-based measures are currently supporting resilience. CRE is more directly exposed to higher financing costs, and further increases in medium to long-term rates could put valuations and activity under renewed pressure. The risks are greatest for non-prime offices, developers and weaker markets, while prime assets remain more resilient. Given the size of banks’ RRE mortgage books, a broad-based decline in household credit quality could have serious implications for bank solvency at the euro area level. Banks’ CRE exposures are smaller but they are also unevenly distributed, so stress could still be material for the most exposed institutions. Hence, while real estate risks are not elevated by historical standards, there are pockets of vulnerability that warrant continued monitoring.

Box 1
Diverging developments in euro area residential real estate markets: risks and policy considerations

Prepared by Giorgia De Nora, Barbara Jarmulska and Cristian Perales

Since 2022 residential real estate (RRE) price growth has diverged significantly across the euro area. While one group of euro area countries is currently experiencing strong RRE price growth and has had price levels remaining relatively high since 2022, another has seen more subdued RRE price dynamics over the same period (Chart A, panel a). This box examines the factors underlying this variation in RRE price developments, with the aim of shedding light on the channels through which possible vulnerabilities may be building up.

More

2.1 An interplay of global risks might challenge euro area financial markets

Geopolitical tensions, rising yields and AI-related developments are shaping euro area and global financial markets. Since the previous edition of the Financial Stability Review was published in May, most broad asset classes have been hit by losses, although euro area equity markets saw gains (Chart 2.1, panel a). Increased geopolitical tensions – especially the re-escalation of the conflict in the Middle East – have been the main driver of trends in commodity and sovereign bond markets. Sovereign yields have increased significantly, reflecting inflation fears and the resulting change in monetary policy expectations, as well as fiscal concerns and, potentially, increased hyperscaler debt issuance (see Section 2.2). Higher global yields and a reassessment of sovereign debt trajectories have also triggered a widening in intra-euro area spreads and have weighed on the euro. By contrast, equity markets appear to have remained relatively resilient to geopolitical events and were largely being driven by earnings reports and AI-related news instead. The pricing of AI-related sectors and stocks has been volatile over recent weeks (see Section 2.3).[11] The continuing AI infrastructure buildout is further increasing AI-related exposures across various markets, increasing the risks of correlated sell-offs.

Chart 2.1

Markets are being driven by developments in AI and the conflict in the Middle East, with volatility picking up across markets

a) Returns across broad asset classes since the previous edition of the Financial Stability Review

b) Option-implied volatility across asset classes

(27 May-10 Nov. 2026, percentages)

(1 Jan.-10 Nov. 2026, indices)

Sources: Bloomberg Finance L.P., ICE BofA and ECB calculations.
Notes: Panel a: the previous edition of the ECB’s Financial Stability Review was published on 27 May 2026. “EA equity” reflects the EURO STOXX; “US equity” reflects the S&P 500 index; “Commodities” reflects the Bloomberg Commodity Index Total Return index. Corporate and sovereign bond indices are from ICE BofA. EA stands for euro area; IG stands for investment grade; HY stands for high yield. Panel b: “Start of Middle East war” refers to 2 March 2026, the first trading day after the start of the conflict on 28 February 2026; “Ceasefire announced” refers to 7 April 2026; “Ceasefire breaks down” refers to 27 June 2026.

Before a recent spike, volatility in equity markets had been low, despite the re-escalation of the conflict in the Middle East and bond market volatility. Forward-looking equity volatility was elevated during the initial phase of the Middle East war but came down after the ceasefire was announced before spiking again briefly following the recent US midterms (Chart 2.1, panel b). This contrasts with bond market volatility in the United States and the euro area, which increased alongside sizeable moves in sovereign bond yields. According to survey results, market participants identify the biggest tail risks as the rise in bond yields turning disorderly, the AI financing boom proving to be a bubble and a surge in inflation.[12] Low forward-looking equity market volatility might be indicating that markets are underestimating the risk of an abrupt, broad-based repricing.

Corporate bond and equity valuations have moderated but the latter remain sensitive to ambitious earnings expectations, especially in the technology sector. Valuations of US tech firms have fallen from their record highs (Chart 2.2, panel a), while euro area equity valuations, as measured by one-year forward price/earnings ratios, remain close to their historical average since 2005.[13] Earnings growth has been strong in both the euro area and the United States, which has fuelled further upward revisions to earnings expectations, most notably in the technology sector (Chart 2.2, panel b). Although large US tech firms have largely met earnings expectations, the earnings growth expectations currently embedded in market pricing are so high that they may be difficult to sustain, especially in the case of AI-related firms. Market reactions to recent disappointing earnings releases of large US tech firms and other AI-related companies have illustrated that markets continue to be exposed to any reassessment of the returns on AI investment (see Section 2.3). There are concerns that high reported earnings expectations may be inflated by sizeable valuation gains on equity stakes in other AI-related firms, reflecting circularity within the AI ecosystem rather than underlying operating performance (see Box 3).[14] Also, euro area corporate bond spreads have seen some recent widening amid rising government bond yields (see Box 2). Despite the recent moderation shown by valuation metrics, markets remain at high risk of abrupt further repricing, which is exacerbated by a rise in leveraged investments (see Section 4.2).[15]

Chart 2.2

Equity valuations have moderated but rely on very ambitious earnings expectations

a) Equity and bond valuations

b) Long-term earnings growth expectations

(1 Jan. 2005-10 Nov. 2026; left-hand scale: one-year forward P/E ratios, right-hand scale: z-spreads)

(1 Jan. 2020-10 Nov. 2026, percentages)

Sources: Bloomberg Finance L.P., LSEG and ECB calculations.
Notes: Panel a: for equities, the boxplot shows the interquartile range of one-year forward price/earnings (P/E) ratios relative to the distribution since 2005. The previous edition of the ECB’s Financial Stability Review was published on 27 May 2026. For the world reading (Wld), the MSCI World Index is used; for bonds, the boxplot shows interquartile range of the latest z-spread compared with the distribution since 2005. JP stands for Japan; EA stands for euro area; IG stands for investment grade; HY stands for high yield. Panel b: the chart depicts the weighted average long-term growth forecast for the EURO STOXX 50 and the S&P 500, as well as their technology sub-indices, based on the median analyst estimate for each constituent over the next three to five years (the next full business cycle).

Risks can interact and are aggravated by a rise in policy-related uncertainty. This is most evident in the way risks interact between AI-related assets and sovereign bond markets. Higher sovereign yields may weigh on long-duration AI-related investments, while AI-related bond issuance competes with long-dated government debt. Expectations of AI-driven productivity gains also underpin both equity valuations and debt sustainability. A shock in either market could therefore propagate to the other, amplifying longer-term losses in both. However, short-term reactions might differ, as illustrated by the recent sell-off in AI-related assets, which has triggered safe-haven flows and supported sovereign bond markets. Difficult-to-read inflation signals − from tariffs and energy to AI-related demand − compounded by a shift to less prescriptive communication at the Federal Reserve System, have made policy decisions harder to anticipate. This raises the likelihood of significant monetary policy surprises, even as uncertainty around ECB decisions remains low (Chart 2.3, panel a). More broadly, weakening multilateral cooperation leaves markets exposed to abrupt policy shifts, while upcoming elections in large euro area countries are adding further uncertainty. The recent US midterms also highlighted growing political opposition to large-scale AI infrastructure projects, heightening policy uncertainty around AI-related capital expenditure and casting further doubt on expected returns.

Supply-driven inflationary pressures have eroded the diversification benefits that bonds usually provide, leaving investors more exposed to losses. Sovereign bonds have traditionally acted as a hedge in balanced portfolios. With concerns rising about a pick-up in inflation and interest rates, however, stock and bond returns are again positively correlated, with the correlation in the euro area climbing to its highest level since 2005. A decomposition of the variance of a 60/40 stock-bond portfolio shows that, historically, a negative covariance between stocks and bonds reduced portfolio risk (Chart 2.3, panel b). This covariance has now turned positive and, together with a higher contribution from bond volatility, is adding to overall risk. As a result, a sell-off in financial markets triggered by a stagflationary shock would likely be broad based, with the co-movement of stocks and bonds potentially amplifying losses in diversified portfolios, even if sovereign bonds may still act as a hedge in risk-off episodes driven by other types of shocks.

Chart 2.3

Uncertainty has risen, while bonds no longer reduce portfolio risk

a) Distance of market pricing from the nearest fully priced-in outcome, one week ahead of rate decision

b) Variance decomposition of a 60/40 portfolio

(Jan. 2019-Oct. 2026, percentage points)

(14 Jan. 1999-10 Nov. 2026, percentages)

Sources: Bloomberg Finance L.P. and ECB calculations.
Notes: Panel a: based on the market-implied probability of a hike, hold or cut, priced for the Federal Reserve System and the ECB using the overnight index swap model on the day one week before the respective policy rate announcement. The chart shows the distance of this probability from a fully priced-in rate cut, rate hike or hold decision. A reading of zero means a full rate cut, hike or hold decision was fully priced ahead of the meeting; higher readings indicate that pricing sat between two outcomes, with a maximum of 50 percentage points when two decisions were priced as equally likely. Two unscheduled meetings of the Federal Reserve System’s Federal Open Market Committee in March 2020 and one unscheduled monetary policy meeting of the ECB in March 2020 are excluded. Panel b: the portfolio is constructed by investing 60% (wS ) into the EURO STOXX and 40% (wB ) into German ten-year sovereign bonds. Portfolio variance is decomposed as σP2=wS2σS2+wB2σB2+2wSwB CovrS,rB. Variance (σS2 for the equity part and σB2 for the bond part) and covariance between equity and bond returns (CovrS,rB) are calculated over 260-day rolling windows. The chart depicts the relative contributions of the three components to the total portfolio variance.

Box 2
Unwinding of corporate spread compression: implications for euro area market repricing

Prepared by Mar Domenech Palacios, Martina Jančoková, Paraic O’Gorman and Chiara Scommegna

Euro area corporate bond spreads have started to widen after a period of steady compression in recent years. Spreads peaked in October 2022 when disruptions to the supply of energy following Russia’s invasion of Ukraine, high inflation and rapidly tightening monetary policy gave rise to fears of a recession. Thereafter they narrowed substantially in both the investment-grade and the high-yield segments to levels well below their long-term averages (Chart A, panel a). This compression took place despite elevated geopolitical tensions, renewed inflation concerns linked to the conflict in the Middle East, growing debt issuance by AI hyperscalers and subdued growth prospects, all of which could have weighed on risk sentiment and corporate fundamentals. In recent weeks, however, corporate spreads have widened alongside rising government bond yields and tighter monetary policy. This box examines what is driving this widening and how a further repricing of risk would affect euro area borrowers and investors.

More

2.2 Sovereign bond yields at multi-year highs are posing challenges

Rising real rates and inflation concerns have put euro area sovereign bond markets under renewed pressure. Sovereign yields have increased in tandem over the past year in advanced economies and are now close to their highest level since the aftermath of the global financial crisis, especially at the long end of the curve (Chart 2.4, panel a). This has been driven by several factors, including rising energy prices adding to inflationary pressures and concerns over global fiscal expansion. There is limited evidence so far that AI-related debt issuance is crowding out euro area sovereign bonds. However, the scale of AI financing in the United States at higher durations may be contributing to higher yields there, which could add to upward pressure on euro area yields through spillovers. Model results suggest the increase in the United States mainly reflects the real rate component of the yield, with rising inflation expectations and term premium also playing a role in the euro area (Chart 2.4, panel b). Markets have continued to function in an orderly manner overall, although strains have prompted official action, particularly in the United States, including a joint US-Japanese intervention to support the yen in July and a US Treasury decision to increase its buybacks of longer-dated bonds.

Chart 2.4

Long-term yields are at decade highs, driven by real rates

a) 2-, 10- and 30-year sovereign yields in selected advanced economies

b) Decomposition of 10- and 30-year yield changes since January 2026

(1 Jan. 2010-10 Nov. 2026, percentages)

(1 Jan.-10 Nov. 2026, basis points)

Sources: Bloomberg Finance L.P., LSEG, Haver Analytics, Federal Reserve System and ECB calculations.
Notes: Panel a: the previous edition of the ECB’s Financial Stability Review was published on 27 May 2026. Panel b: the euro area (EA) term structure decomposition is based on an average of three models following Joslin et al.* (with and without surveys) and Geiger and Schupp**. The US decomposition is based on an average of three models: Adrian et al.***, Christensen-Rudebusch**** and D’Amico et al*****. The term structure models decompose the fitted yield, and the difference to observed yield change due to fitting error is not shown. The EA (US) real rate component is the overnight index swap rate (US Treasury yield) minus the inflation-linked swap rate (break-even inflation rate). The latest observations are for 10 November 2026 for EA decompositions and 9 November 2026 for US decompositions.
*) Joslin, S., Singleton, K.J. and Zhu, H., “A New Perspective on Gaussian Dynamic Term Structure Models”, Review of Financial Studies, Vol. 24, Issue 3, 2011, pp. 926-970.
**) Geiger, F. and Schupp, F., “With a little help from my friends: Survey-based derivation of euro area short rate expectations at the effective lower bound”, Discussion Papers, No 27/2018, Deutsche Bundesbank, 2018.
***) Adrian, T., Crump, R.K. and Moench, E., “Pricing the term structure with linear regressions”, Journal of Financial Economics, Vol. 110, Issue 1, 2013, pp. 110-138.
****) Christensen, J.H.E. and Rudebusch, G.D., “The response of interest rates to US and UK quantitative easing”, The Economic Journal, Vol. 122, Issue 564, 2012, pp. F385-F414.
*****) D’Amico, S., Kim, D.H. and Wei, M., “Tips from TIPS: The Informational Content of Treasury Inflation-Protected Security Prices”, Journal of Financial and Quantitative Analysis, Vol. 53, No 1, 2018, pp. 395-436.

It would appear that euro area yields are being driven in large part by common external shocks, notably higher energy prices and spillovers from US yields. Given the continued geopolitical tension in the Middle East, the prices of oil, gas and refined products are expected to stay elevated for longer than market participants initially expected (see Section 1.1). At shorter maturities, higher energy costs pushed up inflation expectations. So far, medium-term expectations are little changed (Chart 2.5, panel a), but a more prolonged shock could test how well anchored these expectations are. Developments in the United States have also added to the pressure at the long end of the curve, with Treasury yields reaching levels last seen before the global financial crisis amid stronger than expected growth, record investment in AI and concerns about US fiscal fundamentals. These external shocks have lifted sovereign yields across all euro area countries, with domestic factors adding further pressure in some.

Chart 2.5

Higher energy prices raise short-term (not medium-term) inflation compensation, while spreads have widened in some countries

a) Natural gas price and futures, and euro area 5-year and 5y5y inflation swap rates

b) Selected 10-year euro area sovereign spreads over the German 10-year yield

(1 Jan. 2024-31 Dec. 2027; left-hand scale: €, right-hand scale: percentage points)

(1 Jan. 2007-10 Nov. 2026, percentage points)

Sources: ECB, Bloomberg Finance L.P. and ECB calculations.
Notes: Panel a: natural gas price and futures refer to Netherlands TTF Natural Gas Forward Day Ahead (EUR). “Gas future – pre-war” refers to 27 February 2026; “Gas future – May FSR” refers to 27 May 2026. “Gas future – latest” refers to 10 November 2026. Euro area inflation 5y and 5y5y are inflation swap rates. Panel b: “Average euro area spread” is the GDP-weighted spread for Belgium, Ireland, Greece, Spain, Netherlands, Austria, Portugal and Finland. 2025 GDP is used as a proxy for 2026 GDP.

Long-standing fiscal concerns, compounded by political uncertainty, are also pushing intra-euro area spreads wider in some highly indebted countries. The spread between French and German ten-year bonds has risen significantly (Chart 2.5, panel b). Markets are increasingly pricing in political risk ahead of significant elections in 2027. The widening French-German spread has resulted in some spillover effects to other highly indebted sovereigns, while German Bunds have benefited from their status as a relatively safe haven. Although the average euro area spread has widened recently, it remains low compared with levels reached during previous periods of stress. More broadly, changes in spreads since the start of 2025 have aligned broadly with projected debt-to-GDP ratios (Chart 2.6, panel a). The relationship is somewhat steeper at the long end of the curve, indicating that fiscal concerns are priced in mainly for longer horizons, which can provide price signals that encourage prudent fiscal policies. Despite brief episodes of heightened volatility and reduced liquidity, sovereign markets have largely absorbed these pressures in an orderly fashion. However, internal and external challenges continue to leave them exposed to the risk of a disorderly adjustment.

Chart 2.6

Long-term yields reflect fiscal concerns, with sovereign issuance shifting to shorter-term debt

a) Changes in spreads vs OIS rate and debt outlook for euro area countries

b) Euro area gross sovereign debt issuance by low-debt and high-debt countries

(1 Jan. 2025-10 Nov. 2026; percentage points, basis points)

(Q1 2021-Q3 2026, percentages)

Sources: Bloomberg Finance L.P., IMF, LSEG, ECB projections database, Böninghausen and Vladu* and ECB calculations.
Notes: Panel a: the chart shows changes in the spreads between ten-year and thirty-year sovereign yields versus overnight index swap (OIS) rates since 2025 and IMF 5-year projected changes in government gross debt-to-GDP ratios. Dotted lines show regression lines. Panel b: low-debt countries have debt-to-GDP ratios below the euro area average: Bulgaria, Germany, Ireland, Croatia, Cyprus, Estonia, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Austria, Slovenia and Slovakia. High-debt countries have debt-to-GDP ratios above the euro area average: Belgium, Greece, Spain, France, Italy, Portugal and Finland. Debt issuance is shown as a rolling four-quarter average.
*) Böninghausen, B. and Vladu, A.L., “Sloping up: the repricing of euro area yields in 2025”, The ECB Blog, ECB, 16 January 2026.

Pressures are amplified by a changing investor base and a shift towards shorter-dated issuance. Given that the Eurosystem is no longer an active buyer, more procyclical investors (e.g. foreign investors, investment funds and hedge funds) are increasingly absorbing new issuance. Participation by players located outside of the euro area particularly reflects growing hedge fund activity, which has become a significant − albeit procyclical − source of liquidity (see Box 7). Japanese investors are significant holders of euro area sovereign debt, particularly French government bonds. They have been reducing their holdings of long-dated euro area debt since 2022 and could reduce them still further if rising domestic yields or exchange rate movements made domestic assets more attractive. Competition for long-dated funding has also intensified, with a surge in AI-related corporate issuance proving attractive for investors. Sovereigns face weaker demand at the long end, including reduced demand following the recent reform of the Dutch pension system. As a consequence, they have shifted issuance towards shorter maturities, especially in highly indebted countries (Chart 2.6, panel b). While this will keep funding costs contained in the near term, it leaves sovereigns more exposed to shifts in investor sentiment, with a quarter of outstanding euro area debt maturing before the end of 2028 and very high sovereign financing needs in a number of countries.

Rising long-term yields run the risk of tightening financing conditions beyond sovereign markets. This would compound the effect of monetary policy tightening as it feeds through the wider financial system. Euro area corporate bond yields have risen since the previous edition of the Financial Stability Review was published, and some corporate borrowers are vulnerable to a further deterioration in refinancing conditions (see Box 2). This risk is mitigated somewhat as the corporate maturity wall is still some years away. Investment-grade redemptions will peak in around 2028 and high-yield maturities will fall due almost entirely in the next five to seven years. Meanwhile, the floating-rate structure of private credit means higher rates pass through immediately to these riskier borrowers. AI-related investment is adding to these pressures (see Box 4). Higher yields are also compressing asset valuations by raising the rate at which future cash flows are discounted. This is weighing most on growth stocks where earnings lie furthest ahead.

2.3 AI external financing needs are increasingly linking exposures across different market segments

Market pricing for AI-related assets has turned volatile, as worries emerge about profitability, competition, cyber risk and a slowdown in AI development. In June enterprise-facing AI applications and agentic products moved to consumption-based pricing. The result has been a rapid increase in lower-cost models, including cheaper Chinese open-source AI models (Chart 2.7, panel a). In addition to broader concerns around overinvestment in AI-related assets, this has raised questions over which sectors and companies will be able to reap AI-related profits. On the one hand, hyperscalers have underperformed broad equity markets, perhaps because of uncertainty over the extent to which large capital expenditure will generate future profitability (Chart 2.7, panel b). On the other hand, semiconductor stocks, including those in emerging market economies, had first been assessed as more likely to be beneficiaries of the AI buildout. After a sizeable rally, however, semiconductor stocks corrected, as general risk appetite for AI-related assets dropped significantly. Although some of the losses were quickly recovered, markets still appear to be less convinced of productivity gains from AI, as the performance of companies considered to be AI winners converged with the performance of companies at risk of being displaced by AI. The outperformance shown by cybersecurity stocks also reflects worries about an increase in cyberattacks stemming from the improved capabilities of AI models (see Special Feature B). There have been calls for a slowdown in AI model development and more regulation due to security concerns and the buildout of large-scale AI infrastructure projects could slow down due to growing political opposition.

Chart 2.7

Market pricing on the AI narrative has turned volatile, as worries emerge about profitability, competition, cyber risk and a sequence of adverse news

a) Chinese and US AI model token usage through OpenRouter

b) Selected equity index returns

(4 Dec. 2025-5 Nov. 2026; left-hand scale: trillion tokens,
right-hand scale: index)

(1 Jan.-10 Nov. 2026, index: 1 Jan. 2026 = 100)

Sources: Bloomberg Finance L.P., LSEG and ECB calculations.
Notes: Panel a: usage of Chinese and US models measures aggregate token volume processed through the OpenRouter application programming interface, which is estimated to account for approximately 2% of current global AI token consumption. The Silicon Data LLM Token Expenditure Index is a benchmark for the price of large language model (LLM) tokens calculated from the usage-weighted price of one million tokens across a defined LLM universe. Panel b: the UBS EU AI Winners basket tracks the performance of euro area listed stocks of firms enabling and adopting AI technologies. The UBS EU AI Risk basket tracks the performance of euro area listed stocks of firms considered to be potentially at risk of disruption from AI. “MSCI World incl. emerging markets” refers to the MSCI World Index including emerging markets.

As free cash flow has diminished, the AI infrastructure boom has come to rely on external sources of funding. For most hyperscalers, free cash flow has declined considerably since early 2025. It has turned negative in 2026 and is expected to remain so until 2028. The AI investment boom, which in its earlier stages was largely funded from internal profits, has therefore started to rely heavily on external sources of funding (see Box 3). This is reflected in large equity issuances by big tech firms, including seasoned equity offerings and record initial public offerings (IPOs) conducted or announced by AI-related companies (Chart 2.8, panel a). These large IPOs could be a sign of late-cycle market exuberance and run the risk of introducing additional volatility into equity markets (see Special Feature C). Big tech firms are also increasingly taking on debt (including euro-denominated debt) through private credit and public bond markets, with a notable rise in issuance. Such issuance may crowd out other investments and is cited as a potential contributor to the rise in US sovereign bond yields. However, there is only limited evidence of crowding out in euro-denominated markets so far (see Box 4). In recent weeks, concerns about the sustainability of this debt funding have risen, with credit default swap spreads for a number of big tech firms climbing to record levels.

Chart 2.8

AI financing boom increasingly links different segments of capital markets

a) Big tech firms’ annual equity and debt issuance

b) AI concentration across markets

(1 Jan. 2020-10 Nov. 2026, € billions)

(31 Dec. 2024, 10 Nov. 2026; percentages)

Sources: Bloomberg Finance L.P., Dealogic, a service of ION Analytics, PitchBook, ICE BofA and ECB calculations.
Notes: Panel a: for 2026, issuance is until 10 November. The sample covers Alphabet, Amazon, Broadcom, Meta, Microsoft, NVIDIA, Oracle and SpaceX. “Other currencies” includes the Australian dollar, pound sterling, Canadian dollar, Japanese yen and Swiss franc. “Ten-year average debt issuance” refers to issuance by tech firms. Panel b: EA stands for euro area. AI firms are defined as those involved in the buildout of AI data centres and infrastructure, the development of frontier AI models or the manufacturing of chips used for AI. Concentration in equity indices is the share of market capitalisation of AI-related firms in the total market capitalisation of the respective index. Concentration in bond indices is computed likewise, based on market value. “US IG 10y+” and “EA IG 10y+” refer to the subsets of the ICE BofA US Corporate Index and the ICE BofA Euro Corporate Index that include investment-grade (IG) bonds with a maturity of more than ten years. 2024 refers to end-2024 data for equity and bond indices; 2026 refers to the latest available data point, which is 10 November 2026. Due to missing data on outstanding amounts in private markets, AI funding shares for venture capital, private credit and private equity are based on transaction-level data for 2024 and 2026 (deal volumes for venture capital and private equity, and debt volumes for private credit). AI-related firms are identified using the PitchBook-defined verticals “Artificial Intelligence & Machine Learning”, “Big Data” and “CloudTech & DevOps”, supplemented by targeted keywords identifying AI compute and data centre infrastructure. The cited data has not been reviewed by PitchBook analysts and may be inconsistent with PitchBook methodology.

With the associated surge in financing and rise in valuations, the AI narrative is reinforcing the role of AI as a common source of vulnerability across markets. Ongoing issuances in equity and bond markets, as well as sizeable deals related to AI companies and infrastructure in private markets, are continuing to increase AI-related exposures in a range of markets (Chart 2.8, panel b). Previously, these markets were driven by different factors, so risks could be diversified. This concentration on a single market narrative, which is expected to increase, sets the scene for further correlated losses in the event of AI-related volatility.

Euro area equity markets are less tech exposed, but still suffer from tech-related sentiment spillovers, although they are recovering quickly. Euro area equity markets are less concentrated in technology or AI-related sectors than their US peers (Chart 2.9, panel a). Structurally, their lower tech share implies that while they miss out on valuation gains, they may benefit from inflows from investors who appreciate the diversification they offer. The evidence to date suggests that in the event of an AI or a technology-related shock, euro area markets fall in tandem with US equity markets, likely driven by a spillover in global risk sentiment. They do, however, see smaller losses and appear to recover faster than their US peers (Chart 2.9, panel b). Nevertheless, euro area investors remain highly exposed to AI-related investments in non-domestic markets (see Chapter 4).

Chart 2.9

Less tech-heavy euro area markets suffer from technology-related shocks spilling over to global risk sentiment but also benefit from diversification flows

a) Global equity sector weights as a share of market capitalisation

b) Equity market response to a negative tech-related shock

(10 Nov. 2026, percentages)

(trading days after tech-related shock, percentages)

Sources: Bloomberg Finance L.P. and ECB calculations.
Notes: Panel a: “Euro area” refers to the EURO STOXX; “United States” refers to the S&P 500. Panel b: local projections using daily data starting from January 2022 estimate cumulative equity market responses over 20 trading days to a 1 standard deviation negative technology-related equity shock, defined as the Nasdaq falling by at least 1 standard deviation more than its usual relationship with the S&P 500 would imply. Regressions include five lags of the dependent variable, VIX changes and US and German ten-year yield changes; shaded areas denote 90% Newey-West confidence intervals.

Box 3
The debt financing of AI and its implications for financial stability

Prepared by Federica Bosio, Lara Coulier, Daniel Dieckelmann, Tina Emambakhsh, Maciej Grodzicki and Maria Leonor Puga[16]

The investment needs of firms in the AI supply chain are unprecedented in scale. Cumulative capital expenditure by US big tech firms alone is forecast to reach about USD 5.5 trillion by 2030, although estimates vary widely. On an annual basis, overall AI investment needs would amount to more than 3% of US GDP between 2025 and 2032, which is considerably larger than the financing of other major innovations in the past.[17] Investment in the AI buildout is highly concentrated among a small group of US firms, including hyperscalers and manufacturers of semiconductors. While these firms have in the past generated large free cash flows through their non-AI-related activities, free cash flows turned negative for most US hyperscalers in 2026, as capital expenditure on AI infrastructure exceeded operating cash flows.[18] In response, these firms are increasingly turning to external funding from equity[19] and debt.

More

Box 4
Big tech’s growing presence in the euro-denominated bond market: implications for investors

Prepared by Anne Duquerroy, Oana Furtuna and Lia Vaz Cruz

Hyperscalers are increasingly turning to credit markets to fund an investment cycle which is unprecedented in scale.[20] At individual firm level, hyperscalers now rival financials as the largest issuers in the US dollar-denominated investment-grade corporate bond market, having issued over USD 200 billion this year alone. They have also branched out into other currencies, with the euro playing an important role in the foreign currency mix. As hyperscalers have become increasingly active in the euro-denominated investment-grade (EUR IG) corporate bond market, euro area investors’ exposure to the US economy and the tech sector has risen further. This is increasing risks for investors by limiting diversification and leaving portfolios more vulnerable to a shock related to artificial intelligence (AI). This box looks at the financial stability implications of hyperscalers’ growing presence in the euro area credit market, particularly at the long end of the curve, focusing on concentration and correlation risks. It also considers how exposed euro area investors are and whether shifts in their portfolios suggest that AI-related borrowing is displacing other issuers.

More

3.1 The near-term outlook for bank profitability is still positive, but wide cross-country differences remain

Euro area banks’ profitability remained strong in the first half of 2026, supported by a rebound in operating income and a slight pick-up in leverage. The trailing return on equity (ROE) of significant institutions reached the highest level seen in the last ten years. This was due to an increase of pre-provisioning operating profits that more than offset a modest rise in loan loss provisions (LLPs) (Chart 3.1, panel a). A decomposition of changes in ROE reveals that a moderate increase in leverage has also contributed positively to ROE in recent quarters (Chart 3.1, panel b).

Chart 3.1

Aggregate profitability remains strong, but cross-country disparities persist

a) Decomposition of changes in net income and ROE

b) Contributions from ROA and leverage to year-on-year change in ROE

c) Drivers of profitability differences between country groups

(Q1 2017-Q2 2026; left-hand scale: percentages of equity, right-hand scale: percentage points)

(Q1 2024-Q2 2026, percentage points)

(Q2 2026; percentages, percentage points)

Sources: ECB (supervisory data) and ECB calculations.
Notes: Based on all significant institutions and on four-quarter trailing figures. ROE stands for return on equity. Panel a: “Other P&L items/taxes” includes impairment on non-financial assets, negative goodwill, results from discontinued operations, extraordinary profit or loss and tax expenses, among other things. Panel b: leverage is defined as total assets over equity. ROA stands for return on assets. Panel c: based on weighted averages for countries with below and above-median ROE. Countries with below-median ROE: Belgium, Germany, Estonia, Ireland, Greece, France, Luxembourg, Malta, Netherlands and Slovenia. Countries with above-median ROE: Bulgaria, Spain, Italy, Cyprus, Latvia, Lithuania, Austria, Portugal and Finland. Panel c: NII stands for net interest income; NFCI stands for net fee and commission income; “Other op. inc.” stands for other operating income; OPEX stands for operating expenses; “Imp./prov.” stands for impairments and provisions.

However, there is still wide cross-country variation in bank profitability. Dividing euro area countries into two groups based on their ROE level shows that banks in countries with an above-median ROE generate significantly higher net interest income (as a share of equity) than banks in countries with below-median ROE (Chart 3.1, panel c). One of the main underlying drivers of the net interest income gap is asset and liability composition, with the overperforming group of countries having a significantly higher share of customer loans and deposits in their balance sheets.[21] In addition, the underperforming group of countries has tighter customer loan-deposit spreads.[22] This partly reflects their higher share of fixed-rate loans as banks in these countries tend to benefit less from rising interest rates, as was the case in the previous rate hiking cycle.[23] Furthermore, lower market concentration and pricing power is weighing on loan-deposit spreads in some of these countries.

Net interest income growth has recovered, while net fee and commission income growth has also gained momentum. Net interest income growth turned positive again in the second quarter of 2026, driven by both stronger volume effects and a less pronounced negative impact from the tightening of margins (Chart 3.2, panel a). Accelerating growth in net fee and commission income also contributed positively to overall profitability. The recent positive momentum seen in fee income generation was driven mainly by fees related to asset management and product distribution (Chart 3.2, panel b). As fee income growth continues to depend strongly on investment management activities, it remains sensitive to abrupt changes in global risk sentiment and to a further rise in long-term rates.

Chart 3.2

Core revenue growth picked up again, while cost ratios improved

a) Decomposition of net interest income change and net interest margin level

b) Net fee and commission income growth and contribution from investment management

c) Banks’ cost/income and cost/assets ratios

(Q1 2022-Q2 2026; left-hand scale: percentage changes, right-hand scale: percentage points)

(Q1 2018-Q2 2026, percentage points)

(Q1 2016-Q2 2026, percentages)

Sources: ECB (CBD, supervisory data) and ECB calculations.
Notes: Based on all significant institutions and on four-quarter trailing figures. Panel a: decomposition of year-on-year change in net interest income (NII). Panel b: investment management includes asset management and distribution of (not managed) products. NFCI stands for net fee and commission income. Panel c: dotted lines represent long-term averages, calculated for the period from Q4 2015 to Q2 2026.

Improved cost ratios also contributed to higher overall profitability, while the impact of rising LLPs was small. The aggregate bank cost/income ratio reached its lowest level since 2015, as a moderate rise in operating expenses was outweighed by robust revenue growth. Following the upward trend observed between 2022 and 2025, the aggregate cost/assets ratio declined in the first half of 2026 and has remained below its long-term average (Chart 3.2, panel c). LLPs picked up somewhat in the first two quarters of 2026, rising by 12% year on year, which reflects a more cautious credit quality outlook. However, banks’ cost of risk, as measured by the ratio of LLPs to total loans, edged up by just 2 basis points from its end-2025 level due to the offsetting impact of robust loan growth.

Analyst projections suggest a positive outlook for near-term profitability, but some downside risks remain. For large listed banks, ROE is projected to rise by around 2 percentage points between 2025 and 2027, driven by continued revenue growth. Furthermore, analysts expect the rise in LLPs to have only a modest negative impact on profitability (Chart 3.3, panel a). There are still some downside risks to this positive earnings outlook, however. First, while higher long-term interest rates could support net interest income via steeper yield curves and higher reinvestment yields, this could be offset, at least in part, by the negative impact on funding costs and loan demand. Second, LLPs could rise by more than foreseen if tightening financing conditions and higher oil prices were to have a more pronounced impact on bank asset quality than is currently anticipated. Finally, some components of non-interest income, such as trading revenues or fee income from asset management, remain susceptible to any abrupt deterioration in financial market sentiment (see Chapter 2). In addition, higher long-term rates could lead to valuation losses on debt securities recognised at fair value (potentially affecting profit and loss or equity) as well as unrealised losses on debt securities held at amortised cost.

Chart 3.3

Bank valuations continue to be supported by a positive outlook for earnings, with a partial reversal in gains since late September due to higher equity risk premia

a) Drivers of expected ROE

b) Euro area banks’ price/earnings and price-to-book ratios

c) Decomposition of cumulative changes in bank share prices

(2025-27; percentages, percentage points)

(1 Jan. 2021-6 Nov. 2026, ratios)

(2 Jan.-6 Nov. 2026; percentages, percentage points)

Sources: Bloomberg Finance L.P., LSEG and ECB calculations.
Notes: Panel a: based on a sub-sample of 25 banks from the EURO STOXX Banks index, using the mean as consensus. ROE stands for return on equity; “Op. inc.” stands for operating income; OPEX stands for operating expenses; LLP stands for loan loss provisions. Aggregate figures are equity weighted. Panels b and c: based on weekly data. Panel b: based on the EURO STOXX Banks index. Dotted lines represent the medians of valuation ratios since 2003. Panel c: the decomposition of cumulative changes in the EURO STOXX Banks index is based on the H-model presented in Fuller and Hsia*.
*) Fuller, R.J. and Hsia, C.-C., “A Simplified Common Stock Valuation Model”, Financial Analysts Journal, Vol. 40, No 5, 1984, pp. 49-56.

Bank valuations continue to be supported by a positive outlook for earnings, but they remain exposed to abrupt changes in risk sentiment. Price/earnings ratios and price-to-book ratios have been somewhat volatile since mid-2026, with initial valuation gains followed by a partial reversal due to worsening sovereign risk perceptions in a few euro area countries (Chart 3.3, panel b). A dividend discount model decomposition shows that upward revisions to expected earnings and dividend payouts have aided the increase in bank share prices for much of 2026, while rising equity risk premia have had a partial offsetting impact since late September (Chart 3.3, panel c). Going forward, bank valuations remain sensitive to a possible further deterioration in sovereign risk perceptions or negative earnings surprises.

3.2 Banks’ funding costs could rise as liquidity is drained

Bank bond yields have risen in response to higher benchmark rates and wider sovereign spreads in some countries. Bond financing conditions for euro area banks have tightened, with yields on senior unsecured bonds surpassing the 4% mark in September. The increase in yields has for the most part been driven by an increase in the risk-free market rate, although widening sovereign spreads have also contributed in some countries. Bank bond yields have risen less than sovereign yields in these countries as the compression of spreads between banks and their domestic sovereign has partly offset the rise in sovereign yields (Chart 3.4, panel a). This suggests that market participants are discriminating between the credit risk of banks and their sovereign, and the bonds of some banks even briefly traded at a premium to the sovereign. Despite this decoupling, bank financing conditions could still be vulnerable to spillovers from sovereign risks, and vice versa, in times of severe market stress.

Chart 3.4

Bank bond yields have risen, while the increase in spreads remains contained

a) Decomposition of bank bond yields into spread to domestic sovereign and swap rate

b) Euro area bank bond spreads across seniorities

(2026, percentages)

(3 Jan. 2022-10 Nov. 2026, basis points)

Sources: S&P Dow Jones Indices LLC and/or its affiliates and ECB calculations.
Notes: Panel a: 27 February 2026 is the day before the outbreak of the latest conflict in the Middle East; June was the month when bank bond spreads reached their low point in 2026. “Spread: bank to sovereign” is the difference between the yield of bank bonds and the five-to-seven-year yield on the sovereign bonds of the domestic sovereign. “Spread: sovereign to swap rate” is the asset swap spread on five-to-seven-year domestic sovereign bonds. Panel b: the spread is the asset swap spread over the maturity-matched swap rate.

Although bank bond spreads have increased recently, they remain within historical norms and are tight for the most junior instruments. Bank bond spreads have widened on the back of the recent repricing seen in euro area credit markets. The trend over the past four years reveals that, despite this increase, credit spreads are still well within historical norms and remain compressed for AT1 instruments (Chart 3.4, panel b). This may be an indication that investors consider banks’ credit risk, and in particular the likelihood of tail risks (such as the need to convert eligible bonds to equity), to be contained, given that euro area banks are now better capitalised and more resilient.

Deposit competition remains limited to term deposits, while overnight rates are low and stable. Average overnight retail deposit rates are still far below the policy rate and, in aggregate, have barely moved despite the recent central bank rate hikes and increases in market interest rates. That said, there is a wide range of overnight deposit interest rates across banks and some banks at the upper end of the distribution may be competing more for deposits as they match the recent increase in benchmark rates. Meanwhile, the tight distribution of interest rates on new term deposits around the median and their proximity to the market benchmark rate indicates that term deposit markets are already more competitive (Chart 3.5, panel a).

Chart 3.5

Deposit rate pass-through is still limited but could increase

a) Dispersion of overnight and term deposit rates across euro area banks

b) Ratio of overnight deposit rate to market rate, by ratio of central bank reserves to total assets and country

(Jan. 2022-Sep. 2026, percentages)

(Jan. 2003-Dec. 2011, Sep. 2022-Sep. 2026; percentages)

Sources: ECB (MIR, BSI, bank treasurer survey) and ECB calculations.
Notes: Panel a: weighted average rate for households and non-financial corporations. Rates on new business for term deposits. The benchmark rate is the deposit facility rate for overnight deposits and the 12-month EURIBOR for term deposits. Panel b: deposit pass-through is calculated as the ratio of the level of overnight deposit rates to the benchmark rate, meaning the €STR/EONIA at a monthly frequency (i.e. each dot indicates one monthly observation). Values where the €STR was below 0.5% were excluded. The blue dots on the right graph show the country-level overnight deposit rate divided by the benchmark rate.

Looking ahead, deposit rates may start to come under greater pricing pressure as excess liquidity is drained. Compared with the situation before the low interest rate environment, the ratio of overnight deposit rates to benchmark market rates is still low (Chart 3.5, panel b). As banks’ liquidity positions tighten, however, they may have a greater incentive to source deposits. This would exert upward pressure on deposit rates, increasing bank funding costs. Banks in countries where deposit rates are still very low might be more vulnerable to rising deposit rates, which could weigh on their profitability. The growing market share of more aggressive competitors such as digital banks − which could also benefit from reduced frictions due to technological advancements − might exert further pressure on the deposit rates of traditional banks.[24] Agentic AI models could make it quicker and easier to compare deposit rates online and shift funds automatically. Euro area deposits tend to be quite sticky, however, and euro area residents prefer to keep their savings in bank accounts rather than invest in financial markets.[25]

While aggregate liquidity remains high, excess reserves are declining and are being redistributed across the system. The aggregate liquidity coverage ratio remains fairly stable, but the share of excess reserves in banks’ liquidity buffers is continuing to decline as the Eurosystem’s balance sheet normalises (Chart 3.6, panel a). At the same time, excess reserves are being redistributed across banks, reducing differences in reserve levels.

Chart 3.6

Liquidity buffers remain stable while excess reserves decline

a) Excess reserve and liquidity coverage ratios

b) Sovereign bond holdings and home bias of euro area banks, by country

c) Borrowing by banks through repurchase agreements, by counterparty

(Q1 2017-Q2 2026, percentages of LCR denominator)

(Q2 2022, Q2 2026; percentages of total assets and total sovereign bond holdings)

(Q1 2021-Q2 2026, percentages of total liabilities)

Sources: ECB (supervisory data) and ECB calculations.
Notes: Panel a: excess reserves are in euro for banks that report the split by currency and total reserves for banks that do not report this split. The excess reserve ratio is the ratio of excess reserves to the LCR denominator. LCR stands for liquidity coverage ratio. Panel b: home bias is defined as the ratio of domestic sovereign bond holdings to total sovereign bond holdings. EA stands for euro area. Panel c: total liabilities are defined as all non-equity, non-derivative liabilities of banks.

Liquidity buffers have mainly been supported by purchases of non-domestic sovereign bonds, contributing to a modest reduction in home bias. Sovereign bond holdings have increased rapidly in recent years across the entire euro area banking sector, but the relative level of exposures still varies significantly across countries (Chart 3.6, panel b). The growing stock of sovereign bonds held directly supports banks’ liquidity buffers as high-quality liquid assets (HQLA) but can also serve as collateral in repo markets to obtain additional cash funding.[26] Indeed, liquidity redistribution via repo markets has increased recently, especially where funding is provided by non-bank counterparties (Chart 3.6, panel c). While repo funding is secured on liquid assets, which can be eligible for other funding sources and mitigate liquidity risk, non-bank funding can be volatile and vulnerable to disruptions during periods of market stress. The expansion of repo markets is strengthening the ties between banks and non-banks and could also increase their vulnerability to spillovers from sovereign bond markets.

3.3 Asset quality remains sound, albeit with some variation at the country and the sector level

Aggregate euro area bank asset quality remains sound, with a record-low NPL ratio and falling default rates. The aggregate non-performing loan (NPL) ratio of euro area significant institutions stood at 2.2% in the second quarter of 2026, representing a marginal decrease since the end of 2025 and a record low (Chart 3.7, panel a). While the stock of NPLs increased slightly in the first two quarters of 2026, driven mainly by countries with previously low NPL ratios, this has been outpaced by the growth of total loans. Default rates on all exposure types have been also trending downwards since 2025. For mortgage lending, they have fallen below levels observed before mid-2022, which was when the previous monetary policy tightening cycle started (Chart 3.7, panel b). At the same time, the declining trend for exposures to corporates and small and medium-sized enterprises (SMEs) has come to a halt, with SME default rates remaining above their previous low recorded in late 2021.

Chart 3.7

Aggregate euro area bank asset quality remains sound

a) NPL stocks by country group, total NPLs and the aggregate NPL ratio

b) Annual default rates for bank exposures, by sector

(Q1 2021-Q2 2026; left-hand scale: € billions, right-hand scale: percentages)

(Q1 2021-Q2 2026, percentages)

Sources: ECB (supervisory data) and ECB calculations.
Notes: Panel a: based on all significant institutions. Euro area countries are divided into two groups based on whether their non-performing loan (NPL) ratios were above or below the median in Q1 2021. Panel b: includes on- and off-balance-sheet exposures subject to credit risk. Four-quarter trailing figures for significant institutions reporting under the internal ratings-based method. The sectoral definitions under COREP used for default rates differ from those used elsewhere (e.g. FINREP). SME stands for small and medium-sized enterprise.

A benign overall picture masks some variation in asset quality at the country and the sector level, particularly for SME and consumer loans. While banks’ aggregate SME NPL ratio has not increased from end-2025 levels, SME lending in a few large countries continues to show signs of vulnerability. In particular, the SME NPL ratio has been gradually trending upwards in France and Germany since 2023 (Chart 3.8, panel a). This reflects, in part, a weaker macroeconomic environment in these countries. The quality of consumer loans has also continued to worsen as net NPL flows of consumer loans, which had been positive since 2023, picked up in the first two quarters of 2026 (Chart 3.8, panel b). Aggregate unemployment in the euro area remains low, and the overall financial position of households appears robust (see Section 1.4). However, consumer loan asset quality could deteriorate further as higher interest rates and inflation weigh on households’ real income and debt servicing capacity. In addition, given that exposures to non-euro area regions account for around 30% of consumer loans and 40% of consumer loan NPLs, the overall portfolio also remains sensitive to worsening macroeconomic conditions outside the euro area. That said, the share of consumer loans in the total loans of euro area significant institutions remained contained at 6.2% in June 2026, despite varying widely across the five largest euro area countries, from 1.4% (the Netherlands) to 13.2% (Spain).

Chart 3.8

SME and consumer loan quality showed signs of weakening in some parts of the euro area, while there are also downside risks stemming from the oil supply shock

a) SME NPL ratio and contributions from the largest and other euro area countries

b) Net NPL flows for consumer loans

in the

largest and other euro area countries

c) Share of exposures and NPL ratios for sectors vulnerable to the oil supply shock

(Q1 2021-Q2 2026; percentages, percentage points)

(Q1 2022-Q2 2026, € billions)

(Q2 2026, percentages)

Sources: ECB (AnaCredit, RIAD, supervisory data) and ECB calculations.
Notes: Panels a and b: based on all significant institutions and the home country of the reporting bank. Panel a: euro area aggregate SME NPL ratio and contributions from the largest/other euro area countries. Panel b: four-quarter trailing net NPL flows. Panel c: based on all banks. “NPL ratio vuln.” stands for the NPL ratio for sectors vulnerable to the oil supply shock, defined as the most energy-dependent sectors, exhibiting high sensitivity to upstream energy supply disruptions. Exposure shares only indicate potential sensitivity to the oil supply shock and should not be interpreted as explaining contemporaneous NPL ratios, as the impact of oil supply shocks on bank asset quality materialises with a lag. Banks’ estimated exposures to energy-intensive sectors should be interpreted with caution, as their magnitude may vary depending on the definitions, sectoral coverage and methodology adopted. NFC stands for non-financial corporation.

The oil supply shock caused by the conflict in the Middle East has not yet led to a material deterioration in asset quality, but downside risks have increased. The oil supply shock is mainly affecting firms in energy-intensive sectors, through higher energy costs which could reduce their profit margins and eventually result in more defaults. However, the overall asset quality of exposures to these sectors had not shown signs of deterioration until the second quarter of 2026, with the NPL ratio unchanged from the end of 2025 (at 4.2%). Looking ahead, the impact of the oil supply shock is expected to have a negative impact on bank asset quality in the upcoming quarters (see Box 5). The impact is likely to vary across the largest euro area countries, partly in line with the size of exposure to energy-intensive sectors (Chart 3.8, panel c).[27]

NPL coverage ratios continue to decline, but provisioning needs might increase, given the ongoing uncertainty related to high energy prices. In the 12 months to June 2026, the coverage of NPLs by loan loss reserves continued to decline for total, corporate and household loans (Chart 3.9, panel a). Furthermore, the decline in the total coverage of NPLs, also including collateral and financial guarantees, was even more pronounced over the same period. The change in NPL coverage ratios has varied somewhat across countries, with countries showing patterns of increasing NPL ratios typically reporting declines in their NPL coverage ratios (Chart 3.9, panel b). The declining trend in NPL coverage ratios partly reflects a shift towards newer NPL vintages within the total stock, as these require lower minimum coverage. However, provisioning needs might increase further in the period ahead, also given the continuing uncertainty around the possible impact of the oil supply shock on banks’ asset quality.

Chart 3.9

NPL coverage ratios continue to decline, particularly in countries in which NPL ratios increased

a) Change in coverage of NPLs by LLRs, collateral and financial guarantees

b) Change in NPL ratios and NPL coverage ratios, by country

(Q1 2021-Q2 2026, percentage points)

(Q1 2021-Q2 2026, percentage points)

Sources: ECB (supervisory data) and ECB calculations.
Notes: Based on all significant institutions. Panel a: LLRs stands for loan loss reserves (i.e. accumulated impairments). Combined coverage is defined as the ratio of LLRs, collateral and financial guarantees to NPLs. NFC stands for non-financial corporation; HH stands for household. Panel b: based on countries with at least three significant institutions. Excludes an outlier.

Direct euro area bank lending to firms in the AI supply chain remains limited, with loan quality weakening in a number of sectors. Based on available data on banks’ direct lending exposures, euro area banks appear to play a relatively limited role in providing funding to firms in the AI supply chain.[28] However, this narrow measure may underestimate total exposures due to data limitations and should be treated with caution.[29] When an alternative, broader measure is used, total AI-related lending amounts to around €75 billion, just 1.1% of euro area banks’ total committed exposures to corporates.[30] The quality of exposures to software companies and semiconductor manufacturers has weakened since 2023, while loans to data centres have been performing extremely well (Chart 3.10, panel a). The NPL ratio for software and semiconductor firms was above the total corporate NPL ratio in the first quarter of 2026. For the software sector, increases in the NPL ratio may have been driven in part by non-AI-related firms being crowded out or replaced, given that AI disrupts traditional software development processes.[31]

Chart 3.10

Risks to asset quality might emerge from exposures to some AI-related sectors, while cyberattacks continue to be a key source of operational risk

a) NPL ratio of exposures to AI-related firms and related sectors

b) Top four drivers of operational risk seen by euro area banks

(Q1 2019-Q2 2026, percentages)

(Sep. 2024-Mar. 2026, percentage agreeing)

Sources: ECB (AnaCredit, RIAD), Moody’s, GLEIF, EBA and ECB calculations.
Notes: Panel a: based on all banks. For definitions of AI-related firms and AI-related sectors see footnotes 29 and 30. Panel b: banks were asked in the European Banking Authority’s Risk Assessment Questionnaire to indicate what they consider to be the main drivers of operational risk. They could select up to three options out of seven.

Besides posing risks to banks’ asset quality, digitalisation and frontier AI developments are also increasing banks’ operational risk. Banks rank cyber risk as the largest driver of operational risk by a large margin (Chart 3.10, panel b). Agentic AI systems are enabling cyberattacks to become faster and more sophisticated. This is favouring attackers as AI models are becoming available at lower cost after just a short time lag. While euro area banks have a good level of cyber security, defenders are disadvantaged by the cost of deploying frontier models, governance challenges and slower decision-making. Combining this with banks’ reliance on a few key third-party ICT providers, the risk of systemic disruptions from cyber incidents is increasing (see Special Feature B).

Box 5
The effects of oil supply shocks on corporate resilience, bank asset quality and lending

Prepared by Stephan Fahr, Emanuele Franceschi, Dejan Krusec and Cosimo Pancaro

The global supply of oil was suddenly curtailed when the latest conflict in the Middle East broke out, resulting in higher and more volatile prices for the commodity. Around 20% of global supply has been hit by the effective closure of the Strait of Hormuz to commercial traffic since March 2026.[32] As a result, oil prices have reached levels similar to those observed after the Russian invasion of Ukraine and their volatility has increased significantly. Against this background, this box investigates the impact of oil supply shocks on corporate resilience, bank asset quality and lending activity.

More

3.4 High aggregate capital ratios are supporting banking sector resilience

The banking sector remains well capitalised overall, supported by high profitability. The aggregate Common Equity Tier 1 (CET1) ratio and headroom have stabilised at high levels (Chart 3.11, panel a), thanks in part to the regulatory reforms implemented since the global financial crisis. Euro area banks have channelled about a quarter of their profits into additional equity, and retained earnings remain the largest driver of capital formation (Chart 3.11, panel b), while allowing for steady shareholder remuneration. A growing number of banks are also using synthetic risk transfers (see Box 6) to transfer credit risk outside the banking sector and lower their capital requirements. While these factors have supported banks’ capital headroom in aggregate, the dynamics differ across profitability quartiles. Indeed, a quarter of banks with the largest headline profitability have reduced their capital headroom over time and are now the least capitalised (Chart 3.11, panel c). As high profitability is a first line of defence against solvency risks, these banks might feel comfortable operating at tighter voluntary buffers.[33] However, banks should remain vigilant and ensure they are maintaining sufficient capital headroom in case earnings disappoint.

Chart 3.11

Stable aggregate capital ratios mask diverging trends in capital management

a) Euro area aggregate CET1 ratio and headroom

b) Use of euro area banking sector profits

b) CET1 headroom, by profitability quartile

(Q1 2020-Q2 2026, percentages of risk-weighted assets)

(Q2 2026, € billions)

(Q2 2022-Q2 2026, percentages)

Sources: ECB (supervisory data) and ECB calculations.
Notes: Panel a: capital headroom is the difference between the implied CET1 capital need and banks’ actual CET1 ratios. The implied CET1 need is banks’ overall capital requirement + Pillar 2 guidance to be covered by CET1, plus any AT1/T2 shortfall that needs to be absorbed by CET1 and any eligible liability shortfall that must also be covered by CET1, also accounting for capital needs arising from leverage ratio and MREL regulations. Panel b: four-quarter trailing sum of banks’ total profits and their allocation between retained earnings, dividends and repurchased treasury shares. Panel c: ratio of CET1 capital need to risk-weighted exposures. Profitability quartile is defined as the quartile of a given bank in Q2 2026 in terms of return on equity (ROE), such that banks do not move across quartiles over time. Data are shown only for the second quarter in each year.

Box 6
Bank reliance on credit protection providers via synthetic risk transfers

Prepared by Allegra Pietsch, Daniel Dieckelmann, Maciej Grodzicki, Francesca Lenoci and Manuela Storz

Synthetic risk transfers (SRTs) allow a bank to transfer the credit risk of a loan portfolio to a third party, while the loans themselves stay on the bank’s balance sheet. SRTs are the most prevalent form of capital relief securitisation in the euro area.[34] Unlike cash securitisations, SRTs do not involve selling the loan portfolio. Instead, the bank retains the loans on its balance sheet and collects interest payments, manages credit risk and maintains customer relationships. Meanwhile, credit losses on the portfolio are covered by investors up to a limit, against a premium.[35] SRTs lead to lower capital requirements by reducing banks’ risk weights on securitised loans although, in effect, they substitute bank equity for temporary loss-absorbing capital from investors.[36],[37]

More

3.5 Geopolitical headwinds and elevated financial stability risks underscore the need to preserve resilience

Preserving resilience remains the key priority for macroprudential policy.[38] Although the euro area banking sector is healthy and credit dynamics are favourable, albeit unevenly so across countries, financial stability risks remain elevated. Should these risks materialise, it is crucial that there is adequate macroprudential space in the form of releasable capital buffers for the relevant authorities to act swiftly and effectively. This is because releasing these buffers gives banks greater capacity to absorb losses and supports the continued provision of credit to households and firms.[39] As capital requirements are not currently constraining the supply of bank credit, and in the absence of signs of widespread loss materialisation, macroprudential capital buffers should remain in place to preserve banking sector resilience.

Borrower-based measures should continue to serve as a structural backstop, where feasible under national legislation, to safeguard sound lending standards throughout the real estate cycle. The recent recovery in residential real estate markets has become more apparent in several euro area countries, driven by improving financing conditions and a gradual strengthening of housing demand. While vulnerabilities remain contained in the euro area as a whole, developments vary substantially across countries, with some already experiencing strong dynamics in residential real estate prices and in some cases also mortgage lending (see Section 1.5 and Box 1). In light of this, existing borrower-based measures should be maintained to prevent an erosion of lending standards and contain the build-up of vulnerabilities among borrowers. Countries in which the residential real estate market is gaining momentum more rapidly might also consider tightening existing measures, while other countries could – where feasible under national legislation – adopt such measures as permanent structural backstops to prevent the build-up of vulnerabilities in the future.[40]

The ECB supports the European Commission’s communication on the competitiveness of the EU banking sector and welcomes its strategic direction.[41] Banking sector resilience is a key prerequisite for economic growth and competitiveness.[42] The regulatory reforms that were implemented following the global financial crisis have helped to ensure that the banking system is well capitalised. This has enhanced banks’ capacity to maintain credit provision in times of stress, mitigated excessive risk taking and promoted sound risk management practices. Furthermore, there seems to be no evidence to date that reducing capital requirements in normal times would foster competitiveness by increasing lending. Efforts are needed to reduce undue complexity in the regulatory framework, while maintaining a level playing field that ensures fair competition as well as supporting innovation and minimising risks.

Financial and regulatory fragmentation of the Single Market, including the incomplete banking union, continues to hold back European banks. Europe’s competitiveness would benefit from a genuinely single European banking market that enables cross-border banks to compete on a global scale. In this context, the Eurosystem has repeatedly called for the finalisation of a European deposit insurance scheme and is looking forward to the new proposal from the European Commission. The aim of the proposal would be to ensure strong and uniform protection of depositors across Europe and simplify the structure of the deposit insurance framework by better aligning the decision-making of national and EU authorities under the crisis management framework with the financial responsibilities for financing resolution and deposit insurance measures. The ECB has already taken concrete steps to support cross‑border banking integration within the existing framework.[43] However, further steps are needed to enable a more efficient allocation of resources within cross-border banking groups, for example by applying cross-border liquidity and capital waivers, subject to adequate safeguards. Such safeguards should promote resilience across credit institutions at the solo, subsidiary, branch and consolidated levels, including through fair and timely transfers of resources within banking groups, particularly in times of stress.

While acknowledging the benefits of the current regulation, the ECB considers the cyber risk stemming from frontier artificial intelligence (AI) models to be substantial. The EU Digital Operational Resilience Act[44] strengthens the incident reporting framework, enhances banks’ preparedness to deal with cyberattacks and facilitates closer monitoring of third-party risks by supervisors. At the same time, however, progress in the field of frontier AI models is making it possible to carry out cyberattacks more quickly and cheaply, and with greater sophistication. The ECB is paying close attention to these developments, assessing their implications from a microprudential and macroprudential perspective (see Special Feature A), and taking action where needed.[45]

Lastly, amendments to the Markets in Crypto-Assets Regulation (MiCAR)[46] could help realise the potential of distributed ledger technology and tokenisation to foster innovations and efficiency in the financial sector. Distributed ledger technology and tokenisation could contribute to the development of more integrated and efficient financial market infrastructure in Europe. In this context, MiCAR provides a robust, balanced and harmonised regime for crypto-asset service providers and issuers. The ECB welcomes the European Commission’s intention to review MiCAR to ensure that it remains fit for purpose while safeguarding the stability of the financial system. In this regard, the ECB considers that the minimum deposit requirements for stablecoins could be removed and replaced with minimum liquidity requirements for reserve assets. This would provide issuers with more flexibility and still maintain their ability to meet redemption requests promptly.[47] It could also reduce potential channels for contagion between banks and stablecoins while ensuring that reserve assets remain sufficiently liquid. Other key issues to be considered in MiCAR include the need for additional measures to address financial stability concerns about stablecoin issuance and the further harmonisation of the legal framework for tokenised assets.[48]

4.1 The non-bank financial sector remains vulnerable to rising risks from correlated valuation losses

High equity valuations and elevated duration risk are leaving non-banks’ portfolios vulnerable to simultaneous valuation losses. Equity holdings of euro area non-banks have increased significantly during 2026, driven primarily by rising stock market valuations that have partially reversed more recently. By contrast, debt portfolios have grown to a lesser extent, as non-banks have purchased additional debt securities while simultaneously experiencing valuation losses amid rising interest rates following the start of the Middle East conflict (Chart 4.1, panel a). Looking ahead, a deterioration in the macroeconomic outlook and renewed energy price pressures could raise credit risk and expose portfolios to further valuation losses. In addition, duration risk in non-banks’ bond portfolios is still high, despite declining recently for pension funds and investment funds, leaving their longer-duration holdings in particular exposed to a further rise in sovereign bond yields (Chart 4.1, panel b). Non-bank portfolio valuations therefore continue to be vulnerable to geopolitical shocks, concerns about the positive narrative around artificial intelligence (AI) and increases in interest rates, which may trigger simultaneous losses on equity and bond holdings. Furthermore, these concurrent risks are being compounded by the increased stock-bond correlation (see Chapter 2), which limits the benefits of portfolio diversification.

Chart 4.1

High equity valuations and elevated duration risk are leaving non-banks’ portfolios vulnerable to simultaneous valuation losses

a) Changes in non-banks’ portfolio holdings and valuations, by asset type

b) Duration risk in non-banks’ portfolios

(1 Jan.-10 Nov. 2026, percentages)

(Q1 2023-Q2 2026, estimated pp losses for a 1pp increase in interest rates)

Sources: ECB (CSDB, SHS), LSEG, Bloomberg Finance L.P., EPFR Global and ECB calculations.
Notes: Panel a: changes in holdings from July onwards are estimates based on flows into equity and bond funds. Panel b: duration risk is measured as modified duration, which estimates the approximate percentage change in a portfolio’s value for a 1 percentage point (pp) change in interest rates; higher values indicate greater interest-rate sensitivity.

Non-banks continue to play an important role as absorbers of sovereign debt, increasing their purchases amid record high issuance. Non-banks have continued to be a key purchaser of newly issued sovereign debt, acquiring around 20% of issued debt with a maturity of less than five years and more than 30% of issued debt with a maturity of more than five years over the last 12 months. During this period, the absorption of bonds at the very short end (below one year) and the very long end (over 15 years) has declined, while absorption of issued debt with a maturity of one to two years and five to 15 years is at a historical high (Chart 4.2, panel a). Reflecting their respective preferred investment horizons, money market funds have been highly active in absorbing very short-term issuances, while insurance corporations and pension funds have continued to purchase securities at the long end (Chart 4.2, panel b).

Chart 4.2

Non-banks’ sovereign debt absorption has remained high amid strong issuance

a) Distribution of net issuance of sovereign bonds and non-bank absorption

b) Euro area sovereign bonds absorption, by sector

(Q1 2014-Q2 2026; residual maturity buckets, € billions)

(Q3 2025-Q2 2026; residual maturity buckets, € billions)

Sources: ECB (CSDB, SHS) and ECB calculations.
Notes: Panel a: interquartile (box) and minimum-maximum (line) range of annual net issuance of sovereign debt and non-bank absorption, along with the latest observation. Net issuance is defined as cumulative quarter-on-quarter net changes in issued securities. Panel b: absorption refers to cumulative quarter-on-quarter net changes in the holdings of the euro area non-bank sector.

Non-banks’ portfolios are highly sensitive to shifts in AI sentiment, given their large and concentrated exposures to US technology and AI-related firms. The share of non-banks’ holdings in US technology and AI-related firms has increased on the back of substantial valuation gains and now accounts for between 16% and 26% of their equity investments. In addition to equity, non-banks have also started to invest in debt issued by AI-related firms, and particularly by US hyperscalers seeking to fund their capital expenditure (Chart 4.3, panel a; see Section 2.3 and Box 3). While holdings of AI-related debt have risen most notably among investment funds and pension funds, they are so far limited in absolute terms and do not appear to be crowding out non-bank absorption in euro area bond markets (see Box 4). The growing portfolio allocation to AI-related firms reflects broader developments in global financial markets. Although the share of AI-related firms in euro area non-banks’ equity portfolios is still below that in major US equity markets, such exposures are more concentrated than in the broader euro area equity market (Chart 4.3, panel b).[49] The concentration in predominantly US dollar-denominated equities also increases non-banks’ sensitivity to US-specific shocks and, in the absence of hedging, related exchange rate fluctuations. Furthermore, concentration risk has increased as valuations have continued to rise, especially among the largest US technology and AI-related firms, amid a strong earnings outlook centred on optimistic expectations for future AI demand and adoption (see Special Feature A). The valuations of non-banks’ equity holdings have followed this trend and, in recent years, have remained above those of the broader market (Chart 4.3, panel c). A sudden shift in expectations, as for instance observed after the recent disappointing earnings reports of large US tech firms, could trigger renewed price declines across firms and amplify portfolio losses.

Chart 4.3

Non-banks’ equity portfolios are increasingly concentrated in AI-related equities, while holdings of debt issued by AI-related firms are rising

a) Share of AI-related firms in euro area non-banks’ portfolios, by financial sector

b) Share of AI-related firms in euro area non-banks’ equity portfolios and benchmark indices

c) CAPE ratios of non-banks’ portfolios and selected equity indices

(Q2 2024, Q2 2025, Q2 2026; percentages)

(Q2 2026, percentages)

(Q1 2020-Q2 2026, CAPE ratios)

Sources: ECB (CSDB, SHS), LSEG and ECB calculations.
Notes: Panel a: bubble size represents the total holdings of AI-related equity and debt for each sector. Panels a and b: “AI-related firms” comprises 167 firms (and their respective subsidiaries) identified as being involved in AI model development, AI chip production, AI cloud provision, AI data centre and infrastructure provision, or related utilities. Panel c: cyclically adjusted price/earnings (CAPE) ratios at portfolio and index level, using earnings for the last ten years, adjusted for inflation and averaged. Portfolio-level prices and earnings are weighted by the instrument-level holdings of non-banks.

4.2 Rising leverage and redemption pressures could expose liquidity vulnerabilities in investment funds

Investment fund flows are showing an increasingly divergent picture, as demand for duration remains muted while appetite for risk assets remains high. Flows into long-term sovereign bond funds have remained subdued since the beginning of 2026, against a backdrop of fiscal concerns and expectations that interest rates will rise. Flows into short-maturity funds, by contrast, have experienced an upward trend. Risk assets – in particular high-yield bond funds – rebounded strongly after the first few weeks of the conflict in the Middle East (Chart 4.4, panel a). There were particularly strong inflows into global leveraged equity exchange-traded funds (ETFs) from June, indicating increased investor appetite for leveraged equity exposures. At the same time, flows into euro area-domiciled leveraged ETFs have been very limited in scale and are lagging behind demand for global leveraged ETFs, while leveraged ETFs overall remain smaller in size than non-leveraged ETFs (Chart 4.4, panel b).

Chart 4.4

Fund flows reflect preference for lower duration and AI-fuelled risk-on sentiment

a) Cumulative flows into euro area-domiciled bond funds

b) Cumulative flows into equity ETFs, by domicile

(2 Jan.-10 Nov. 2026, percentage shares of total net assets)

(2 Jan.-10 Nov. 2026, percentage shares of total net assets)

Sources: EPFR Global and ECB calculations.
Notes: “Start of Middle East War” refers to 2 March 2026, the first trading day after the start of the conflict on 28 February 2026. “MoU announced” refers to 12 June 2026, the day on which the temporary truce framework (MoU) between the United States and Iran was unofficially announced. Panel a: the sample is based on euro area-domiciled funds investing primarily in western European bonds, which include bonds issued in euro area countries, Norway, Sweden, Switzerland and the United Kingdom. Short-term bond funds invest in bonds with durations of between zero and four years. Long-term bond funds invest in bonds with durations of six years or more. A bond fund with a duration of five years is classified as intermediate term.

The increasing use of leverage in funds with concentrated artificial intelligence (AI) exposures could amplify a potential market downturn. Larger euro area investment funds have a higher share of their portfolios allocated to AI-related firms than their smaller counterparts and are also exposed to risks from synthetic leverage in equity derivatives (Chart 4.5, panel a). While funds’ exposures to AI-related assets have benefited from strong market performance and continued investor demand, they could also make funds more vulnerable to losses if there were an abrupt reassessment of growth prospects and earnings expectations, even for just a few AI-related firms. This was also illustrated by the recent sell-off of AI-related stocks after disappointing earnings releases of large US tech firms. Concentration risks may be particularly pronounced where they interact with leverage, as for example in the case of the US hedge fund Situational Awareness in July 2026.[50] While synthetic leverage at euro area equity funds remains well below the levels observed at hedge funds, sudden price swings could lead to significant portfolio losses and pressure to deleverage to meet margin calls. This could result in forced asset sales and procyclical fund share redemptions, which would have the potential to further amplify any market downturn.[51]

Chart 4.5

High concentration in AI-related assets among large funds is compounded by rising leverage and cross-border interconnectedness between hedge funds

a) Euro area investment funds’ share of AI-related equities and synthetic leverage

b) Gross repo borrowing by US and global hedge funds

(June 2026; GNV/IM, percentages of total assets)

(Q1 2023-Q2 2026; left-hand scale: € trillions,
right-hand scale: percentages)

Sources: LSEG Lipper, Office of Financial Research, ECB (IVF, SFTDS, EMIR) and ECB calculations.
Notes: Panel a: euro area equity funds are grouped into terciles by assets under management (AuM). Synthetic leverage is the gross notional of equity derivatives divided by the initial margin (GNV/IM), shown as the AuM-weighted average within each group. Bubble size represents the gross notional amount of equity derivatives holdings in each group. “AI-equity holdings” refers to equities issued by AI-related firms, which comprise 167 firms (and their respective subsidiaries) identified as being involved in AI model development, AI chip production, AI cloud provision, AI data centre and infrastructure provision, or related utilities. The hedge fund bubble covers only euro area hedge funds whose holdings contain more than 50% in equities. Panel b: US hedge funds are global hedge funds managed by US-based asset managers. End-of-quarter data presented.

Hedge fund leverage and cross-border activity could carry the risk of spillovers to euro area sovereign bond markets. Leverage in the hedge fund sector remains elevated globally. This is particularly the case for US funds, which have more than doubled their repo borrowing since 2023 (Chart 4.5, panel b). This is also reflected in the euro-denominated repo market, where hedge fund borrowing has increased, implying that global hedge funds are increasingly using euro area sovereign bonds as collateral in their investment strategies.[52] While rising repo borrowing can enhance market resilience by facilitating secured funding and supporting liquidity redistribution, it also increases the risk of shocks originating outside the euro area spilling over into euro area sovereign bond markets. This risk is particularly acute during periods of market stress when hedge funds may be forced to reduce leverage, especially if their liquidity preparedness is inadequate. As a result, hedge funds’ procyclical deleveraging may amplify falling prices and rising volatility in sovereign bond markets (see Box 7).

Risks from liquidity mismatches in parts of the investment fund sector remain elevated, as rising interest rates may increase redemption pressures. High-yield corporate bond funds may be at risk of incurring renewed outflows, as valuation losses and higher interest rates could encourage investors to rebalance their portfolios. While euro area high-yield bond funds have increased their holdings of high-quality liquid assets (HQLA) recently, levels remain low overall, particularly among the largest funds. Around half of the largest high-yield bond funds currently have HQLA holdings that are insufficient to cover the outflows associated with a redemption shock (Chart 4.6, panel a). Furthermore, real estate funds could face rising redemption requests, as interest rates are expected to move higher. Despite a rebound in commercial real estate prices in 2025, flows into euro area real estate funds have remained subdued and there may be new procyclical outflows as interest rates rise (Chart 4.6, panel b). It remains essential to align redemption terms with asset liquidity to avoid forced asset sale spirals that would have the potential to transmit stress to other parts of the financial system (see Section 4.4).

Elevated credit risk and an uncertain macroeconomic outlook could increase the risk of further redemptions in private credit funds. Redemptions from US private credit funds remain elevated and are still being gated to avoid forced deleveraging or asset sales (Chart 4.6, panel c).[53] A high share of comparatively weak borrowers in private credit portfolios and the high proportion of floating-rate loans increase the risk of fund performance declining in the event of a worsening macroeconomic outlook or rising interest rates. Although private credit funds are unlikely to pose systemic risks to financial stability in isolation, their increasing involvement in AI financing makes them more susceptible to spillover risks from changes in investor AI sentiment (see Chapter 2).

Chart 4.6

Larger funds face higher liquidity mismatches, while real estate and private credit funds are exposed to increased redemption risk

a) HQLA levels in euro area-domiciled high-yield corporate bond funds, by fund size

b) Cumulative net flows into euro area-domiciled real estate funds and CRE prices

c) Redemptions from US private credit funds

(Q1 2025-Q1 2026, percentages of total net assets)

(Q1 2023-Q2 2026; left-hand scale: percentages of total net assets,
right-hand scale: index: Q4 2022 = 100)

(Q1 2024-Q2 2026, percentages of total net assets)

Sources: LSEG Lipper, company SEC filings, ECB (IVF, RESC) and ECB calculations.
Notes: Panel a: the average redemption shock is defined as the average redemptions observed during the top 10% worst 30-day outflows since 2007. Panel b: Q1 2026 and Q2 2026 data for commercial real estate (CRE) prices are approximated using German prices due to a longer reporting lag for euro area data. Panel c: data for a total of 26 non-traded business development companies that reported redemption requests in the periods shown.

Box 7
The footprint of hedge funds in euro area sovereign bond markets via repo markets

Prepared by Carlos Miguel Aguiar da Glória, Maurizio Michael Habib, Andrea Sante and Michael Schmidt

The sudden unwinding of large hedge fund positions in sovereign bond markets can pose risks for financial stability. Understanding hedge funds’ behaviour during periods of market stress is increasingly important, given the rapid growth of the sector and its substantial exposure to US and European sovereign bonds. Hedge funds now account for a sizeable share of trading activity in euro area sovereign bonds.[54] These funds typically use flexible trading strategies that rely heavily on leverage, particularly when trading such bonds. These include relative value strategies such as basis trades, swap spread trades and auction cycle arbitrage, as well as directional strategies such as yield curve arbitrage or sovereign spread trades.[55] Although hedge funds may support market liquidity in normal times, tighter financing conditions or internal risk limits during periods of stress can force them to quickly unwind their positions.[56] Such actions can dry up market liquidity, disrupt benchmark asset prices and further exacerbate market turmoil.[57] The objective of this box is to analyse how global hedge funds adjusted their exposures to euro area sovereign bond markets via the repo market, focusing on the March 2026 period of bond market volatility triggered by the start of the latest conflict in the Middle East. This unexpected geopolitical event provides a unique opportunity to analyse how hedge funds responded to a large shock in these markets.[58]

More

4.3 Insurance and pension fund sectors weather rising yields and market volatility

Euro area insurance corporations and pension funds remain resilient, supported by strong solvency positions, stable liquidity buffers and generally good levels of profitability. Insurance corporations in the euro area continue to demonstrate strong solvency positions, underpinned by robust own funds and prudent risk management practices. Liquidity risks are contained, with insurance corporations able to meet short-term obligations without incurring excessive costs. Profitability metrics for the sector remain stable, as market perceptions continue to be favourable. Evidence for this is provided by the sector’s positive stock market performance relative to broader indices and its solid price/earnings ratios.[59] Pension funds, particularly those offering defined benefit schemes, are maintaining adequate levels of reserves and funding, reflecting their stable solvency metrics. Liquidity risks for pension funds are being managed effectively, with institutions continuing to meet payment obligations.[60]

Despite their strong aggregate solvency and liquidity positions, euro area insurance corporations and pension funds remain exposed to geopolitical, sovereign and market shocks. The balance sheets of insurance corporations and pension funds have evolved gradually in recent years, their composition reflecting an increasing appetite towards riskier, alternative and less liquid asset classes such as private credit, private equity and infrastructure (Chart 4.7, panel a).[61] Insurance corporations have reduced their allocation to debt securities, particularly euro area sovereign debt, while increasing their exposure to investment funds and alternative assets. Pension funds have bolstered their holdings of equity and non-tradable assets, while also lowering the share of euro area government securities in their debt portfolios. Nevertheless, the euro area’s insurance and pension fund sectors remain important sovereign bond investors, holding 10.7% and 4.1% of the region’s sovereign debt respectively, with approximately half of these holdings concentrated in maturities above ten years. Home bias has declined somewhat over the last decade and non-domestic issuers now account for the majority of public sector bond holdings in both sectors (Chart 4.7, panel b). These exposures create a potentially important channel through which sovereign stress and geopolitical shocks can be transmitted to insurance corporations and pension funds through valuation losses, collateral pressures and broader macro-financial spillovers.

The liabilities side of both sectors is undergoing structural transformation, reducing the market sensitivity of capital positions. The insurance sector has experienced a growing uptake of unit-linked products, while pension systems are continuing to transition from defined benefit to defined contribution arrangements.[62] These developments are generally reducing the interest rate sensitivity of capital positions. Unit-linked products pass investment risk directly to policyholders, meaning that when interest rates change, the value of assets and liabilities held in such products move in tandem, so there is no direct interest rate-related impact on the insurer’s capital position.[63] By contrast, the value of assets held for non-unit-linked products, such as traditional life insurance policies, typically changes by less than that of related liabilities, as the duration of liabilities is longer than that of the assets backing them (negative duration mismatch). When interest rates fall, the value of liabilities rises more than the value of assets, eroding capital; when interest rates rise, the value of liabilities falls more than assets, which benefits capital.[64] Similarly, the ongoing shift towards defined contribution pension schemes is transferring a larger share of market risk to beneficiaries (Chart 4.7, panel a). It is also reducing the duration and sensitivity of pension fund equity to interest rate developments while also affecting the markets for long-duration securities.

Chart 4.7

There has been a shift to defined contribution models, coupled with increased diversification into riskier assets while maintaining low concentration in liquid portfolios

a) Aggregated balance sheets of euro area insurance corporations and pension funds, by asset and liability type

b) Marketable securities holdings of euro area insurance corporations and pension funds, by asset type

(Q4 2021, Q4 2023, Q2 2026; € trillions)

(Q2 2026; left-hand scale: € trillions, right-hand scale: percentages)

Sources: ECB (CSDB, ICB, PFBR, SHS) and ECB calculations.
Notes: Panel a: technical reserves recorded on the assets side of an insurance corporation’s balance sheet specifically relate to exposures that have been reinsured. Panel b: HQLA stands for high-quality liquid assets.

Euro area insurance corporations and pension funds have recently demonstrated resilience to rising yields and market volatility. When yields rose sharply in 2022, the decline in marketable securities (including sovereign bond holdings) was less than the fall in technical reserves and pension entitlements (Chart 4.8, panels a and b).[65] This specifically reflects the liability-driven nature of insurers’ balance sheets, where liabilities heavily influence balance-sheet dynamics. The data, which cover developments up to the second quarter of 2026, also reflect the sectors’ ability to withstand geopolitical tensions and inflation-driven rate hikes. All the adjustments occurred without significant capital depletion, underscoring the sectors’ capacity to absorb shocks.

Chart 4.8

Resilience amid interest rate shocks and structural shifts in insurance and pension fund portfolios and liabilities

a) Insurance corporations: quarterly changes in yields, yield-driven valuations and technical reserves

b) Pension funds: quarterly changes in yields, yield-driven valuations and pension entitlements

(Q1 2020-Q2 2026; left-hand scale: € billions,
right-hand scale: percentages)

(Q1 2020-Q2 2026; left-hand scale: € billions,
right-hand scale: percentages)

Sources: ECB (ICB, PFBR, YC) and ECB calculations.
Notes: The analysis uses aggregate data for the euro area insurance and pension fund sectors, focusing on the aggregate marketable securities positions on the assets side and insurance technical reserves or pension entitlements on the liabilities side. Capital is replenished when the change in marketable securities exceeds the change in technical reserves/entitlements. Conversely, capital is depleted when the change in marketable securities is smaller than the change in technical reserves/entitlements. Whereas derivative positions are not taken into account in these charts, separate calculations show that adding net financial derivatives as an approximation of hedging to the above does not significantly change the results. Panel a: insurance technical reserves are primarily relevant for the life insurance business. As at Q2 2026, the euro area sector’s total assets were distributed across reinsurance corporations (10.1%), life insurance corporations (36.8%), non-life insurance corporations (16.2%) and composite insurers offering both life and non-life insurance (36.9%). This segmentation reflects the diverse business models within the sector and their differing exposures to macroeconomic and market risks.

Although the structural characteristics of euro area insurance corporations and pension funds mitigate risks, they warrant continual monitoring. While structural features such as long liability durations, diversified portfolios and risk transfer mechanisms enhance resilience, the pace and magnitude of future interest rate adjustments remain a key risk factor. The 2022 liability-driven investment episode in the United Kingdom highlights how rapid increases in sovereign yields can generate systemic liquidity pressures, even for fundamentally solvent institutions. Although euro area insurance corporations and pension funds have not exhibited similar vulnerabilities, the episode underscores the importance of liquidity preparedness and robust risk management practices in mitigating potential risks.[66]

Insurance protection gaps, particularly for natural catastrophes including those attributable to climate change, pose significant financial stability concerns, as uninsured losses strain households, firms, governments and the financial system. The summer of 2026 underscored these risks, with wildfires, droughts and heatwaves ravaging Europe – the world’s fastest-warming continent.[67] Historically, only about a quarter of natural catastrophe-related economic losses across all hazard types in the region have been insured (Chart 4.9).[68] This insurance protection gap is shifting the financial burden to society at large, increasing economic fragility. Closing insurance protection gaps is essential to bolster resilience, ensuring households, businesses and governments can better absorb future shocks in an era of intensifying climate hazards.

Chart 4.9

Economic losses from natural catastrophes in euro area countries are far more likely to be uninsured than insured across all types apart from windstorms

Insured and uninsured economic losses from natural catastrophes in the euro area, by peril

(2025; left-hand scale: € billions, right-hand scale: percentages)

Source: EIOPA.
Notes: EIOPA insurance protection gap dashboard based on CATDAT and national data provided by national competent authorities.

4.4 A macroprudential lens is needed across the NBFI policy framework to protect market liquidity in times of stress

Concentration-related risks can amplify vulnerabilities across the non-bank financial intermediation (NBFI) sector during periods of stress, with implications for the non-bank policy framework. A trigger event, related to a shift in sentiment on artificial intelligence (AI), geopolitical developments or fiscal concerns could pose risks for the non-bank sector. It could, for example, lead to a rapid unwinding of positions held by highly leveraged funds, as well as selling pressures from redemption requests in open-ended funds (OEFs) and margin calls from the wider sector. Previous stress episodes have shown how large market footprints, together with concentrated portfolios of cohorts of similar funds, can amplify the market impact due to vulnerabilities related to leverage and liquidity mismatch. This becomes especially relevant where the NBFI sector draws on the same sources of liquidity. Hence, it is important that measures to address non-bank vulnerabilities are appropriately calibrated to mitigate risks from concentration and crowded positioning.

Implementation of international reforms in the EU should be followed by a review of their effectiveness in preserving market liquidity during stress episodes. Drawing on the 2021 Financial Stability Board (FSB) proposals to enhance the resilience of money market funds (MMFs), EU authorities are currently implementing supervisory guidance on MMF liquidity buffers. Instead of raising minimum liquidity requirements as proposed by the ESRB,[69] in May 2026 the European Commission issued guidance on how liquidity buffers might be used, particularly to address rising redemption requests during times of market stress as well as identifying market resilience levels for weekly liquid assets above the minimum regulatory levels.[70] National authorities are now implementing this guidance in their respective domestic frameworks.[71] As MMFs are key providers of short-term funding to banks, companies and public entities, liquidity shortfalls in times of stress can force asset sales that disrupt this funding, as was the case in March 2020. Authorities should therefore apply the guidance strictly to ensure that MMFs maintain liquidity buffers in line with supervisory expectations. In addition, in line with the guidance, MMFs should make effective use of these buffers in meeting redemptions in periods of stress rather than preserving them by selling less liquid assets, which amplifies strains in core short-term funding markets. Further work is also needed on classifying open-ended funds according to the liquidity of their assets, and the FSB’s 2024 recommendations on liquidity preparedness for margin and collateral calls have yet to be implemented.

EU authorities need a dedicated instrument to address ex ante structural liquidity mismatch in OEFs, including for cohorts of similar funds. The existing measures to mitigate liquidity risk, such as requirements for funds to have liquidity management tools in place, have generally been implemented from a microprudential perspective. For similar types of fund that have concentrated portfolios and crowded positions, however, such an approach may be insufficient to ensure that their collective behaviour does not lead to market dysfunction. A key policy gap in this regard is a dedicated tool to address structural liquidity mismatch in OEFs. As set out by the FSC high level task force on NBFI, such a tool should be similar to the existing legislative power to address systemic risk stemming from excessive leverage (see Article 25 of the AIFMD).[72] It could be conferred through targeted amendments to the Undertakings for Collective Investment in Transferable Securities (UCITS) Directive[73] and the Alternative Investment Fund Managers Directive (AIFMD).[74] The tool should allow authorities to address systemic risk ex ante, for example, by specifying longer notice periods for particular cohorts of funds. The intention is that it should be used to mitigate the build-up of vulnerabilities in benign times rather than as a crisis management tool.

Supervisory authorities need to take steps to ensure that the NBFI sector has adequate liquidity to meet margin calls. During stress episodes, the forced selling of assets to meet margin calls can depress asset prices and, in turn, trigger further margin calls. Moreover, as historical correlations between offsetting positions break down, non-banks might need to mobilise cash or collateral across several portfolios at once. It is important that the EU implement the FSB’s 2024 recommendations on margin preparedness, which cover the broader NBFI sector’s liquidity risk management and governance, its stress testing and scenario design, as well as its collateral management practices, and aim to mitigate liquidity risks arising from spikes in margin and collateral calls. Crucially, they also require non-banks, as part of their liquidity risk management processes, to (i) consider whether they participate in crowded trading strategies or concentrated market segments, and are therefore more likely to liquidate the same assets at the same time as others; and (ii) incorporate estimates of how deteriorating market liquidity affects their anticipated liquidity needs, as well as the actual liquidity of assets held and the market impact of how they meet those liquidity needs.

The current EU framework for non-bank leverage is not fully equipped to monitor and address risks stemming from complex cross-border funds. The rapid rise of hedge funds in EU sovereign bond markets poses several challenges for EU authorities. At present there is limited information on the wider positions held by cross-border hedge funds that are dominant in sovereign markets, highlighting a key need to close data gaps to monitor risks from hedge fund leverage (see Box 7). Closing these data gaps is also crucial to assess the potential for cross-border contagion arising from EU banks’ credit exposures to hedge funds outside the EU, which warrant further scrutiny. It is crucial that the FSB makes progress in addressing these data challenges. In particular, removing barriers to effective cross-border information sharing and strengthening public disclosure would help authorities to assess the systemic risk from hedge fund leverage. There is a major policy gap regarding how best to address risks from complex leveraged strategies. A breakdown in the effectiveness of hedging strategies, along with the increased potential for a crowding of trades, could leave core financial markets vulnerable to adverse market dynamics. Implementation of the FSB’s 2025 recommendations on NBFI leverage is important. However, there is also a need for additional FSB guidance on how to implement activity-based and entity-based measures, including how to address risks from concentration and crowded positioning. A key priority is to address risks from zero haircuts and ensure that margin add-ons capture concentration risks.

The rise of private credit highlights how data gaps and the absence of common definitions can hinder the assessment of emerging risks. Euro area financial institutions currently appear to have limited direct exposure to private credit, implying that private credit in isolation is unlikely to be a source of systemic financial instability at present. That said, large data gaps, including data sharing and issues relating to definitions and taxonomies, prevent a comprehensive assessment of the financial stability risks stemming from private credit. The recent Call for Technical Advice[75] from the European Commission to the European Supervisory Authorities to enhance transparency of private credit exposures is very welcome. Global efforts are needed to close data gaps and develop consistent definitions and taxonomies that will enhance cross-jurisdictional comparability. In the EU, it is important to ensure greater data access across authorities to facilitate monitoring of cross-sectoral exposures and introduce a dedicated reporting category for private credit funds within the AIFMD.

Progress on a macroprudential framework for asset managers would help ensure that deeper integration within the sector is accompanied by greater resilience. To fully unlock the benefits of the savings and investments union capital markets should be a resilient source of funding, even in times of stress. This would require a prudential framework that can identify and address emerging systemic risk across the financial system. A more integrated approach to the supervision of funds and managers, as proposed by the European Commission in its market integration and supervision (MISP) package,[76] should be accompanied by targeted amendments to the UCITS Directive and the AIFMD with a view to developing a macroprudential approach alongside microprudential oversight.

The ECB looks forward to an agreement on the MISP package which includes key amendments to the supervisory framework for capital markets and measures to support integration across the asset management and funds sectors. Further integration of the asset management sector – a welcome goal of the MISP package – also entails greater cross-border contagion risk for financial stability. A more integrated supervisory framework would include direct supervision of certain large, cross-border capital market players at the European level, as well as enhanced coordination and additional powers for the European Securities and Markets Authority (ESMA) to promote supervisory convergence. This would help to ensure the consistent treatment of risk, promote a level playing field and reduce the potential for regulatory fragmentation or arbitrage.[77] Providing members of the European System of Central Banks (ESCB) with access to granular, cross-border data on investment funds and central counterparties (CCPs), together with clarification that central banks may use such data to fulfil their core mandates, including safeguarding financial stability, is also essential to support these objectives.

It is important that there are no carve-outs in the criteria included in the legislation for identifying systemic players. As the EU legislative process for the MISP package proceeds, discussions will focus on the criteria used to identify the large EU cross-border actors that would be subject to direct supervision by ESMA. In this context, it is key that the selected criteria correctly identify the sources of the risk posed by those institutions and ensure a level playing field in financial market infrastructure. Regarding crypto-asset service providers (CASPs), the ECB supports the European Commission’s proposal to transfer the authorisation, monitoring and enforcement powers for all CASPs to ESMA.[78] It also suggests maintaining, in the Markets in Crypto-Assets Regulation (MiCAR),[79] the category of significant CASPs – which should be subject to tighter prudential requirements – and expanding the existing criteria to include objective metrics of significance (e.g. size, cross-border activity, systemic relevance, volume of trades for platforms, volume exchanged against funds and consideration of group-wide activities).

A Financial stability implications of AI adoption in the financial industry and the real economy

Prepared by Leonardo Andaloro, Maria Leonor Puga, Sándor Gardó, Benjamin Klaus, Urban Rupnik and Christian Weistroffer[80]

The rapid adoption of artificial intelligence (AI) is reshaping the financial industry as well as the broader economy, creating both opportunities and risks. Banks and asset managers, which represent key channels of financial intermediation in the euro area, are increasingly using AI for tasks such as credit scoring, fraud detection and portfolio management. While AI adoption can support innovation and enhance operational efficiency, it may also introduce new risks and amplify existing vulnerabilities, including a reliance on a small number of technology providers that creates concentration and third-party dependencies across the financial system. At the same time, financial institutions are also increasingly exposed to AI-driven developments in the real economy. A novel firm-level indicator suggests that a sizeable share of firms, concentrated mainly in the IT and financial sectors, is significantly exposed to AI-related opportunities and risks. Policies should focus on improving the quality of data that monitor the use of AI, strengthening the oversight of critical AI providers and adapting market safeguards and supervisory capabilities to emerging risks.

More

B Decoding the cyber-AI nexus

Prepared by Sándor Gardó, Benjamin Klaus, Guido Pistorius and Jonas Wendelborn

The rapid evolution of artificial intelligence (AI) is reshaping the cyber threat landscape, with significant implications for financial stability in the euro area. This special feature explores the complex interplay between AI and cybersecurity, highlighting how advanced AI models are enabling attackers and defenders alike to enhance their capabilities. Highly capable and agentic AI systems are increasing the scale, speed and sophistication of attacks while also introducing new vulnerabilities. These include AI models themselves being exposed to novel types of attack and AI agents failing to act as their human operators intend. Critically, new AI capabilities could favour attackers, as the availability of low-cost AI agents lowers entry barriers. Meanwhile defenders remain constrained by the cost of deploying frontier models, governance challenges and slower, regulation-bound decision-making. This leaves defenders structurally disadvantaged, at least in the near term. Euro area banks, despite their high level of cybersecurity maturity, face asymmetries in access to cutting-edge AI models and rely heavily on concentrated third-party providers. This increases the risk of systemic disruption, as AI adoption transforms the channels through which cyber incidents propagate. Mitigating these risks requires banks, AI developers, cybersecurity firms and regulators to coordinate their efforts to enhance resilience, improve governance and address emerging systemic vulnerabilities.

More

C Financial stability risks from large IPOs, equity issuance and passive investing

Daniel Dieckelmann, Lavinia Franco, Paraic O’Gorman and Chiara Scommegna[81]

A new wave of equity issuance, concentrated in AI-related sectors, has brought public equity financing back to the forefront of capital markets. SpaceX’s record initial public offering (IPO) has led the way, with OpenAI and Anthropic planning to follow. This special feature assesses the financial stability implications of large IPOs and equity issuance in general, as well as how these interact with the trend towards passive investing. Primary equity financing supports financial stability by absorbing losses and limiting leverage at the firm level, but aggregate equity issuance is highly cyclical and strongly influenced by investor sentiment. Issuance peaks have historically been associated with late-cycle exuberance and weaker future returns, while equity funding tends to dry up in downturns when the ability to recapitalise matters most. The rise of passive investing is increasing the relatively inelastic demand for equities, leading to larger equity offerings and making equity financing conditions more dependent on investor flows. Although these developments do not currently constitute a systemic financial stability risk, they could amplify market cyclicality and the sensitivity of equity financing to changes in sentiment.

More

Acknowledgements

The Financial Stability Review assesses the sources of risks to and vulnerabilities in the euro area financial system based on regular surveillance activities, analysis and findings from discussions with market participants and academic researchers.

The preparation of the Review was coordinated by the ECB’s Directorate General Macroprudential Policy and Financial Stability. The Review has benefited from input, comments and suggestions from other business areas across the ECB. Comments from members of the ESCB Financial Stability Committee are gratefully acknowledged.

The Review was endorsed by the ECB’s Governing Council on 12 November 2026.

Its contents were prepared by Desislava Andreeva, Katharina Cera, Tina Emambakhsh, John Fell, Sándor Gardó, Sofia Gori, Philippe Molitor, Csaba More, Charles O’Donnell, Paraic O’Gorman, Allegra Pietsch, Manuela Storz, Marcello Tumino and Jonas Wendelborn.

With additional contributions from Ana Bandeira, Maria Leonor Carrilho Puga, Lara Coulier, Paola di Casola, Johanne Evrard, Marco Forletta, Maciej Grodzicki, Barbara Jarmulska, Domenic Kellner, Davide Magliano, Mateusz Miklas, Edward O’Brien, Guido Thomas Maria Pistorius, Jaime Hinojosa Laso, Oscar Schwartz Blicke, Chiara Scommegna, Jens Tapking, Kea von Daak, Ville Voutilainen, Fabian Wagner, Christian Weistroffer, Gilmar Witschge, Regine Wölfinger, Stefan Wredenborg and Balázs Zsámboki.


© European Central Bank, 2026

Postal address 60640 Frankfurt am Main, Germany
Telephone +49 69 1344 0
Website www.ecb.europa.eu

All rights reserved. Reproduction for educational and non-commercial purposes is permitted provided that the source is acknowledged.

For specific terminology please refer to the ECB glossary (available in English only).

PDF ISSN 1830-2025, QB-01-26-244-EN-N
HTML ISSN 1830-2025, QB-01-26-244-EN-Q


Legal Disclaimer:

EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.

Share this page:

Advanced Search Options

Search for:

Search scope:

Type:

Search in:

Date range:

The last

Sort by:

Sign up for:

Cyprus Healthcare Press

The daily local news briefing you can trust. Every day. Subscribe now.

By signing up, you agree to our Terms & Conditions.