Financial Stability Report - July 2026

Financial Stability Report Summary

Overall summary

• Vulnerabilities in risky asset valuations, sovereign debt markets, and risky credit markets, including in private credit, previously highlighted by the FPC remain, and some have become more pronounced since the December 2025 Financial Stability Report. Notably, there has been a substantial increase in the use of leverage in equity markets.
• Developments in the Middle East have affected the global risk environment materially. Despite this, the UK financial system has remained resilient and has continued to support the UK real economy, emphasising the benefits of the resilience that has been built.
• Recent rapid advances in frontier Artificial Intelligence (AI) capabilities have increased financial stability risks related to cyber and operational resilience.

The macroeconomic environment

• The conflict in the Middle East has resulted in a substantial negative supply shock to the global economy and triggered significant market reactions. Energy and some other commodity prices, as well as sovereign bond yields, have been volatile, initially rising sharply above pre-conflict levels. Market interest rates globally, including in the UK, have risen, thus tightening financial conditions. Meanwhile, risky asset prices have reached high levels.
• The signing of the Memorandum of Understanding between the US and Iran has led energy prices to fall back to just above pre-conflict levels, reducing near-term risks, and sovereign bond yields have also declined. However, substantial uncertainty remains and energy prices and interest rate markets have remained volatile.
• Persistent vulnerabilities in the financial system previously identified by the FPC could interact with further developments in the Middle East. In addition, leverage has increased in equity markets. The likelihood of these vulnerabilities crystallising at the same time has increased since the December FSR, potentially amplifying their combined impacts on financial stability.

Vulnerabilities

• Equity prices have increased especially for AI-related stocks. In global equity indices this has, in part, been driven by continuing positive earnings news since the December FSR. However, on some metrics, valuations have also become more stretched. Rising equity prices have been driven, in part, by a narrow set of AI-related companies, increasing market concentration in some global indices. There has been a significant rise in hedge fund leverage in equity markets, creating risks, including via the prime brokers which facilitate this activity and through markets that are interconnected via hedge funds’ exposures, such as sovereign debt. Retail inflows, including via exchange traded funds (ETFs), may have added momentum to the rise in equity prices, and there has also been rapid growth in assets under management of levered ETFs.
• AI-related companies’ use of credit markets has accelerated rapidly, including in public markets, private credit, leveraged and structured finance, and is set to increase further as financing needs continue to expand. This pace of investment is unprecedented historically. As of yet, there is little evidence that AI activity in these markets is crowding out the ability of other businesses or governments to access funding markets.
• AI has the potential to raise productivity across a range of sectors, and, in turn, support long-term economic growth. Developments in AI have already provided a tailwind to growth in some regions, but there is uncertainty over the scale and timing of future productivity gains and the ability of companies to monetise these. While AI-related equity valuations are underpinned by forecasts of strong long-term earnings growth, those forecasts are highly uncertain and depend on the successful build-out of infrastructure, continued access to financing, and the pace at which AI is adopted across the economy. A reassessment of these prospects could trigger a fall in equity prices that might be amplified by high concentration, correlated momentum-driven positions that can exacerbate volatility as markets fall, and increased leverage.
• Considerations around the future earnings potential for AI-related companies will also be relevant to the sustainability of these companies’ debt, where the increasing complexity and opacity in debt structures used could increase risks to financial stability. In this FSR, the FPC sets out its forward-looking framework to monitor the macrofinancial implications of AI developments for financial stability.
• Risky credit markets, including private credit, remain vulnerable to a tightening in financing conditions. Investor sentiment in parts of private and riskier credit markets had already weakened ahead of the Middle East conflict, reflecting growing concerns around asset quality, valuations and liquidity. Redemption requests have been elevated in several retail funds, with some limiting redemptions, underlining both liquidity mismatch and valuation concerns.
• Higher interest rates could increase debt servicing pressures for leveraged borrowers and lower growth could decrease asset quality, further increasing pressure on, for example, private credit funds.
• Vulnerabilities in these markets remain, including high leverage, complexity and opacity. The Bank’s second system-wide exploratory scenario (SWES) exercise is examining how stress could propagate in these markets.
• Global sovereign bond market issuance is at historically high levels, with higher proportions issued at shorter maturities. Debt-to-GDP ratios are continuing to trend upwards globally, which raises risks of increased volatility and reduces the capacity of governments to respond to future shocks. Current market expectations are that increases in growth driven by AI developments will support debt sustainability. Any negative change in expectations could have wider consequences for sovereign debt markets. In addition, many of these markets are characterised by a relatively high use of leverage by a small number of hedge funds pursuing similar trading strategies across jurisdictions. The high use of leverage also increases the risk of a disorderly unwind of positions that could cause a jump to illiquidity in core markets. Should investor sentiment on the sovereign debt outlook deteriorate abruptly, this could affect the UK market directly, or indirectly via spillovers from other sovereign debt markets, with potential implications for domestic financing conditions.
• During the most significant period of volatility following the onset of the Middle East conflict, moves in gilt yields were amplified by hedge fund deleveraging. Markets absorbed high volatility and volumes, underpinned by functioning repo markets. This partly reflects resilience built in recent years, particularly in Liability Driven Investment funds and Money Market Funds (MMFs). This underlines the importance of locking in this resilience through reforms such as those recently announced for MMFs across the UK and Europe, and through the changes the Bank is exploring to support gilt repo market resilience following its first SWES.
• The FPC is particularly concerned that a number of these vulnerabilities could crystallise simultaneously. The FPC emphasises the importance of market participants ensuring that they have a clear and comprehensive understanding of the size, concentration, and interconnections of their exposures. Strengthening this understanding will support effective risk management and help sustain the resilience of the markets in which they operate.

UK household, corporate and banking system resilience

• UK household and corporate aggregate indebtedness remains low relative to historical averages and debt vulnerability metrics remain around their long-run averages, providing resilience against shocks. Debt-servicing burdens are projected to rise moderately due to higher energy prices and borrowing costs, but it would take a severe shock to return aggregate debt servicing pressures to near previous peaks. Some vulnerable, low-income households, and smaller and more leveraged corporates, particularly those which are financed by riskier credit markets like private credit and leveraged loans, remain more exposed. Overall, the FPC judges that in aggregate households and corporates remain resilient, even in a challenging external environment.
• The UK banking system remains appropriately capitalised with high levels of liquidity. Past stress test results demonstrate that the UK banking system is able to absorb a severe energy price shock and associated economic downturn while continuing to support lending to the real economy. Consistent with this, banks have continued to supply credit to households and businesses despite the deterioration in the macroeconomic outlook and heightened uncertainty, with no sign of lending being restricted to protect capital positions.
• Banks’ interlinkages with non-bank financial institutions, including those that support credit provision to the real economy, create channels through which risks can be transmitted back to banks. Effective monitoring and risk management of these interconnections is important in the current environment. An example is significant risk transfers (SRTs) which allow banks to transfer credit risk to third parties. The Prudential Regulation Authority (PRA) continues to engage with banks on appropriate use of SRTs and is testing thoroughly any proposed structures that appear to be more complex or less robust.
• The FPC has maintained the UK countercyclical capital buffer (CCyB) rate at its neutral setting of 2%. Maintaining a neutral setting of the UK CCyB in the region of 2% would help to ensure that banks continue to have capacity to absorb unexpected future shocks without restricting lending in a counterproductive way.

Developments in frontier AI

• Rapid progress in frontier AI capabilities since the December FSR present a significant increase in the risks to financial stability from cyber and operational vulnerabilities. Frontier AI models are increasingly capable of identifying and exploiting software vulnerabilities at greater scale and over multiple stages. Whilst frontier AI will offer opportunities to improve cyber defence, it will also increase the sophistication and impact of cyber-attacks on firms, including financial institutions and market infrastructure. Operational risks are also likely to increase as frontier AI accelerates vulnerability discovery and exploitation, requiring firms to identify, patch and mitigate vulnerabilities more quickly and frequently, increasing the risk of disruption if change is not managed effectively.
• These developments reinforce the importance of firms acting on the May 2026 joint statement from the Bank, Financial Conduct Authority (FCA) and HM Treasury on frontier models, and on existing cyber and operational resilience frameworks. UK authorities are providing further support through supervision and sector engagement, including through the Cross Market Operational Resilience Group, and through the Bank and PRA’s upcoming consultation on Cyber, and Information and Communication Technology risk management. These developments also require firms and authorities to revisit whether current deep cyber recovery capabilities, coordination arrangements and the resilience of key technology providers remain sufficient. Deepening coordination across authorities and key vendors domestically and internationally will be critical. These developments also underline the importance of operationalising the UK’s Critical Third Party regime.

Bank capital review

• The FPC, working with the PRA, is modernising the capital framework (Box D). The proposed changes will help ensure the framework is simpler, more effective, more proportionate and better calibrated to the risks in today’s financial system, while ensuring that the system remains resilient so that it can support the economy when it needs it most. The reforms will address unintended consequences in the leverage framework and strengthen the releasability and usability of buffers. In doing so, they will make it easier for banks to use capital to provide credit to households and businesses and support the functioning of core sterling markets, while maintaining overall consistency with international standards.
• The Committee will work with the PRA and international authorities to pursue broad reform of the bank capital buffer framework, moving towards the goal of a single buffer that is releasable in stress. In the near term, the FPC and PRA are taking steps towards this vision domestically. The FPC welcomes the PRA’s intention to release the other systemically important institution (O SII) buffer that applies to certain domestic systemically important firms in the event of systemic stress, engaging with the FPC when doing so.
• The FPC and PRA intend to consult on a package of measures to make the leverage ratio framework more proportionate and more effective by being better targeted. This package includes removing the countercyclical leverage buffer from banks’ leverage requirements; moving the calibration of the additional leverage ratio buffer for firms with systemic buffers into line with international standards; and making a greater share of leverage requirements and buffers releasable.

1: Developments in the global risk outlook and financial markets

1.1: Developments in the global risk environment

The conflict in the Middle East caused a material shock to global commodity supplies, and a sharp rise in global energy prices.

The onset of the Middle East conflict in late February disrupted global commodity flows as a result of the closure of the Strait of Hormuz and attacks that caused damage to energy infrastructure. Consequently, energy prices as well as those of other commodities, such as fertilisers, rose sharply, representing a substantial negative supply shock to the global economy.

Oil prices were volatile and fluctuated around the $100 per barrel level for several months (Chart 1.1). By comparison, oil prices had averaged around $68 per barrel over the previous decade. Gas prices also increased as a result of the conflict. However, oil and gas prices remained below the peaks reached in 2022 following the Russian invasion of Ukraine.

The United States and Iran signed a Memorandum of Understanding (MoU) in June, and energy prices subsequently declined to just above pre-conflict levels. The agreement provides for a subsequent 30-day period for the restoration of traffic in the Strait and a 60-day period for negotiations to reach a final deal. As such, while near-term risks around the conflict have reduced, material uncertainty persists in respect of the wider normalisation of the region, including commodity flows.

In the event of re-escalation, there would be a risk of further volatility and sharp upward repricing in energy markets. In that event, there may be limited scope for previous partial mitigants to continue moderating upward price pressures. These mitigants included the coordinated release of strategic oil reserves by the member countries of the International Energy Agency, and there is evidence of reduced demand for energy in response to higher prices, particularly in Asia. Nonetheless, the MoU has made it less likely that energy markets will cross the supply ‘tipping point’ that some experts had warned was approaching over the summer months.

Chart 1.1: International energy prices have been elevated as a result of the Middle East conflict

Brent crude oil (US$ per barrel) and European natural gas (€ per MWh)

• Sources: Bloomberg Finance L.P. and Intercontinental Exchange (ICE).

The conflict has implications for the global economy, with likely impacts on inflation, growth and interest rates globally. These impacts are still playing out.

Elevated global energy prices constitute a material macroeconomic supply shock to the global economy, generating inflationary pressures, which are likely to weigh on growth and put upward pressure on interest rates globally. While the higher costs of some energy products, such as petrol, have already fed through to prices faced by households and businesses, increased costs for many non-energy products, such as food and manufactured goods, take longer to be passed on.

Some Asian economies are particularly dependent on energy imports from the Gulf. The UK and the euro area are less directly reliant on supplies from the Gulf, but are net importers of energy and so are exposed to higher international prices. The United States, as a net energy exporter, is relatively less exposed – although there can still be impacts on US inflation and growth. The impact of higher energy prices on inflation globally will depend on the combination of both supply and demand factors.

Leaning against the adverse impacts of the conflict on the global economic outlook, there have been some favourable tailwinds to growth. In particular, AI-related investments have supported growth across a number of jurisdictions, especially in the United States and some East Asian economies, and the US effective statutory tariff rate has decreased since 2025.

While near-term risks around the Middle East conflict have reduced with the signing of the MoU, the global risk environment as a whole remains highly unpredictable, and the possibility of further and overlapping shocks persists.

The conflict in the Middle East has occurred against the backdrop of several other major global sources of risk, which remain elevated. These include the Russia-Ukraine conflict, and continuing global trade tensions and associated policy uncertainty. For example, despite declining relative to the peaks seen during 2025, the Trade Policy Uncertainty Index remains well above its historical average.^{footnote [1]}

An unpredictable global environment raises the likelihood of additional large global shocks occurring. Global shocks could interact with one another, increasing their combined impact.

Increased geopolitical tensions also create an environment of heightened risk of cyberattacks – including those enabled by frontier AI – which could coincide with, and amplify, other stresses. Refer to Section 3 for discussion of frontier AI and cyber-risk.

Higher energy prices could create headwinds for households and corporates globally, and interact with existing pockets of risk.

Energy-driven cost increases and higher market interest rates globally could increase pressures on some corporates (Section 5.3 discusses the potential impacts on the UK corporate sector). In particular, there could be risks to heavily indebted corporates, and to smaller, consumer-facing corporates via weaker demand, higher costs and interest rates.

Households globally might also come under pressure as a result of these macroeconomic dynamics, impacting existing debt vulnerabilities. In the United States, the proportion of consumer credit loans more than 90 days in arrears has increased markedly over the past three years. In particular, the share of credit card accounts in arrears has risen particularly quickly to around 13% in 2026 Q1. However, the share of those accounts in arrears but not yet written off – a more relevant measure of potential future losses for lenders – levelled off at around 4% at the end of 2025, near its long-run average level.

1.2: Global interest rates and sovereign debt markets

The Middle East conflict led to a sharp increase – and subsequent period of elevated volatility – in government bond yields in advanced economies.

The energy-driven supply shock to the global economy has generated inflationary pressures, and this drove market interest rates higher across advanced economies, although they have declined somewhat since the signing of the US-Iran MoU (Chart 1.2). These moves reflected, in part, higher policy rate expectations.

After the onset of the conflict, gilt and euro area government bond yields increased by more than US Treasury bond yields, reflecting concerns over those economies’ exposures to imported energy costs, though that trend has since reversed. Despite high levels of volatility, UK gilt and gilt repo markets have remained resilient, with liquidity conditions commensurate with market conditions (Section 4.1).

Japanese 10-year government bond yields have risen to levels last seen in 1997. This has taken place in the context of the Bank of Japan’s monetary policy normalisation, as well as the broader global repricing of long-term bond yields.

Chart 1.2: Some government bond yields have reached their highest level for over a decade

UK, US, German and Japanese 10-year government bond yields

• Source: Bloomberg Finance L.P.

Public debt-to-GDP ratios are expected to continue on an upward trajectory in many major economies.

The upward trajectory in debt-to-GDP ratios is a long-run trend in many advanced economies, with the most recent IMF forecasts indicating that the global average will continue to climb in coming years, driven primarily by the United States and China (Chart 1.3).

An important determinant of debt-to-GDP is the growth-adjusted interest rate (the difference between nominal interest rates on government debt and expected nominal GDP growth rates). While the stock of debt relative to GDP is slow-moving, the Middle East conflict has the potential to worsen the ratio by pushing up market interest rates and weighing on growth.

Another factor behind the sustained rise in debt-to-GDP ratios is the range of spending pressures on governments globally, including in respect of defence, demographics, and climate change. The conflict and resulting energy shock have the potential to exacerbate these pressures, for example by generating demands for government fiscal support for households and businesses, or by increasing defence spending pressures.

Chart 1.3: Debt-to-GDP ratios globally continue to trend upwards

Sovereign debt as a percentage of GDP ^(a)

• Sources: International Monetary Fund (IMF) and Workspace from LSEG.
• (a) Use of IMF Content and Data is subject to the IMF’s terms of Copyright and Usage.

There are various channels through which pressures on sovereign debt globally could impact UK financial stability, including by triggering vulnerabilities in market-based finance.

Investor concerns about the sustainability of government debt can lead to higher yields and increased volatility. In addition to directly tightening financial conditions for the real economy, this dynamic has the potential to trigger vulnerabilities in leveraged parts of market-based finance, which would amplify stress.

The IMF estimates that relatively more price-sensitive investors (which they define as investment funds, foreign investors and households) now hold at least half of US, Japanese, euro area and UK sovereign debt.^{footnote [2]} In particular, hedge funds have increased their participation in government debt markets.

Globally, sovereign debt issuance is at historical highs and there has been a trend of governments issuing debt at shorter average maturities, partially in response to steepening term premia. This increases rollover risks and in an inflationary environment can lead to short-term debt being refinanced at higher interest rates. For UK sovereign debt, while the average term to maturity is among the highest of the OECD countries,^{footnote [3]} there has also been a trend towards issuing shorter-dated bonds. This has been driven in part by a reduction in demand from defined benefit pension funds for long-dated debt. Section 4 explores this trend further, specifically how the ‘preferred habitat’ of gilt investors has shifted notably shorter over recent years – including within the hedge fund and asset manager sectors.

More generally, increased indebtedness and higher debt-servicing costs could reduce the capacity of governments to respond to future shocks, including those stemming from the heightened global risk environment, leaving households and businesses more exposed to their impact.

1.3: Global risky asset markets

Despite the challenging global risk environment, by some metrics equity valuations continue to be stretched relative to earnings…

While the immediate onset of the conflict in the Middle East led to a fall in equity prices globally, that drop quickly reversed. And since the signing of the US-Iran MoU in June, market participants appear cautiously optimistic on it leading to a sustained de-escalation.

Certain equity indices have been largely driven by other factors, including continuing favourable earnings news in Q1, and with limited weight given to the impact of the conflict on US growth, even before the MoU was signed. Many stock market indices are now higher than at the time of the December FSR, for example the S&P 500 and the FTSE 100 have both increased by around 10% over that period.

AI-focussed companies have been a particular source of growth in equity prices, supported by positive earnings news further boosting expectations. Companies focussed on AI hardware have been especially high performing, with the Philadelphia Semiconductor Index up 97% since the start of the year.^{footnote [4]} This increase in AI-related valuations has led to increased concentration in US equity markets. Several share offerings for large AI companies this year could materially increase this concentration. It could also test the market’s appetite for further exposure to this sector. Section 2.2 discusses AI-related equity valuations in more detail.

Consistent with sustained price increases, by some metrics equity valuations have become further stretched by historical standards. For example, the excess cyclically-adjusted price-to-earnings (CAPE) yield for the S&P 500 has continued to edge downwards towards levels not seen since the dot-com bubble (Chart 1.4). Even when excluding the top 30 AI-related stocks,^{footnote [5]} its level is around the lowest seen since 2007, indicating that this stretch in valuations is more broad-based than AI-related optimism. While excess CAPE is backward-looking with respect to earnings, the excess yield for 3-year forward earnings expectations for the S&P 500 is also around historically compressed levels – despite positive news in Q1 leading to increased earnings expectations.

Chart 1.4: Equity valuations continue to be stretched relative to historical levels

S&P 500, Excess cyclically adjusted price to earnings (CAPE) yield ^(a)

• Sources: Bloomberg Finance L.P., S&P Dow Jones Indices, a subsidiary of S&P Global and Bank calculations.
• (a) ‘Top 30 AI’ are defined as those stocks which appear in JP Morgan’s JPAMAIDE Equity Basket, which is comprised of 30 S&P 500 stocks that are particularly impacted by AI.

…and the growing use of leverage in equity markets further heightens the risk of sharp corrections.

The risk of a sharp correction in equity markets remains high. Equity prices could rapidly adjust in the event that investor expectations shift, for example because of a change in the macroeconomic outlook. A significant fall in stock prices would entail losses for individual investors and could trigger a general pullback from risky assets. This would make it harder for corporates to finance their activities.

The scope for sudden adjustments is increased by the use of leverage in equity markets. This is because leveraged positions can be forced to unwind quickly during market stress events, thus amplifying the volatility.

In wholesale markets, there is evidence of increased hedge fund leverage. Hedge funds’ equity prime brokerage balances are at record levels, with supervisory intelligence indicating that these balances globally have increased by around 40% over the past year. In addition, market intelligence suggests that hedge fund equity positions have become more concentrated in particular sectors, such as semiconductors, in recent months, coinciding with the ongoing AI-linked stock price momentum. Higher and more concentrated leverage can pose risks to hedge funds, and create financial stability risks, via potential losses to prime brokers and cross-market interconnections if funds making losses on equity positions deleverage across other markets – including sovereign debt – in response (Section 4 and Box B).

There has also been a significant increase in assets under management of levered exchange-traded funds (ETFs) in recent months, particularly in those holding AI-related stocks. Market intelligence indicates that retail investors are the main holders of levered ETFs. This increase has occurred in the context of retail investor inflows more broadly, including into unlevered ETFs, which may have added momentum to the rise in equity prices.

While total levered equity ETF holdings (at around $200 billion in the United States) remain small so far relative to total unlevered equity ETFs holdings (at around $15 trillion in the Untied States), their use of leverage means they can amplify price moves. For example, their daily rebalancing mechanism causes ETFs to buy more stocks when prices rise, and to sell more – thereby exacerbating volatility – when they fall. These dynamics echo those which were prevalent ahead of the summer 2024 equity market sell-off (discussed in the October 2024 FPC Record), where volatility was substantially amplified. It is important that lessons from this episode are learned, notably so that effective risk management by market participants can prevent amplification and spillovers.

Corporate credit spreads globally are also compressed by historical standards.

Corporate credit spreads are at similar levels as at the December FSR, in the lower half of their historical distribution (Chart 1.5).

Issuance of sterling- and euro-denominated investment grade bonds since the start of the year has been broadly in line with previous years. For US dollar debt, issuance has been significantly higher than average, driven by AI hyperscalers. The influx of issuance from AI-related companies does not – as of yet – appear to be negatively affecting the terms and availability of public investment grade credit (or other credit markets) more broadly, but such a ‘crowding out’ dynamic could have implications for the supply of credit to the real economy (Section 2.4).

Chart 1.5: Credit spreads generally remain at compressed levels

Current levels of selected risk premia as a percentile of their historical distribution, compared to levels as at the March 2026 FPC meeting and the December 2025 FSR^(a)

• Sources: Bloomberg Finance L.P., S&P Dow Jones Indices, a subsidiary of S&P Global, Datastream from LSEG, ICE BofAML, PitchBook a Morningstar company, Inc. and Bank calculations.
• (a) Risk premia data are a percentile of a five-day rolling average (except for leveraged-loan (LL) spreads, which are a percentile of a monthly average). Percentiles are calculated from 1998 for investment-grade spreads and high-yield bond spreads, from 2008 for LL spreads and from 2006 for excess CAPE yields. Data updated to 25 June 2026 apart from LL which is to 19 June 2026. All data are daily except for LL spreads which are weekly. Investment-grade spreads are adjusted for changes in credit quality and duration.

Energy-driven cost increases and higher interest rates globally can be expected to increase pressures on some corporate borrowers. UK credit markets are highly exposed to global developments and investor sentiment, especially in riskier lending segments (such as high yield bonds and leveraged lending). A loss of investor confidence could be triggered by rising defaults or refinancing stress globally or domestically, leading to losses, spillovers across markets, and a tightening in financing conditions for UK corporates.

Recent developments in private credit markets are discussed in Section 4.2.

2: The macrofinancial implications of AI

2.1: The relationship between AI and financial stability

AI has the potential to deliver significant productivity gains, driving broader structural change across the economy and financial system. It is already beginning to reshape patterns of investment and economic activity.

AI may prove to be one of the most transformative technologies of our time, with the potential to deliver substantial efficiency gains, raise productivity, and transform financial services as well as the broader economy. However, the transition to widespread adoption – including the investment required to build the supporting infrastructure – also comes with financial stability risks that must be understood and managed.

Progress in AI models has been faster than many experts expected. The FPC has recently considered the financial stability implications of rapid advances in the cutting-edge – ‘frontier’ – AI models (refer to Frontier AI: implications for financial stability). Over the past year, the FPC has also considered a range of developments in the AI ecosystem and their implications for the financial system, including equity market valuations and the increasing use of AI in the financial sector (refer to the December 2025 FSR and the FPC’s Q1 Record).

This section introduces our assessment of the near and medium-term channels through which the ‘AI transition’ could affect UK financial stability, focusing on the effects of AI-related investment and broader adoption on financial markets, including sovereign bond markets, and financing conditions for the real economy (the ‘macrofinancial’ channels). It does not consider wider macroeconomic impacts, including via the labour market, which will be considered in further work by the Bank. It also sets out how the FPC plans to develop this work further, including by monitoring these channels over time, assessing how they are contributing to structural change in the financial system, and embedding the analysis in the Bank’s mainstream monitoring of risks to and assessment of financial system resilience.

The AI transition could have macrofinancial implications for financial stability through two interdependent channels: the financing of the required infrastructure, and the pace and extent of adoption of the technology.

Building the infrastructure needed to support both the development of AI and its widespread adoption requires significant investment. As noted in the December 2025 FSR, AI-focused companies reached an inflection point in 2025, when required investment exceeded their capacity to finance it from internal cashflows, leading them to turn to external finance. This trend accelerated substantially in the first half of 2026, with companies accessing a broad range of financing channels, including private markets, public debt markets, and bank lending (Section 2.3).

At the same time, shifts in expectations around the pace of AI adoption – which in turn has consequences for anticipated productivity effects and companies’ earnings – could increase volatility in asset prices and funding conditions, before those effects are realised in the wider economy. Considerations around future earnings potential would also be important in determining the sustainability of these companies’ debt, where the increasing complexity and opacity in debt structures used could increase risks to financial stability.

And, while developments in AI are already providing a tailwind to growth in some regions, there is uncertainty over the scale and timing of future productivity gains and the ability of companies to monetise these. Any future revaluation of the impact of AI on growth could also, by extension, impact sovereign debt markets, particularly in relation to highly indebted countries (Section 2.4).

Many factors will determine the pace of AI adoption and the impact of AI-related companies’ investment demand on financial stability.

Four broad factors include:

1. The capacity of AI companies to build the infrastructure that supports the development of capabilities and their adoption.
2. The impact of AI on how finance flows around the financial system and the economy, and the ability and willingness of capital markets to finance AI investment.
3. The capabilities of AI, including to reliably execute tasks at speed and scale.
4. Decisions by businesses and individuals on when and how to adopt AI.

These factors are interdependent, and their interaction will shape both the scale of infrastructure investment and the pace and extent of adoption across the economy:

• Risks through the adoption channel will depend on decisions by businesses and households to adopt AI, which in turn will depend in large part on whether AI capabilities are sufficient to support its adoption, the cost of using AI, and whether AI companies have the capacity to deliver the infrastructure required to support widespread deployment.
• Risks through the infrastructure financing channel will depend on AI companies’ capacity to build that infrastructure and their ability to access and service the necessary finance.
• Expectations around the scale and speed of AI adoption, and around the productivity and returns it may generate, will also influence investment decisions, including what infrastructure is built, when, and the cost of financing it.

By monitoring the trends underpinning these factors – including technological developments, survey data on adoption, labour market developments and financial data – it is possible, despite considerable uncertainty, to build a clearer picture of the likelihood of different outcomes for the paths of AI investment and adoption, and their implications for the financial system. This will form the forward-looking framework the FPC will use to monitor the macrofinancial implications of AI developments for financial stability.

2.2: Near-term risks to UK financial stability

The valuations of AI companies have grown faster than relevant aggregate equity indices in 2026 Q2. High and growing concentration risk in global equity markets increases the potential impact of a revaluation.

Equity prices have increased, especially for AI-related stocks (Section 1.3). This has been driven, in part, by a narrow set of AI-related companies, increasing market concentration in some global indices. This dynamic is most pronounced in geographies, such as the United States (US) and East Asia, where the largest businesses in the AI supply chain (eg Nvidia, Alphabet and Samsung) are publicly listed. The S&P 500 accounts for around half of global equity capitalisation. Within that index, AI companies now account for around half, up from around a quarter in 2022. This share could increase further if a number of companies, including OpenAI and Anthropic, access public equity markets through initial public offerings (IPOs) as expected (Section 2.3). SpaceX, which is in part an AI company, recently achieved the largest IPO in history. In other jurisdictions, AI companies also account for a high share of major equity indices – refer to Chart 2.1.

While UK equity indices continue to be considerably less exposed to AI companies directly, a correction in the valuation of these companies could nevertheless have a significant effect on global equity markets more generally. This would affect the UK macroeconomy and financial system via spillovers, largely due to the adverse impact a scenario like this would have on financing conditions – effects explored in the scenario analysis presented below.

Chart 2.1: AI-related companies’ share of equity indices has grown materially in some jurisdictions since 2022

AI-related companies’ market capitalisation as a proportion of the relevant indices ^(a)

• Sources: LSEG Workspace and Bank calculations.
• (a) The chart shows the share of total index market capitalisation accounted for by AI-related companies across major equity indices: the United States (S&P 500), Taiwan (TAIEX), South Korea (KOSPI 200), Japan (TOPIX 100) and the United Kingdom (FTSE 100). The numerator captures the combined market capitalisation of companies classified as AI-related, based on a predefined basket constructed from the holdings of AI-focused investment funds. The denominator is the total market capitalisation of each index. Values are reported as at 31 December 2022, 31 December 2025 and 25 June 2026. In all periods, market capitalisation is calculated using index constituents as at 25 June 2026.

While AI-related equity valuations are underpinned by forecasts of strong long-term earnings growth, those forecasts are highly uncertain and depend on the successful build out of infrastructure, continued access to financing, and the pace at which AI is adopted across the economy.

The share prices of many AI companies are underpinned by forecasts of strong long-term earnings growth, something that traditional valuation metrics do not typically capture. However, embedded in these earnings expectations are a number of important assumptions which, if incorrect, could lead to a revaluation of investor sentiment around these companies’ earnings prospects, and accordingly to a correction in equity valuations.

Among these factors, which the FPC has previously highlighted, are:

• Risks to capacity: The projections presume that AI infrastructure will be built out successfully. AI capacity development relies on a considerable amount of hardware (eg logic and memory chips, networking, cooling systems), the raw inputs to produce this hardware (eg silicon wafers, lithography machines and the chemicals and commodities used in the manufacture of AI hardware) and the availability of utilities (eg electricity and water to operate the data centre facilities which house the hardware). The availability of these inputs may constrain the scale and pace of build-out implied by current expectations for revenues and earnings. In addition, significant parts of this supply chain are geographically concentrated (for example, the semiconductor supply chain is concentrated in East Asia), and in many cases, production is highly specialised and dominated by a small number of companies. Events that disrupt the supply chain could materially disrupt the development of AI capacity.
• Risks around capabilities and adoption: Long-term earnings projections assume strong AI adoption. Adoption relies on AI being capable (eg able to execute tasks with a sufficiently high degree of accuracy and at scale) and cost-effective. More complex workflows and operations involving longer chains of reasoning – such as some agentic applications – may require larger volumes of input and output tokens, increasing usage-based costs and potentially making them prohibitively expensive for widespread use. Even where capabilities are sufficient, businesses may take time to adopt AI at scale while they build confidence that it will deliver reliable productivity gains, meaning technical capability may emerge before revenues are realised.
• Risks around finance and capital flows: Building AI capacity depends on continued access to external finance. However, there are also some fragilities in the financing arrangements around AI investment, which could amplify the impact of a shock. For example, AI companies’ revenue forecasts may reflect ‘circular financing arrangements’. One example of this is when technology companies invest in AI companies which in turn purchase those technology companies’ products. This creates self-reinforcing capital loops. There is also growing use of off-balance sheet financing structures as a source of funding for AI investment which may represent an additional source of fragility in stress (Section 2.3). The AI supply chain is also dependent on the investment decisions of the five large AI-focused technology companies – the ‘AI hyperscalers’. These companies’ spending and commitments provide revenues for upstream semiconductor, hardware and data centre providers, while their built capacity supports revenue generation for AI application providers (including labs like Anthropic and OpenAI). Taken together, these factors increase the likelihood that a negative shock could result in a large and correlated negative earnings re-evaluation across a wider set of AI-related businesses.

Bank staff have assessed the impact on the UK economy of a potential global correction in AI valuations, using a scenario in which such a correction occurs in the near term but with long-term implications.

The hypothetical scenario considers an AI-related equity price correction, driven by a repricing of AI productivity and profitability, that starts in the US. This implies a sharp equity market correction (a 45% fall in the US equity market over six quarters), a repricing of corporate credit spreads (increasing by 350 basis points) and a persistent dollar depreciation. The shocks spill over into lower UK output (a 2.2 percentage point fall in GDP), largely driven by financial channels, with equity market effects accounting for about 36% of the response and credit spreads about 50%. While a shock of this scale would undoubtedly have an impact on households and corporates, the macroeconomic impacts are encompassed by both the 2025 Bank Capital Stress Test and the 2024 Desk-Based Stress Test scenarios, to which UK banks were resilient.

An equity shock in isolation would be unlikely to present a direct risk to UK financial stability. Nevertheless, a reassessment of AI-related companies’ prospects could trigger a fall in equity prices that might be amplified by high concentration, correlated momentum-driven positions that can exacerbate volatility as markets fall, and increased use of leverage (Section 1.3).

In addition, as discussed in Section 2.4, such a scenario could have broader implications for sovereign debt dynamics. In the Bank’s analysis, while debt-to-GDP ratios rose in the hypothetical scenario, the US Treasury and gilt markets continued to function well; had these markets come under pressure, the financial stability consequences could have been more significant.

2.3: A growing role for debt in financing AI infrastructure

The risks from the infrastructure investment channel are growing rapidly, as the AI ecosystem is accelerating its use of external finance.

The pace of investment is unprecedented historically. AI companies are increasingly turning to the financial system for external financing – particularly debt financing – to support investment in infrastructure, and this has accelerated in the first half of 2026.

As this section sets out, the financial stability risks posed by this increasing demand for debt financing have to date been contained by modest stocks of outstanding debt. This is changing at pace. If the scale of AI debt financing grows as expected over the coming years, an adverse shock to AI companies that results in losses or affects their ability to service debt could more materially affect global financing conditions. In turn, this could affect the provision of finance to the UK real economy.

Market participants’ projections of the size of future AI investment have repeatedly been revised upwards. These expectations of more significant investment needs are likely to lead to even greater demands for external finance.

For example, at the time of publication of the December 2025 FSR, Bloomberg consensus estimates were for AI hyperscalers’ capital expenditure for the year 2028 to be less than $600 billion. The same statistic is now over $1 trillion.

Some of this additional investment need will be met via equity issuance. Alphabet raised just under $85 billion through equity issuance in 2026 Q2, and more AI hyperscalers are expected to follow this year. Additionally, as noted earlier, a number of companies in the AI ecosystem have accessed, or are expected to access, public equity markets through IPOs. Equity financing of AI investment could be positive for financial stability: equity capital can preserve debt capacity, strengthen credit confidence and diversify funding sources, supporting continued investment. However, given the higher-than-expected financing needs of these companies, it remains likely that their need for debt financing will continue to grow alongside equity issuance. Chart 2.2 summarises Morgan Stanley’s latest estimates of how global data centre investment will be financed between 2026 and 2028.

Chart 2.2: More than half of the external financing need for data centres from 2026 to 2028 could be funded via debt

Morgan Stanley estimates of funding sources for data centre capital expenditure (2026–28) ^(a)

• Source: Morgan Stanley Research.
• (a) The chart shows Morgan Stanley forecasts of global data centre capital expenditure from 2026–28, broken down by expected funding source. This includes hyperscalers and non-hyperscalers. ABS are asset-backed securities and CMBS are commercial-mortgage-backed securities.

AI-related companies’ use of credit markets has also accelerated rapidly, including in public markets, private credit, leveraged and structured finance, and is set to increase further as financing needs continue to expand.

At the start of 2026, the stock of outstanding debt from AI companies was relatively modest, which the FPC judges has helped to contain the immediate risk to financial stability. However, the fast growth of activity in credit markets in the first half of 2026 is resulting in potential risks building rapidly.

Across all currencies, hyperscaler year-to-date investment-grade debt issuance is broadly comparable in scale to UK gilt issuance over the same period.

The largest footprint is currently in US dollar investment-grade (IG) credit. The five AI hyperscalers accounted for just 3% of the stock of outstanding US IG debt at the end of 2025, but as of early May accounted for over 15% of year-to-date issuance.^{footnote [6]}

However, the scale of this issuance is material, as is its expected impact on the composition of the market. For example, analysis by Barclays projects $240 billion of the AI hyperscalers’ investment needs in 2026 will be financed through IG credit issuance; by comparison, they note the ‘big six’ US banks^{footnote [7]} typically issue $150–170 billion of senior debt per year. Analysis by JP Morgan suggests, by the end of this year, technology is expected to overtake US banks as the largest sector in US dollar IG debt, by share of outstanding debt.

So far, this issuance has been easily absorbed by the market, reflecting high appetite from investors for assets from these companies – the majority of which have low debt ratios and high credit ratings (AA- or above). More broadly, credit spreads in US dollar IG credit markets remain compressed.

Chart 2.3: AI hyperscalers’ bond issuance in the first half of 2026 has already exceeded 2025 issuance

Bond issuance by the AI hyperscalers, by year and currency ^(a)

• Source: LSEG Workspace.
• (a) The chart shows IG bond issuance by hyperscalers (Meta, Alphabet, Amazon, Microsoft and Oracle), aggregated by year and broken down by currency of issuance. Currency conversions are based on the end‑of‑day spot rate as at 25 June 2026. All figures are denominated in US dollars.

AI companies’ use of leveraged finance, structured finance and private credit is also growing rapidly, though their share of outstanding debt remained low entering 2026. AI issuers have accounted for 41% of non-refinance-related US high-yield (HY) issuance this year so far, despite only accounting for 1% of the JP Morgan HY bond index at the end of 2025. The OECD estimated that the share of private credit financing AI investment increased from 9% in 2024 to 34% in 2025.

The growth and projected scale of debt issuance to finance AI investment introduces new risks – particularly around the sustainability of medium-term debt servicing.

Considerations around future earnings potential for AI-related companies will also be relevant to the sustainability of these companies’ debt, where the increasing complexity and opacity in debt structures used could increase risks to financial stability.

The declining free cash flows of the AI hyperscalers increases their dependence on accommodative future refinancing conditions in capital markets to sustain their infrastructure investment over time (Chart 2.4). It is also leading to their growing use of off-balance sheet financing arrangements. This off-balance sheet financing trend is resulting in capital being sought for AI investment from across the credit ecosystem – including through securitised data centre and other asset-backed structures, special purpose vehicles, and other bespoke financing arrangements – broadening the AI debt footprint across leveraged, structured, and private credit. It can result in higher asset-level leverage and increases the complexity of identifying where risk ultimately sits. The riskiness of this debt depends on the underwriting terms, in particular the quality of the leases and guarantees which back debt holders’ claims.

Chart 2.4: Hyperscalers’ free cash flows are declining and are projected to continue to do so

Unlevered free cash flow (bars, left axis) and free cash flow to revenue (line, right axis) for the AI hyperscalers ^(a)

• Sources: LSEG Workspace and Bank calculations.
• (a) This chart shows aggregate unlevered free cash flow and free cash flow as a share of revenue for the five AI hyperscalers (Meta, Alphabet, Amazon, Microsoft and Oracle). Data are presented on a calendar-quarter basis. Observations are based on reported data where available, with consensus forecasts used for periods where company results had not yet been reported.

There is also a potential source of fragility where long-term debt is issued to finance investment in AI assets which may have shorter lifecycles.

To date, debt issued to finance AI investment has tended to have longer (more than ten year) duration, in line with the appetites of investors in the deepest capital markets. The majority of this debt has funded data centre buildings (ie the physical shell of the data centres) and facilities – rather than the servers and AI chips inside them. However, frontier AI data centres can quickly become outdated if they are no longer able to support the latest chips. This could shorten the facilities’ economic lives and reduce collateral values and the cash flows available to service debt.

Funding needs for AI chips are expected to grow considerably. JP Morgan estimates that over $2 trillion of aggregate funding may be needed for the AI chips used to train and run AI models over the next five years. However, chips have short and uncertain lifecycles, which may make it harder to attract enough capital. In particular, issuing debt maturities aligned with the asset life of AI chips would imply shorter tenors than are typically preferred by most credit investors. Additionally, there is mixed evidence on how quickly AI chips depreciate: current shortages and demand for older chips support the argument for longer useful lives, while rapid innovation and efficiency gains could shorten them.

The scale of this maturity mismatch risk across the debt financing of data centre facilities and AI chips, and how that risk is allocated across lenders, borrowers or tenants, will depend on the terms of the debt agreements.

2.4: Interaction of macrofinancial dynamics with growth

The UK is expected to see a significant buildout of AI data centre infrastructure in the coming years.

The UK has the largest data centre pipeline in Europe. There will be a need for significant investment to deliver this roll-out. Consistent with its secondary objective of supporting the government’s economic policy, including growth, the FPC is seeking to identify and support the removal of material frictions to the investment of long-term capital into improvements in the productivity of the economy. The FPC will report on progress in the November 2026 FSR.

If these AI data centre infrastructure projects are delivered, they could support UK growth, including through their aggregate effects on investment. In turn, this would yield benefits for financial stability. Market intelligence indicates that capital is ready to be deployed in support of such investments in the UK.

AI is shaping expectations for growth. Investment and trade dynamics are supporting growth in some regions already, and potential productivity improvements could support growth more broadly into the medium term.

The surge in AI-related capital expenditure is supporting investment and export-led growth in geographies with links to the AI supply chain, including in the US and East Asia respectively. Looking ahead, AI also has the potential to raise productivity across a range of sectors, and, in turn, support long-term economic growth. However, and as set out in Section 2.1, there are risks that could prevent these outcomes arising. For example, significant energy bottlenecks – including from insufficient power grid connections and labour shortages – could limit the buildout of AI data centre capacity. It could also influence the speed and scale of adoption in the wider economy.

As of yet, there is little evidence that AI activity is ‘crowding out’ the ability of other businesses or governments to access funding markets.

Currently, terms and availability of debt globally (and public IG credit in particular) do not appear to be responding negatively to the influx of issuance from AI-related companies, potentially reflecting high appetite for these instruments from investors.

But it is possible that other businesses may find their future capacity to issue in these markets ‘crowded out’. This could have broader implications for the provision of credit to businesses, and, therefore, also for growth.

Current market expectations are that increases in growth driven by AI development will support debt sustainability. Any negative change in expectations could have wider consequences for sovereign markets.

Expectations around AI-focused companies’ contributions to growth are a factor influencing expectations in a wider range of financial markets. A higher expected growth rate improves the perceived path of public debt ratios. All else equal, if the adoption of AI results in higher long-run productivity growth, it could materially improve sovereign debt sustainability. On the other hand, large, sustained investment in AI infrastructure can place upward pressure on real interest rates, particularly where the supply of savings is relatively inelastic – raising debt-servicing burdens. The balance between these two dynamics matters particularly given elevated sovereign debt levels globally.

A negative reassessment of the impact of AI could weaken both AI-related earnings and broader growth expectations. Such a shift could prompt a negative repricing in sovereign debt markets, particularly in jurisdictions where the growth outlook is increasingly tied to AI developments. This could affect market participants’ assessment of the sustainability of current debt trajectories, especially in jurisdictions with high and rising sovereign debt-to-GDP ratios. In turn, that could result in spillovers to global financial markets.

2.5: Monitoring and assessing risks across the financial system

The FPC will continue to monitor the potential risks that the macrofinancial implications of AI could pose to the UK financial system, embedding these risks into mainstream assessments of the financial system’s resilience.

As noted earlier, AI has the capacity to create new macrofinancial vulnerabilities (eg increased concentration risk in global markets), to act as a trigger that crystallises existing financial vulnerabilities (eg leverage), and to increase the potential for exogenous shocks to the UK (eg via international spillovers). The AI transition is expected to result in an increasingly broad and deep set of interactions with the financial system. Therefore, in addition to monitoring developments using the approach set out in this chapter, Bank staff will embed analysis of the risks presented by AI into their mainstream assessment of risks to systemic firms, market-based finance, and the provision of vital financial services to the real economy. This will support the FPC in assessing the resilience of the financial system to the potential structural change that AI presents and promoting system-wide resilience.

The risks outlined in this section reinforce the importance of financial firms remaining up to date on new or growing correlations between asset prices and developments in their lending and investment portfolios, that might not otherwise be readily embedded in their risk management and stress testing approaches.

The growing interconnectedness between sectors in the AI supply chain – such as technology and utilities – may require financial firms to reassess their assumptions on asset price correlations in stress and ensure their approach to risk management keeps pace with the evolution of the external environment.

AI advances may also be correlated with negative shocks in other sectors (as the Q1 events in the software sector show), where new model releases or developments trigger investor reassessment of the sustainability of business models in other business sectors. Some financial firms are already, as a matter of course, assessing the potential impact of AI on businesses as part of loan-level decision-making.

Given the different funding sources from which AI companies are drawing, and different levels of transparency in those arrangements, it may be difficult for financial firms to be aware of the full extent of their direct and indirect exposures to the AI ecosystem and AI companies. This could increase the risk that exposures to developments in AI are greater than anticipated, or not fully reflected in firms’ risk management.

3: Frontier AI: implications for financial stability

In 2026 Q1, the FPC assessed the risks to financial stability from the adoption of AI in the financial sector across four channels, building on the framework set out in its April 2025 Financial Stability in Focus, namely: (i) greater use in banks’ and insurers’ core financial decision-making; (ii) greater use in financial markets; (iii) operational risks in relation to AI service providers; and (iv) the changing external cyber threat environment. The FPC noted the considerable growth in AI capabilities in recent years and judged that risks were likely to increase. The FPC also previously noted that, given the pace of technological change, rapid and unforeseen shifts in the implications of AI for financial stability were plausible. Advances since the December 2025 FSR in the most capable general purpose AI models (‘frontier AI‘) are consistent with that assessment and could, in particular, materially increase risks through cyber and operational vulnerabilities. Separately, Section 2 of this FSR provides an assessment of the channels by which AI investment and adoption across the wider economy could impact financial stability in the near and medium term.

3.1: Recent progress in frontier AI

Recent frontier AI models have shown marked gains in capabilities and progress has been faster than many experts expected.

One way to measure that progress is to assess the length of software task that an AI system can complete successfully with limited human input, using the time a human expert would take as the benchmark. Research by Model Evaluation and Threat Research (METR), an independent research organisation, indicates that, in 2020, frontier AI models could only complete tasks that would take a human a few seconds. By 2025, frontier AI models could complete tasks that would take a human several hours.

The pace of progress appears to have accelerated further. In March 2025, METR estimated that the length of software tasks that frontier AI models could complete at a given success rate had been doubling roughly every seven months since 2019 (the so-called ‘doubling time’). By early 2026, METR had revised that estimate to a ‘doubling time’ of around four months since 2023. More recent evidence suggests that frontier models may have improved faster still. METR estimates that the latest frontier models can now complete software tasks with a 50% success rate that would take a human expert 16 hours to complete, which exceeds previous projections.

Agentic AI systems – systems that can plan and carry out multi-step tasks at machine speed using external tools with limited human oversight – are not new. What recent advances in frontier AI demonstrate is a marked increase in (a) how reliably these systems can perform; and (b) the length and complexity of software tasks that can be completed autonomously. In particular, the latest models can sustain work over more steps, make better use of tools such as code editors and test environments, recover more effectively from mistakes, and require less human intervention to reach a useful result.

These systems could help companies develop software and digital services more quickly, improve customer-facing services, support research and decision-making, and reduce operational costs. Because software and digital processes underpin a wide range of business functions across many sectors, gains in planning, tool use and task completion could therefore support productivity and innovation across the economy. Over time, they could also enable new products, more personalised services and more efficient business processes.

While frontier AI will offer opportunities to improve cyber defence, it will also increase the sophistication and impact of cyber-attacks on firms, including financial institutions and market infrastructure.

For the financial sector, this matters because important business services depend heavily on complex digital infrastructure and common third-party technology, so a faster pace of vulnerability discovery and exploitation could have system-wide consequences, including through correlated exposures at shared suppliers, software components and critical infrastructure sectors on which finance depends. Notably, respondents to the Bank and FCA’s 2024 AI Survey, rated cyber security as the highest perceived systemic risk related to the use of AI.^{footnote [8]}

The UK’s AI Security Institute (AISI) has found that the latest frontier models (eg Anthropic’s Mythos Preview and OpenAI’s GPT 5.5) represent a clear step up from earlier models in an area where cyber performance was already improving quickly (Chart 3.1). Two years ago, frontier AI models could complete only short, self-contained cyber tasks. In controlled tests, AISI found that the latest frontier AI models could now carry out multi-stage attacks on vulnerable systems and identify and exploit weaknesses with little human input – tasks that would take skilled professionals days of work. In addition, since agents operate at machine speed, they are likely to be able to complete these longer and more complex tasks much faster than humans.^{footnote [9]} In one of AISI’s advanced reverse-engineering challenges, GPT-5.5 completed the task in 10 minutes and 22 seconds, compared with around 12 hours for a human expert.

Chart 3.1: The cyber capabilities of the latest frontier AI models exceed previous projections

Estimates of the time horizon of cyber tasks that can be completed with 80% reliability by selected frontier AI models (log scale) ^(a)

• Source: AISI.
• (a) 80% reliability cyber time horizon on AISI's narrow cyber task suite (2.5 million token budget). Shaded bands show each model’s bootstrap distribution (1000 hierarchical resamples of tasks/runs). Inner band is the central 50%, outer band is the central 95%. The longest task in AISI’s cyber suite is 12 hours.

Recent frontier models have also successfully completed both of AISI’s harder ‘cyber range’ tests – simulated attacks against small, undefended company networks requiring sustained planning and execution. This is a notable result because (a) these tests require models to sustain multi-step offensive activity over longer sequences of work, rather than simply solve short, isolated tasks; and (b) no previous model had completed either range. AISI has also noted several factors that may mean their evaluations understate the peak capabilities of current frontier models. At the same time, success in these simulated environments does not on its own show that frontier models can already carry out reliable attacks against well-defended real-world targets. That reinforces the importance of firms getting the basics of cyber resilience right now, as discussed further in Section 3.4.

Beyond model evaluations, frontier models have already been used by organisations that have been granted trusted access to identify large numbers of weaknesses in existing software, demonstrating the potential benefits of frontier AI for cyber security. For example:

• As of 22 May 2026, Anthropic had identified 23,019 candidate software vulnerabilities in open-source software using its Claude Mythos Preview model. Of those, they disclosed 1,596 vulnerabilities to software maintainers across 281 open-source projects.
• A frontier model was used to identify long-standing flaws in widely used software, including in a security-focused operating system (OpenBSD) and a heavily tested open-source software library (FFmpeg).
• The US cyber security company Palo Alto Networks said its May 2026 security update identified 75 security issues across 26 publicly-recorded software vulnerabilities, compared with its usual monthly volume of fewer than five such vulnerabilities, and that the majority of those findings came from frontier AI models scanning its code.
• Mozilla reported fixing 423 Firefox bugs in a single month, around 20 times its normal rate.

These examples suggest frontier AI may materially increase the flow of vulnerabilities requiring assessment and remediation by firms. That could itself become a source of systemic risk, even if frontier AI does not immediately enable reliable attacks against well-defended targets. A materially higher volume of identified vulnerabilities would require firms and suppliers to patch and validate changes at much greater speed and frequency, increasing the risk of errors, outages, and disruption across interconnected systems. Their systemic significance will therefore depend not only on how material those vulnerabilities are, but also on how effectively firms and suppliers can absorb a persistently higher operational burden of triage, patching, testing, and recovery.^{footnote [10]}

The UK’s National Cyber Security Centre (NCSC) has similarly concluded that frontier AI is already changing the cost, speed and scale of cyber operations, including by helping malicious actors identify weaknesses, write code to exploit those vulnerabilities and automate parts of multi-step attack chains. Consistent with that assessment, the NCSC has stressed that organisations should focus on getting established cyber security fundamentals right – including patching known vulnerabilities quickly, managing access, securing networks and preparing to respond and recover from incidents – as the most effective near-term defence against AI-enabled cyber threats.

Today’s frontier AI models still have important practical limitations. But progress is fast enough that firms and authorities should not assume those limitations will persist.

Progress has been uneven across cyber tasks. AISI’s testing suggests that the strongest recent gains have been in finding software weaknesses, prioritising them, and carrying out attacks in controlled test environments. These are capabilities that could already be useful to malicious actors. But importantly, AISI’s cyber tests did not include active defenders, and so do not show that frontier models can already carry out fully autonomous, reliable, undetected attacks against well-defended real-world targets. There is separate evidence that current frontier models still lack some of these capabilities needed to meet that more demanding threshold. METR’s May 2026 Frontier Risk Report concluded that leading internal AI agents plausibly had the means to carry out unauthorised self-directed deployments, for example attempting to install or replicate themselves without approval, but could not do so robustly. While not directly testing offensive cyber capabilities, this finding suggests that the most advanced models still struggle to operate reliably, persistently and covertly in more demanding real-world conditions.

However, recent progress has been so rapid that firms and authorities should not assume those limits will persist. In several areas, frontier models have advanced faster than many experts expected and have overcome constraints that previously appeared material. The capabilities that matter most – sustained reasoning, tool use, error recovery and the ability to complete longer task sequences with limited human input – have all improved materially in a short period. Because these are closely related to the capabilities still missing for reliable, covert attacks against defended targets, further progress could narrow some of these remaining constraints quickly. That would reduce the time available for firms and authorities to adapt. This uncertainty is reinforced by emerging evidence that frontier AI models are already being used to accelerate parts of AI research and development, particularly coding and evaluation workflows, though fully autonomous recursive self-improvement remains unproven.^{footnote [11]}

3.2: Drivers of cyber risk to the financial sector

Whether frontier AI benefits malicious actors more than defenders will depend on three factors: how much further the technology improves, how quickly malicious actors can access frontier models, and how quickly firms strengthen the way they identify and fix vulnerabilities.

Based on current capabilities, frontier AI models are increasingly capable of identifying and exploiting software vulnerabilities at greater scale and over multiple stages, potentially increasing the number of attacks, shortening the time between a weakness being discovered and exploited, and conceivably lowering the barrier to entry for a wider range of malicious actors. At the same time, it could help defenders reduce the time between a weakness being found and fixed. A key question is therefore not only whether frontier AI improves cyber capabilities overall, but whether it is more useful in practice for offensive than defensive cyber tasks. If frontier AI is better able to support vulnerability discovery, exploitation and parts of multi-step attack chains than the more complex end-to-end defensive workflow needed to validate, prioritise, test and safely remediate weaknesses, the balance of advantage could shift towards malicious actors.

The balance of advantage between malicious actors and defenders will depend on at least three factors:

1. Further capability improvements. The latest frontier models already appear to have cyber capabilities that could assist both malicious actors and defenders, but their capabilities remain uneven, which limits their utility for fully autonomous cyber operations (whether offensive or defensive). If limitations that currently constrain offensive cyber use cases were overcome, malicious actors would be more likely to be able to find and exploit weaknesses faster than defenders can identify and fix them. That asymmetry risk is amplified by the fact that attackers need only identify and exploit one overlooked weakness to succeed, whereas defenders must identify, prioritise and remediate a much wider set of vulnerabilities across their own systems and any third-party dependencies.
2. How quickly these capabilities spread to malicious actors. Where frontier models already have cyber capabilities that could benefit attackers, the extent of the risk today also depends on how far those capabilities spread to malicious actors. Frontier AI models are currently proprietary or ‘closed-weight’, ie model weights – which are needed to replicate the model – are kept confidential by the developer and users can only interact with the models via chatbot interfaces or APIs. Access is either restricted to trusted users, or models are released with guardrails designed to detect and block harmful use, though current guardrails are not fully reliable.^{footnote [12]} The latest proprietary frontier models can be costly to use at scale, which may currently limit sustained use by lower-resourced malicious actors. For now, defenders may therefore be able to use frontier AI more easily than attackers. But these barriers are more likely to delay misuse than prevent it, for three main reasons:
   1. Malicious actors may still gain access through workarounds such as bypassing safeguards, exploiting weak customer controls, compromising accounts, abusing trusted access, leaking models or stealing model weights.
   2. Model inference costs have fallen materially over time as hardware, model efficiency and serving infrastructure have improved (Epoch AI), and competition between providers could put further downward pressure on user prices. If that continues, advanced frontier capabilities could become economically viable for a wider range of malicious actors.
   3. Open-weight models – models of which the core parameters are made available for others to run and adapt – are an important medium-term issue. Evidence suggests they may lag the most advanced closed models by only four to eight months, and that their safeguards are often easy to remove (AISI). Because they can also be run more cheaply than proprietary models, the point at which open-weight models become dangerous and practical to use may arrive before they fully catch up with frontier models. Restrictions on access to proprietary frontier models may therefore buy defenders time but are unlikely to prevent malicious use indefinitely. That assessment is consistent with the June 2026 joint statement by the NCSC and other Five Eyes cyber security agencies, which warned that the cyber capabilities of frontier AI are developing on a timeframe of months, not years.
3. How quickly firms adapt their existing systems. Frontier AI could help firms strengthen their defences, but access to these systems alone will not be enough.^{footnote [13]} Firms will need to speed up their end-to-end management of software vulnerabilities since, if AI helps malicious actors move faster, each manual step in firms’ processes becomes more costly. In practice, that is likely to be a significant operational challenge as vulnerability management is a multi-step process. Identified vulnerabilities must be checked to rule out false positives, assessed for materiality, mapped to specific systems and business services, assigned to the right teams, fixed, tested and then deployed safely into live environments. Factors such as dependence on legacy technology, complex change-management processes and large numbers of external applications and software components will likely exacerbate this adaptation challenge as they can make rapid remediation difficult even when a weakness is already known. Faster patching can itself create operational risks if changes are rushed, insufficiently tested or difficult to coordinate across interconnected systems.

Increased geopolitical tensions also create an environment of heightened risk of cyber-attacks – including those enabled by frontier AI – which could coincide with, and amplify, other stresses. See Section 1 for developments in the global risk outlook.

3.3: Potential risks to financial stability

There is significant uncertainty around each of the three factors set out above. The FPC has therefore considered three indicative scenarios for how these dynamics could evolve (Figure 3.1). Crucially, the required frequency, intensity and speed of patching for the financial system will be substantially higher in any scenario going forward, which will increase the rate of errors and risk of operational disruption.

In the best-case scenario, defenders would use frontier AI well enough to stay ahead of attackers. The backlog of unfixed weaknesses would remain broadly stable, though the sector would face a higher baseline of operational strain and increased dependence on frontier AI vendors. Cyber incidents could still crystallise in the real economy, but their financial stability implications would remain manageable.

This outcome would depend on three conditions: frontier AI not rapidly gaining further offensive cyber capabilities; those capabilities not spreading rapidly to malicious actors; and defenders adapting quickly enough. In that case, major firms, key suppliers and important open-source software projects could retain the advantage. They would be able to find and fix weaknesses faster than attackers could find and exploit them at scale, keeping the overall backlog of unfixed weaknesses broadly stable.

In this more favourable scenario, the main risk to the financial system would be ongoing operational strain, including continuous patching at high volume, pressure on firms to make changes quickly and where the fixes themselves may result in incidents that cause outages. Resources would need to be diverted on a sustained basis towards defensive adaptation, potentially weighing on innovation and other investment. To the extent that firms come to rely on a handful of frontier AI providers to improve their defensive cyber capabilities, to identify vulnerabilities, and to fix processes, this could also represent a new channel of concentration risk. Cyber disruption in other sectors of the economy could still affect financial stability indirectly, for example where disruption to core operating systems or supply chains temporarily impairs non-financial firms’ cash flows and debt servicing capacity. But in this scenario, those effects are assumed to remain contained, with limited impact on bank balance sheets.

In an alternative scenario, financial institutions would strengthen their defences materially, but systemic risk would still arise through correlated disruption at shared providers, common software components and critical infrastructure sectors.

In this scenario, frontier AI capabilities improve further, but important practical limitations remain, preventing reliable, fully autonomous attacks at scale against well-defended targets. Large firms, major suppliers and authorities adapt, using frontier AI and other measures to strengthen identification, prioritisation and remediation of vulnerabilities. But resilience is uneven across the wider financial system, especially beyond the largest firms and most established providers. Where multiple firms rely on the same providers, software components or essential services, a vulnerability, compromise or defensive shutdown at a common supplier could affect several institutions at once. Because finance is tightly interconnected with other critical sectors, including energy and telecoms, disruption in those sectors could quickly spread to the financial system.

Systemic risk would therefore arise less from a rapid increase in firms’ own backlogs of vulnerabilities than from correlated disruption through common dependencies. Faster attack and remediation cycles would still create significant operational strain, especially for less-well-resourced firms. But the more important systemic concern in this scenario is that smaller shared suppliers, managed service providers, common software components or critical infrastructure sectors (such as telecoms) could become channels through which cyber disruption spreads across multiple firms at once – whether through compromise at a common provider, precautionary shutdown of a shared service, or simultaneous difficulty patching and restoring connected systems. This shared dependency channel was also emphasised in recent IMF analysis of AI-enabled cyber risk.

In a worst-case scenario, the backlog of known but unfixed software weaknesses would rise sharply, materially increasing the risk of a system-wide cyber event.

In a worst-case scenario, frontier AI would continue to improve rapidly, including in ways that significantly enhance offensive cyber capability. Malicious actors would gain access quickly, either through leaks from closed models or fast progress in open-weight models, and firms would not adapt quickly enough. The number of known but unaddressed software weaknesses would then rise sharply, including in third-party software used widely across the sector.

In this scenario, the risk of a system-wide cyber event would increase. A larger backlog of known software weaknesses would raise the likelihood that the same or similar weaknesses could be exploited at several firms and/or third-party suppliers at once, or firms could be forced into emergency action before patches have been properly tested. Suppliers may be unable to support all affected clients at the same time, and disruption could spread through core financial services including payments, trading, clearing and settlement. Severe enough cyber incidents could create uncertainty and weaken confidence in the broader financial system, especially if their scale and impact were hard to judge in real time.

3.4: Priorities for firms and authorities

Cyber risk has long been recognised by the FPC, wider authorities in the UK and globally, and the financial sector as a core threat to financial stability.

The financial sector has long been – and will remain – an attractive target for cyber-attacks. The system depends heavily on digital infrastructure, is highly interconnected, offers potential financial reward to malicious actors, and provides services on which households and businesses rely every day.^{footnote [14]} In the Bank of England’s 2026 H1 Systemic Risk Survey (conducted before the latest frontier models were announced), 82% of respondents cited cyber-attack as one of their top five risks to the UK financial system (Chart 3.2). 26% of respondents said cyber risk was the single biggest risk to the financial system, making it the second most cited response.

Chart 3.2: The perception of cyber-attacks as a key source of risk to the UK financial system has significantly increased over the last decade

Results of the 2026 H1 Systemic Risk Survey

• Sources: Bank of England Systemic Risk Surveys and Bank calculations.

Since 2013,^{footnote [15]} UK authorities have built a substantial framework to strengthen the financial sector’s cyber resilience. This includes the operational resilience regime, cyber assessment tools (including CBEST, STAR-FS, CQUEST and CORST) and publishing effective practices on cyber response and recovery.

Whether frontier AI creates material risks to financial stability will depend heavily on how quickly and thoroughly firms respond.

Firms’ cyber resilience and recovery work should remain anchored in severe but plausible scenarios. But frontier AI cyber capabilities may move the bar for these severe but plausible scenarios by making faster, larger-scale and more correlated disruption more credible. Firms – including PRA-regulated firms and financial market infrastructures – should therefore not treat frontier AI as a minor extension of the existing cyber threat, nor assume that the change in risk profile is static from here on. The issue is not that cyber risk is new, but that frontier AI may materially change its speed, scale and economics. Firms should also plan on the basis that a higher volume of vulnerabilities and a faster pace of patching may become a sustained feature of the threat environment, rather than a temporary surge, and that this could itself create operational and resilience risks if change, testing and recovery processes are not able to keep pace safely. These developments reinforce the importance of firms acting on the May 2026 joint statement from the Bank, FCA and HM Treasury, which the FPC strongly endorses, and which sets out practical steps firms should take now to prepare for and reduce frontier AI-related cyber risks within the existing operational resilience framework, across several domains:

• Governance and strategy: firms should ensure their boards and senior management have sufficient understanding of frontier AI risks. This is important to set strategic direction and oversee how control functions manage risks. Investment and resourcing decisions should reflect the emerging threat, including increased exposure from end-of-life systems or those out of vendor support.
• Identification and risk management of vulnerabilities: frontier AI models can rapidly identify and enable exploitation of a potentially large number of vulnerabilities across firms’ technology estates. Firms should be able to triage, prioritise, risk assess, and remediate vulnerabilities more quickly, more frequently, and at scale, including through automation where appropriate, while mitigating the operational risks from doing so.
• Managing risks from third parties: firms should effectively manage frontier AI cyber risks from third parties and supply chains, including open-source software. This means firms should have the capabilities to identify, monitor and manage external applications, libraries and services integrated into their networks. Firms should be prepared to address and remediate vulnerabilities identified by third parties at scale.
• Protection: effective access management, network security, and data protection should enable firms to reduce the attack surface a frontier AI model might access and limit the likelihood and impact of such attacks. Firms should consider adopting automated and AI-enabled defences to operate at comparable speed to AI-driven attacks.
• Response and recovery: firms should be able to respond to and recover from disruption quickly. Firms should read and consider the effective practices on cyber resilience published by the Bank, PRA and FCA in October 2025.

Some smaller firms may be more reliant on third-party software providers than larger institutions and therefore have simpler technology estates. But many may still find it particularly difficult to adapt their vulnerability management processes to this new threat environment, because they often have fewer cyber resources, less in-house expertise, and less capacity to absorb the operational cost of faster implementation of third-party provider fixes. The FPC has previously noted that the cost of cyber resilience for smaller firms can be a significant barrier to growth. In that context, these considerations may strengthen the case for practical support tools and shared defensive infrastructure that could help reduce the burden on firms that lack the scale to build these capabilities on their own. The FPC therefore welcomes recent work by GCHQ on a blueprint for a new national cyber defence capability that aims to use advanced AI to strengthen machine-speed cyber defence. The FPC also supports initiatives such as the NCSC’s Cyber Action Toolkit, which is designed to help sole traders, micro businesses and small organisations – including financial firms – strengthen their cyber resilience.

The pace at which vulnerabilities are found, disclosed, prioritised and remediated will also depend on actors beyond regulated firms, including frontier AI developers, software vendors, open-source maintainers and trusted cyber-security conveners.

For authorities, the priority is to reinforce existing expectations while adapting efforts to ensure system-wide resilience to this threat environment.

UK authorities are already reinforcing these expectations through supervision and sector engagement. In addition to the May 2026 joint statement, authorities are also engaging firms through the Cross Market Operational Resilience Group (CMORG), which in May 2026 convened firms, authorities and the NCSC to discuss the sector’s response to emerging frontier AI models and subsequently ran a frontier AI risk mitigation webinar for UK regulated financial services firms. CMORG issued guidance on frontier AI and cyber resilience in June. In addition, the Bank and PRA’s upcoming consultation on Cyber and Information, Communication and Technology risk management will consider issues relevant to frontier AI cyber risks. The Financial Stability Board (FSB) is also consulting on sound practices for the responsible adoption of AI, including elements relevant to the cyber risk posed by frontier AI. Developments in frontier AI increase the value of clearer common expectations, sector-wide awareness and more coordinated adaptation across firms and authorities.

System-wide monitoring and preparedness are also likely to become more valuable. As noted in its response to the Treasury Select Committee’s inquiry into AI in financial services, the Bank of England is working to incorporate AI scenarios into different forms of cyber and operational testing for the financial sector, building on existing FPC cyber stress tests that take a system-wide approach (in addition to its broader macrofinancial analysis – refer to Section 2). In this threat environment, that work is likely to need to place greater weight on simultaneous multi-firm disruption, correlated disruption across technology supply chains, and whether existing expectations, tools and coordination arrangements remain sufficient if frontier AI compresses the time available to identify, contain and recover from cyber incidents. This includes system-wide considerations relating to:

• Whether the present standard of recovery capability across regulated firms remains adequate given the scope for more severe and fast-moving disruption scenarios, or whether enhanced response and recovery options should be expected of certain systems supporting important business services that are critical to UK financial stability. For example, 'bare metal rebuild’ in an isolated recovery environment or ‘stand-in facility' to deliver a minimum level of service.
• The importance of operationalising the UK’s Critical Third Party (CTP) regime, and the resilience expected of other material third-party technology providers that may not meet the threshold for CTP designation.
• Whether the current balance of firm and industry level activity on response and recovery remains appropriate, or whether greater coordination and industry level co-operation, including on information sharing, cyber-protection infrastructures and coordinated response and recovery options for important business services, may be warranted.

Deepening international coordination will also be critical. The financial system is globally interconnected, and cyber disruption can spread across borders through common technology dependencies, financial institutions and market infrastructure. Differences in cyber capability, resilience and recovery capacity across jurisdictions could therefore have consequences beyond the jurisdiction in which an incident begins. Macroprudential authorities will need to continue working through international forums, including the FSB and G7, to build common understanding of how frontier AI is changing cyber risk and to support consistent approaches to monitoring, testing and preparedness.

This preparedness agenda also matters beyond the financial sector, emphasising the importance of deepening coordination across domestic authorities too. The Government’s Cyber Security and Resilience Bill is intended to strengthen cyber regulation for essential and digital services, helping over time, if effectively implemented, to improve the resilience of critical sectors such as energy and telecoms on which the financial sector depends. If frontier AI compresses the time between vulnerability discovery, exploitation and disruption across interconnected sectors, resilience in those sectors becomes more directly relevant to financial stability.

3.5: Implications for financial stability beyond cyber risk

Although cyber risk is the most immediate concern from rapidly advancing frontier AI, the technology could also affect financial stability in other ways.

The same underlying developments in frontier AI that have precipitated this change in cyber capabilities will drive new capabilities across a variety of fields. Over time, frontier AI could improve efficiency, innovation and resilience in financial services and the wider economy. But these benefits will depend on firms adopting the technology responsibly. These channels are less immediate, and in some cases more speculative, than the cyber channel discussed above, but they are worth monitoring given the speed of capability development. Developments in frontier AI’s cyber capabilities highlight the rapidity of progress and the transformative potential of those capabilities, which may manifest in other future ‘step change’ moments.

In its 2026 Q1 Record, the FPC highlighted two applications of more autonomous AI systems that could matter for how the financial system works and for financial stability, building on the framework set out in its April 2025 Financial Stability in Focus:

1. Payments: more autonomous AI systems are already becoming an area of rapid innovation in payments. They could change how payment decisions are initiated, optimised and carried out, increasing the speed and scale of financial flows, which could bring significant economic benefits. Such innovation also raises new questions about authorisation, traceability, fraud detection and liability, resilience, legal accountability, and the liquidity arrangements needed to manage failed, disputed or reversed payments. IMF analysis has highlighted a key tension: AI systems are probabilistic, while payment infrastructures usually require predictable and deterministic outcomes. This matters because payment systems must produce clear, consistent and legally certain outcomes even under stress. Governance and system design will therefore be important in determining whether possible benefits can be realised safely, without undermining the integrity of the payments systems on which the real economy relies.
2. Financial markets: evidence to date suggests that trading firms are using more autonomous AI systems, primarily for research, coding support, surveillance and other lower-risk operational tasks, rather than for fully autonomous trading.^{footnote [16]} Key challenges include the difficulty of reliably anticipating AI outputs, and of validating and bounding more autonomous systems in markets where conditions and economic relationships can change quickly. But if AI systems become better at core financial decision-making tasks and start to outperform more traditional models in testing environments then firms may begin to rely on them more directly in trading or portfolio decisions in ways that change the speed and nature of adjustment to new information and increase the risk of correlated behaviour. A key policy question is therefore not only whether firms can understand and govern these models effectively, but whether the financial system, including regulators, has the visibility, indicators, and intervention mechanisms to monitor and contain the resulting collective behaviours. To explore these risks, the Bank is working with the BIS Innovation Hub London Centre and other stakeholders on Project Logos, which aims to enable central banks to observe and analyse the behaviour of large language model (LLM)-based agents acting as portfolio managers in a simulated financial market environment.

Bank staff are therefore currently undertaking deep dives on agentic payments and agentic trading.

More broadly, recent advances in frontier AI also underline the possibility that progress in other technologies – such as quantum computing – could be faster than expected.

Frontier AI, cyber risk and quantum preparedness are converging challenges. Rapid advances in AI highlight the urgency of strong asset management and cyber basics, including accurate mapping of systems, software, data and cryptographic dependences, as immediate priorities, not just future needs.

Given the scale of the task to transition to post‑quantum cryptography, and the need for alignment with multi‑year investment cycles, firms should begin planning now. That is consistent with the G7 Cyber Expert Group’s roadmap, co-chaired by the Bank of England and the US Treasury, which encourages financial sector stakeholders to take a coordinated and timely approach to migration to quantum-resistant cryptography. This includes mapping cryptographic dependencies and the broader IT estate to identify vulnerabilities and prioritise resilience measures. Firms should also retain flexibility to accelerate migration if developments in quantum capability, standards, vendor readiness or the threat environment require it.

The quantum threat goes beyond cryptography. It may entail system architecture changes, complex migration efforts and supply-chain fragilities, including potential hardware or skills shortages, if left too late. In practice, quantum risk management aligns with operational resilience, requiring coordinated upgrades and careful planning to avoid disruptions, akin to other major technology transitions. 

Firms’ quantum safety is only as strong as that of the material third-party providers that support their important business services. In line with existing expectations on management of third-party risks (eg PRA Supervisory Statement 2/21), firms should therefore seek assurance and information on how their material third parties are preparing for post-quantum safety.

4: The resilience of market-based finance

Monitoring and strengthening the resilience of market-based finance (MBF) is crucial to ensure that the financial system can absorb shocks, rather than amplify them. This protects overall economic and financial stability and allows MBF to continue providing vital services to households and businesses.

Market-based finance is an interconnected system of markets, market infrastructure such as central counterparties (CCPs), and non-bank financial institutions (NBFIs) such as insurers, hedge funds and private finance firms. MBF has grown significantly in importance, and NBFIs now account for around half of global and UK financial sector assets. Previous events – such as the March 2020 dash for cash and the Autumn 2022 liability-driven investment (LDI) episode – have highlighted that vulnerabilities in MBF could pose risks to UK financial stability. While such vulnerabilities – including those related to the use of leverage by investors in sovereign debt markets, and opacity and interconnectedness in risky credit markets – have not crystallised following the onset of the conflict in the Middle East, they persist and could amplify stress under adverse macroeconomic conditions. Reflecting this, the Bank is strengthening its ability to identify and monitor risks in MBF. This includes a second system-wide exploratory scenario (SWES), which is exploring risks and dynamics associated with the private market ecosystem (Box A), and enhancements to its market surveillance framework (Box B).

4.1: Developments in the resilience of gilt and gilt repo markets

UK gilt markets have undergone significant structural changes in recent years, including historically high issuance, with higher proportions issued at shorter maturities and a shift in the investor base towards more price-sensitive investors.

UK gilt and gilt repo markets play important roles within the financial system, such as financing government activity and acting as a benchmark for the pricing of other financial instruments such as corporate bonds. In doing so, government bond markets facilitate the provision of vital services, investment, and sustainable economic growth. Repo markets facilitate the flow of liquidity – in the form of cash and gilts – across the financial system and are a source of funding for market participants’ leveraged strategies. NBFIs play a significant role in these markets, including through leveraged gilt and repo positions.

Government bond markets globally – including in the UK – have undergone significant structural change in recent years, including higher issuance with an increasing skew towards shorter-term bonds (Section 1). Alongside this, long-term, less price-sensitive investors such as defined benefit (DB) pension funds have reduced their demand for gilts. Many DB schemes are now closed to new members, reducing the need for liability-driven hedging and demand for long-dated and inflation-linked gilts. This reflects the transition away from DB schemes towards defined contribution (DC) schemes. DC schemes do not guarantee specific benefits in the same way as DB schemes and therefore have less need to hedge long-duration liabilities. Consistent with this, the preferred maturities of the gilt investor base have shifted shorter in recent years. The weighted average maturity of pension fund and LDI gilt net purchases has fallen from around 25 years in 2018 to about 14 years in 2026 (Chart 4.1).

Price-sensitive investors (including hedge funds) have increased their participation in government debt markets. In normal times, hedge funds can support market functioning by warehousing risk and intermediating between different market participants within the sterling markets ecosystem. But these investors require a risk premium and are more active in entering and exiting positions as conditions change, relative to more price-insensitive investors like pension funds, which hold bonds for liquidity and hedging purposes. The relatively high use of leverage by a small number of hedge funds increases the risk of a disorderly unwind of positions and a jump to illiquidity in core UK markets.

Despite high levels of volatility, UK gilt and gilt repo markets have remained resilient, with functioning commensurate with market conditions.

The cash gilt market has experienced significant price volatility since the onset of the Middle East conflict. As set out in Section 1, the energy-driven supply shock has caused inflationary pressures across economies globally. After the onset of the conflict, UK and euro area government bond yields increased by more than US Treasury bond yields, reflecting concerns over those economies’ exposures to imported energy costs. Headline measures of market functioning have remained within historical norms (Chart 4.2). While volumes were elevated throughout March and April, they remained below the levels observed during the April 2025 tariff announcements. Bid-ask spreads widened during the period and there was some deterioration in liquidity. However, this was less severe than observed during the 2022 LDI and 2020 dash for cash episodes and remained consistent with the degree of volatility (Chart 4.2). In gilt repo markets, the spread between overnight gilt repo rates and Bank Rate remained narrow throughout the period, aside from expected increases around period-end reporting dates, indicating that the market was functioning well. More recently, both cash and repo markets have remained resilient, and market conditions have improved since March 2026, though liquidity remains somewhat reduced relative to long-run averages.

Chart 4.2: The gilt market has remained resilient

Weekly gilt volume across sectors (top panel); and bid-offer spread versus realised volatility for the 10-year benchmark gilt (10-day rolling measure) (bottom panel) ^(a)

• Sources: Financial Conduct Authority MiFID II data, Bloomberg Finance L.P and Bank calculations.
• (a) Realised volatility is calculated as the 10-day average of the high-low intraday range in the 10-year benchmark gilt yield.

Hedge funds unwound commonly-held positions prior to and following the onset of the conflict in the Middle East.

Hedge funds significantly reduced their positions across European, UK and US government bond markets, in response to a reassessment of the expected path of policy rates. Hedge fund net sales in the cash gilt market were sizeable, but the adjustment was orderly and more gradual than in previous stress events. For example, during the Autumn 2022 LDI episode, net gilt sales by insurance companies and pension funds prior to the Bank’s intervention were around half the scale of the hedge fund sales observed recently and 80% of the PV01.^{footnote [17]} However, these LDI stress-related sales were concentrated within a single week, whereas recent hedge fund sales were spread across a month. This faster, more concentrated selling during the LDI stress overwhelmed market capacity and caused significant market dysfunction, as shown in Chart 4.2. During these recent events, while hedge funds’ repositioning amplified some moves in gilt yields, markets continued to function, with some hedge funds also stepping in as buyers throughout March and April.

These dynamics were also reflected in gilt repo markets. Net hedge fund gilt repo borrowing declined by around 40% by mid-April, from £100 billion at the time of the December FSR (Chart 4.3). Positions have since fluctuated before increasing again from end-May to around £85 billion.

Markets absorbed high volatility and volumes, partly reflecting resilience built in recent years.

Elevated hedge fund gilt sales were absorbed by real money investors (such as asset managers, insurance companies and pension funds) that were willing and able to buy, helping to maintain market functioning. The LDI sector also remained resilient, reflecting policy measures taken since the autumn 2022 gilt market stress to strengthen liquidity buffers and operational readiness. In gilt repo markets, the stability in functioning was a reflection of the widespread availability of sterling cash. Money market funds (MMFs) were significant providers of liquidity, and net lending reached an all-time high during the period. As a result, markets were not subject to sustained one-way flows, and dealer intermediation capacity remained available.

Large price movements in a number of markets led to margin calls. Commodity markets – particularly energy markets – experienced large price moves and material margin calls, but markets continued to function. In interest rate derivatives, hedge funds in aggregate were positioned for lower short-term policy rates and so faced variation margin calls on these positions. However, aggregate estimated variation margin calls on interest rate derivatives faced by non-bank sectors were relatively modest during this episode, at around £0.4 billion between 27 February and 9 March, and materially below levels seen in previous stress episodes (for example, around £13 billion during the autumn 2022 LDI stress episode).^{footnote [18]} This reflected both the smaller magnitude of the interest rate shock to longer-term rates compared to the LDI episode, and a positive inflation shock that was favourable for insurers, pension funds and LDI vehicles’ inflation swap positions. That offset larger variation margin calls faced by hedge funds. It also reflected changes in sectoral positioning, including reduced exposures among pension funds and LDI vehicles. This limited the overall need for asset sales by non-banks to meet margin calls, reducing the risk of amplification of price moves. The use of Bank facilities, such as the Short-Term Repo (STR) and Indexed Long-Term Repo (ILTR), remained steady throughout the period, helping to anchor rates and ensure that pressures were short-lived.

While core markets have remained resilient through recent events, that resilience could be tested by future shocks.

A critical factor in the continued functioning of these markets was the provision of gilt repo. The first SWES exercise illustrated that this cannot be guaranteed in a more severe stress. In the exercise, liquidity demand led to redemptions from MMFs and open-ended funds, prompting asset sales. Gilt repo market conditions tightened as banks, in response to counterparty credit risk concerns, shortened repo tenors, increased haircuts, and were more selective in extending financing. Since the first SWES exercise, the cash gilt and gilt repo markets have undergone significant structural change (Section 4.1). Hedge funds have increased their footprint in the gilt market and shifted from being net cash lenders in gilt repo to net cash borrowers. This leaves hedge funds more exposed to changes in financing conditions and increases the likelihood that distress in this sector could disrupt gilt markets.

This underlines the importance of locking in resilience through reforms such as those recently announced for UK MMFs, which are aimed at improving their ability to meet redemptions in a stress without amplifying market disruption, as well as the Bank’s ongoing work to strengthen the resilience of gilt repo markets. It also highlights the need to enhance the monitoring framework for gilt market functioning, including through system-wide exercises such as the first SWES, to understand system dynamics, as well as improved data and surveillance capabilities (Box B).

Underlying structural vulnerabilities in MBF remain, including leveraged borrowing by hedge funds in gilt repo markets.

The recent fall in hedge fund net repo borrowing was largely driven by an unwind of positions at the front end of the curve, as interest rate expectations shifted. However, hedge fund net repo borrowing remains elevated by historical standards and continues to be driven largely by leveraged relative value strategies. These strategies use net repo borrowing to take advantage of differences between the prices of gilts, or between gilts and related interest rate instruments, with the cash-futures basis trade being one example. As Chart 4.4 shows, the recent fall in hedge fund net repo borrowing does not appear to have been accompanied by a material unwind of long gilt future positions, which remain high. Additionally, hedge funds rely on shorter-term funding via repo markets, leaving them exposed to rollover risks should repo market conditions deteriorate (Box B, Figure D). These strategies therefore remain vulnerable to simultaneous deleveraging, particularly if funding conditions were to tighten significantly.

Chart 4.4: Data suggests that hedge fund net gilt repo borrowing is related, at least in part, to the popularity of the cash-futures basis trade

Change in hedge fund net gilt repo borrowing and gilt futures open interest since January 2024 ^(a)

• Sources: ICE Futures Europe, Sterling money market data (SMMD) and Bank calculations.
• (a) Spikes seen in gilt futures open interest relate to contract roll periods.

Risks also remain concentrated among a small number of firms pursuing similar trading strategies across jurisdictions and markets. Hedge funds with the highest leverage in the gilt market also have large repo positions in other government bond repo markets. Should investor sentiment on the sovereign debt outlook deteriorate abruptly, this could affect the UK market directly, or indirectly via spillovers from other sovereign debt markets, with potential implications for domestic financing conditions.

In addition, Bank staff analysis using entity-level probability of default (PD) estimates from banks’ internal rating models indicates that hedge funds from offshore financial centres (OFCs) exhibit higher and more volatile default risk than other NBFIs active in UK markets, with a pronounced upper tail of risk that has increased over time.^{footnote [19]} This pattern is particularly evident among hedge fund structures whose associated country of risk is the Cayman Islands. Given their concentrated, leveraged and cross-border activity in core UK markets, stress affecting these firms could be amplified across markets. This underscores the importance of robust counterparty risk management and strengthened international coordination to support effective monitoring and surveillance of cross-border risks.

These vulnerabilities could be amplified further through interactions with other parts of the financial system. Hedge funds’ equity prime brokerage balances are at record levels (Section 1). According to transaction and survey data, many hedge funds are active in both gilt and equity markets (Box B). A sharp correction in equity markets could transmit stress into gilt markets if hedge funds making losses on equity positions were to unwind gilt market positions in response. Box B sets out how data can be combined with insights from firm engagement to undertake scenario analysis and assess how financial market participants can deepen or broaden a shock, by amplifying its impact within a market or transmitting it across markets.

4.2: Private markets and risky credit markets

Private markets and risky credit provided by MBF play an increasingly important role in supporting growth, by channelling funding towards productive investment in the real economy, and by contributing to a competitive and diverse lending landscape that meets the varied needs of UK businesses.

Market-based corporate finance is an important part of the funding ecosystem for UK businesses. It includes a broad range of channels through which corporates raise finance outside traditional bank lending, including public credit markets, leveraged lending and private markets. Private markets are increasingly central in supporting corporates and providing funding for infrastructure, real assets, real estate and venture capital. The long-term nature of private capital can enable fund managers to act less cyclically. This means they can continue to provide finance in macroeconomic downturns, although this benefit can be weakened by high leverage or redemptions.

Risky credit markets comprise high‑yield bonds, leveraged loans and parts of private credit, and primarily fund non‑investment grade borrowers. These markets typically provide finance to more highly indebted corporates. Private markets play an important role in this ecosystem, with private equity and private credit markets having grown significantly over the past decade.

Their growing scale and significance make it vital to understand the resilience of the funding they supply to the real economy in bad times as well as good.

While these markets bring significant benefits, they can also give rise to potential financial stability risks. Stress to these markets could reduce the availability or increase the cost of finance for some businesses, amplifying the impact of an adverse shock on the real economy. Within this ecosystem, the provision of financing from private markets includes the extensive use of leverage (including leverage on leverage), complex structures, and interconnections with banks, insurers and other parts of the financial system. In addition, the opacity and valuation uncertainty associated with private markets can make it harder to assess where risks sit and how losses could be transmitted and amplified. The July 2025 FSR explored the implications of the growth in private markets for financial stability and real economy financing in more detail.

Investor sentiment in parts of private and risky credit markets had already weakened ahead of the Middle East conflict.

Concerns about asset quality in leveraged finance and private credit have grown, including around weak underwriting standards. The outlook has worsened since the December FSR, including in relation to geopolitical risks and uncertainty around the impact of a widespread adoption of AI on incumbent companies’ business models and profitability, including in sectors such as software where risky credit markets have significant concentration (Section 2). These concerns are reflected in valuations for some publicly listed private credit funds or traded US business development companies (BDCs) and software leveraged loan prices. Furthermore, a subset of US private credit funds, mainly non-traded BDCs, have faced liquidity pressures arising from elevated redemption requests. Although most private credit funds are closed ended, non-traded BDCs have retail investors – often high-net-worth individuals. Redemption requests were elevated in several of these retail funds, with some limiting redemptions. These funds generally operate quarterly redemption offers, typically capped at 5% of the fund. While these caps have largely operated as expected, they can encourage investors to redeem pre-emptively due to concerns about their future ability to access their capital or valuations. Furthermore, while the non-traded BDC sector remains relatively small in aggregate, liquidity stress in these funds could spill over to other parts of private credit and private equity markets, as well as to other parts of risky credit markets.

Recent defaults have also illustrated how weaknesses in risky credit markets can crystallise and subsequently pose risks to systemic institutions.

The default of Market Financial Solutions Limited, a non-bank mortgage lender, in February was largely driven by similar underlying vulnerabilities as those highlighted by the earlier defaults of First Brands and Tricolor. These include high leverage, weak underwriting standards, opacity, and complex structures. Such weaknesses may have been a contributing factor that allowed significant alleged fraud to be undetected for a period.

These corporates were financed by both banks and NBFIs and illustrated the complexity and interconnections between them. Some UK banks experienced losses related to these corporate failures (Section 6). In light of recent developments, some banks have increased scrutiny of their exposures to private markets. This may represent a prudent reassessment of credit risk for asset-backed lending and fund financing, but could also contribute to tighter lending conditions for corporates going forward.

In traded risky credit markets, there has been divergence between the riskiest, distressed credit and other sub-investment grade securities.

Despite weaker investor sentiment, and the potential for the conflict in the Middle East to affect these markets, broader risky credit markets do not appear to have repriced materially so far. Credit spreads overall remain compressed relative to historical levels (Section 1) and issuance has continued across risky credit markets. However, spreads between lower-rated (‘CCC’) and higher-rated (‘BBB’) debt have widened in recent months (Chart 4.5), reflecting concerns about the lowest rated corporates.

Chart 4.5: The spread between the riskiest and safest high-yield debt has widened since the conflict began

US high-yield spreads by rating bucket

• Sources: ICE BofAML and Bank calculations.

Some corporates and lenders are taking steps to manage refinancing and cash flow challenges.

A worsening in the outlook could tighten funding availability for UK corporates and amplify stress through lower earnings and valuations. This could be exacerbated where private capital or risky credit markets tighten, particularly as borrowers refinance maturing debt. To date, issuance has continued across risky credit markets. However, many borrowers are managing pressures by pushing refinancing risks into the future. Where borrowers face difficulties, they have amended and extended loans, used payment in kind structures to manage cash flow challenges, or both, among other forms of forbearance. However, these measures may not be available to all borrowers or sustainable in the longer term. Similarly, some funds active in private markets that provide equity and credit have established continuation vehicles that transfer investments in a fund reaching the end of its lifecycle to a new fund often with a similar investor base. This can provide liquidity to existing investors while delaying exit and extending holding periods.

Looking ahead, a material share of UK private credit debt and leveraged loans are due to be refinanced in the coming five years. Some pockets of refinancing risk remain within a small subset of risky corporates’ refinancing needs. A substantial portion of UK private debt taken out in 2021 is due to refinance in the coming year, and some borrowers could face tighter conditions given leverage was acquired in a much lower interest rate environment. While private credit remains a small share of total UK corporate debt, corporates defaulting on private credit debt could worsen investor sentiment and spill over into broader risky credit markets, potentially tightening broader credit conditions.

Market structures continue to evolve, and UK life insurers are expected to increase their allocation to private markets.

Insurers have long-term illiquid liabilities making private markets an attractive asset class given their higher liquidity premium. The July 2025 FSR highlighted the increasing interconnectedness of UK life insurers and private markets, including through funded reinsurance exposures, which have been growing rapidly. The PRA has since reviewed the current regulatory treatment of funded reinsurance and concluded that it does not appropriately reflect the underlying risks and is not aligned to that of economically similar assets. It has recently issued a consultation paper which seeks to address this.

In addition to these changes, other factors, such as changes to the Matching Adjustment under Solvency UK have made it more attractive for insurers to invest directly in structured and private credit assets. These reforms are intended to support investment in productive finance, including long-term assets that can support UK economic growth. While current exposures remain modest in the UK, supervisory intelligence suggests some firms are planning to increase allocations. One driver of these trends may be the recent acquisition of two large UK life insurers by US private capital firms. The FPC and PRA will continue to monitor these changes, as well as other expected changes such as greater retail investment in the sector.

Private markets and leveraged finance are expected to play an increasingly important role in financing AI investment.

The scale of the external financing required to meet AI investment is expected to impact the whole credit ecosystem – with anticipated roles for private credit, leveraged lending and high yield bonds. Debt financing of AI investment is accounting for a significant share of issuance in private credit and high yield bond markets. This could increase vulnerabilities in risky credit markets over time – refer to Section 2 on the macrofinancial implications of AI for more detail.

If vulnerabilities were to crystallise, they could tighten financing conditions and amplify shocks to the real economy.

To date, the impact of these risky credit market concerns and liquidity pressures at international private credit funds on UK financial stability have been limited. However, these markets remain vulnerable to a tightening in financing conditions. Higher interest rates associated with the negative supply shock resulting from the conflict in the Middle East could increase debt servicing pressures for leveraged borrowers. Lower economic growth could decrease asset quality, increasing pressure on private credit funds. At the same time, structural vulnerabilities in these markets persist, including high leverage, complexity and opacity. This highlights the importance of understanding how UK private and related public credit markets would function in a stress, including the channels through which losses and tighter funding conditions could affect the UK real economy. Accordingly, the Bank’s second system-wide exploratory scenario (SWES) exercise is examining how stress could propagate in these markets (Box A).

4.3: Improving the resilience of market-based finance

Market surveillance plays a key role in identifying financial stability risks, including those from leveraged participants.

The Bank monitors leverage, interconnectedness and concentration in core markets using transaction-level data, market intelligence, and system-wide exercises like the first SWES. Data published in this FSR, including in Box B, is intended to contribute to market participants’ understanding of their exposures relative to the aggregate market and help inform their risk management and preparedness for future shocks. The FPC also intends to continue to expand the data it publishes in this area. The Bank welcomes feedback from market participants on which aggregate market data can best help them manage risk. Box B sets out how the Bank is enhancing its surveillance of MBF vulnerabilities, including through improvements in data, analytical tools and system-wide monitoring capabilities.

The FPC seeks to strengthen the resilience of MBF to mitigate risks to UK financial stability. Where effective and practical, it works to address vulnerabilities domestically while recognising that for many risks, mitigations require international co-ordination.

Given the interconnected nature of MBF, reforms to improve resilience are most effective when co-ordinated internationally. Table 4.A provides a summary of progress against building resilience in MBF domestically and internationally.

5: UK household and corporate debt vulnerabilities

Financially resilient households and corporates help to play a key role in supporting economic growth in the UK.

A stable financial system, alongside resilient households and corporates, is a key factor in supporting sustainable economic growth. Household savings and corporate investment support productivity growth. In addition, the sustainable use of debt enables households to smooth consumption and invest in housing and human capital, and supports corporates to finance productive investment and manage cash flow. In turn, this can strengthen aggregate demand and potential output, provided borrowing remains resilient to shocks.

The FPC supports a resilient financial system and healthy household and corporate sectors by assessing the level of harmful debt vulnerabilities from highly indebted households and businesses. In a downturn, these households and businesses may need to cut back sharply on consumption, investment, or employment to make debt repayments, which could act to amplify shocks and worsen economic downturns. The most indebted households and businesses can default, leading to losses for lenders, amplifying economic shocks to the financial sector.

Since the global financial crisis, the UK’s household and corporate sectors have reduced their debt levels as a proportion of UK GDP (Chart 5.1) and they are now at relatively low levels historically. Over the same period, UK Government debt to GDP has increased, as has been the case in other major economies (Section 1).

Chart 5.1: The UK household and corporate sectors have reduced their debt levels relative to UK GDP over the past 20 years

Debt as a percentage of UK GDP ^{(a) (b)}

• Sources: Bank of England, Bayes CRE Lending Report (Bayes Business School (formerly Cass)), Deloitte, ONS, PitchBook, a Morningstar company, LSEG and Bank calculations.
• (a) Corporate debt refers to private non-financial corporations and excludes public sector, financial and unincorporated businesses.
• (b) Household debt includes all liabilities of the household sector, excluding student loans and financial derivatives.

5.1: Overview of UK economic developments

The conflict in the Middle East caused a sharp increase in global energy prices, with implications for the household and corporate sectors.

Elevated global energy prices have generated inflationary pressures and are likely to weigh on growth. While energy prices have fallen back to around pre-conflict levels since the signing of the Memorandum of Understanding (MoU) between the United States and Iran, the pass-through of recently higher energy prices will weigh on household real income growth and corporate earnings and may increase pressure on vulnerable households and corporates. The UK labour market has weakened, with the unemployment rate increasing over the past two years and the April Monetary Policy Report (MPR) projecting this to increase slightly further, which will also weigh on real income growth.

Market interest rates have increased and put upward pressure on rates faced by households and businesses. In the June 2026 Market Participants Survey, median expectations were for Bank Rate to remain unchanged for the year ahead. That represented a tightening in the median path of around 50 basis points relative to expectations prior to the conflict, at which point reductions in Bank Rate had been expected.

5.2: UK household debt vulnerabilities

UK household balance sheets are strong in aggregate, providing resilience against shocks.

Household balance sheets remain strong. The Household debt to GDP ratio (Chart 5.1) has fallen, as has the household debt to income ratio to around 130%, relatively low compared to the average level since 2000 of 155%. Households in aggregate also have relatively high cash buffers at around 140% of income (versus the post-2000 average of 127%). In addition, debt vulnerability metrics – for example, debt-servicing ratios (Chart 5.3) – are around their long-run averages, providing resilience against shocks.

Mortgage rates have increased, consistent with higher market interest rates, and more households are now projected to see increases in mortgage repayments over the next three years. However, the median increase is projected to be materially smaller than the increases already experienced over the last few years.

Quoted rates on new mortgages have increased. The average rate on a two-year fixed 75% loan to value (LTV) ratio mortgage is 4.92%, 72 basis points higher than at the time of the December FSR. The average rate on a two-year fixed 90% LTV mortgage is 5.32%, 75 basis points higher than as of the December FSR. This reflects the increases in market rates following the conflict.

Higher quoted rates mean more households are projected to see increases in their mortgage repayments when they refinance. A little over 5 million households are projected to see their repayments increase by the end of 2028, compared to nearly 4 million at the time of the December FSR (Chart 5.2).

However, for the typical owner-occupier mortgagor rolling off a fixed rate in the next two years, their monthly mortgage repayments are projected to increase by £45. This is significantly smaller than increases experienced over the last few years – the median repayment increased by approximately £120 between the end of 2022 and end of 2024. Based on the current market curve for interest rates, most borrowers on a two-year fixed mortgage expiring by the end of 2028 are projected to remortgage close to their existing rate and therefore see little change in repayments. However, these borrowers are now unlikely to see repayments fall over coming years as was forecast prior to the Middle East conflict.

But there are still a number of households which are projected to see repayments increase significantly, especially those on a fixed-rate mortgage from before interest rates began increasing from 2022. Nearly 750,000 households that are paying less than 3% interest will be rolling off fixes in 2026 and will see an average increase of £170 per month in repayments.

Chart 5.2: More borrowers are projected to see an increase in mortgage payments by the end of 2028 compared to at the time of the December FSR

Number of owner-occupied mortgages by estimated change in monthly mortgage costs, from June 2026 to December 2028, and comparison to the same projections in the December 2025 FSR ^{(a) (b) (c)}

• Sources: Bloomberg Finance L.P., FCA Product Sales Data and Bank calculations.
• (a) The projection uses the OIS curve as at 25 June 2026 and the latest available data on the stock of outstanding mortgages (2025 H2).
• (b) Changes in payments on variable-rate mortgages are calculated using the implied change in the OIS curve, and changes in payment on fixed-rate mortgages are calculated by assuming that mortgagors refinance onto a typical fixed rate implied by the OIS curve at the point that their fixed-rate contract ends.
• (c) Mortgages with less than £1,000 outstanding are excluded. These data do not include buy-to-let mortgages or mortgages that are off balance sheet of authorised lenders, such as securitised loans or loan books sold to third parties.

While higher borrowing costs and energy prices could place additional pressure on household finances, it would take significantly higher energy prices and borrowing costs to take debt servicing pressures to previous peaks.

The share of post-tax household income spent on mortgage repayments (debt-servicing ratio (DSR)) was broadly flat at 7.5% at the end of 2025. Bank staff project the DSR would increase marginally to a little above 8% by the end of 2028 (Chart 5.3) in a scenario of persistently higher energy prices as per Scenario C from the April MPR.^{footnote [34]} This remains significantly below previous peaks.

The FPC also assesses household resilience using DSRs adjusted for essential spending (such as energy). In the same scenario, the cost-of-living adjusted DSR (COLA DSR) is projected to have a more pronounced increase than the gross DSR to 15%, reflecting higher essential spending on energy and related goods. Although this would be above levels observed over the past decade, it would remain below previous peaks and suggests household debt servicing burdens would remain manageable in aggregate. A significant further shock to borrowing costs (as shown in Chart 5.3), household income or the cost of essential goods would be required for COLA DSRs to return to levels observed in the run-up to the global financial crisis (GFC).

Chart 5.3: Aggregate cost-of-living adjusted DSRs are projected to increase, although it would take a severe scenario for them to return close to previous peaks

Gross and cost-of-living adjusted DSRs. Solid lines give historical data. Dashed lines show projections under Scenario C from the April MPR. Dashed orange line shows a projection under Scenario C plus an additional 300 basis point increase in borrowing costs ^{(a) (b)}

• Sources: Bank of England, Bloomberg Finance L.P., FCA Product Sales Data, ONS and Bank calculations.
• (a) Calculated as mortgage interest payments plus principal repayments as a proportion of nominal household post-tax income. Household income is defined as disposable (post-tax) income adjusted for changes in pension entitlements, which is adjusted to exclude gross operating surplus and the effects of financial intermediation services indirectly measured, and to add back interest paid. Mortgage interest payments before 2000 are adjusted to remove the effect of mortgage interest relief at source.
• (b) The illustrative projections to end-2028 use projections for household post‑tax income consistent with Scenario C in the April 2026 MPR forecast. Payment increases are projected using market expectations for Bank Rate based on the overnight index swap (OIS) curve over the 15 days to 22 April 2026 taking into account the distribution of fixed-deal terms from the FCA Product Sales Data and assuming the aggregate mortgage debt to income ratio remains constant.

Vulnerabilities from higher energy prices and borrowing costs are distributed unevenly across households. Those with high debt burdens and lower incomes are likely to be impacted disproportionately.

Although these projections suggest households should remain resilient in aggregate, vulnerabilities remain. Households with higher debt servicing burdens relative to income are more likely to struggle with debt repayments alongside higher spending on energy and other essential goods. These households are more likely to cut back sharply on their consumption, potentially amplifying any potential economic slowdown resulting from the Middle East conflict. The share of households with high COLA DSRs – defined as households which spend over 70% of their income after taxes and essential spending on debt repayments – increased slightly in 2026 Q1 to 1.6%, but remains close to the average level over the last ten years. Under Scenario C from the April MPR, this share would increase marginally further to 1.8% but this remains around half the level seen in the run-up to the GFC.

Lower income households, including renters, are likely to be more exposed to higher energy prices. They spend a larger share of their income on essentials, limiting their ability to adjust spending in response to higher prices. These households have been under pressure lately. In the NMG household survey, the share of households reporting experience of financial difficulty has risen since 2022, and in the latest survey (2026 H1) this share increased slightly to 30%, up from 29% in 2025 H2. However, these households typically hold a smaller share of total outstanding consumer credit and mortgage debt (the lowest income quintile of households hold 8% and 3% of these types of debt respectively) and are a relatively small proportion of aggregate consumption, which limits the potential financial stability implications.^{footnote [35]}

While pressures may build, current signs of distress are low and there are limited signs of pressure building in recent months.

The share of mortgages with more than 2.5% of their balance in arrears is 0.9%, close to long-run averages. Arrears on consumer credit (such as credit cards) have risen a little over the past two years but remain low by historical standards, also at around 1%. Although these measures are lagged and may not capture any increase in distress levels since the onset of the Middle East conflict, market and supervisory intelligence suggests that this picture has continued, with limited signs of distress among mortgage and consumer credit books to date.

Lenders continue to adjust their behaviour following recent policy changes to the FPC’s LTI recommendation.

The FPC's loan to income (LTI) flow limit, alongside the FCA's responsible lending rules, continues to provide appropriate protection against a material and unsustainable increase in household indebtedness and the number of highly indebted households. Following the FPC’s updated Recommendation in 2025 Q2, the share of high LTI lending increased to 13.2% in 2026 Q1, up from 11.8% in 2025 Q4. This leaves the four-quarter rolling average at 10.8%, below the 15% limit (Chart 5.4). Despite the continued increase in high LTI lending, mortgage approvals were broadly flat in Q1 and have been mixed in Q2 so far. Approvals increased in April as borrowers brought forward activity in response to the Middle East conflict, but approvals then fell to the lowest level since 2023 in May. Looking ahead, approvals may continue to be muted as weaker macroeconomic conditions and higher uncertainty weigh on mortgage demand. However, mortgage supply conditions continue to be strong (Section 6).

This increase in high LTI lending is supported by a number of factors, including strong credit supply and competition among lenders, the FCA’s clarification regarding its Mortgage Conduct of Business rules on the rates borrowers are stressed against to test affordability, and a number of lenders taking up the Prudential Regulation Authority's (PRA’s) Modification by Consent to disapply their individual LTI flow limit.

The share of high LTI lending moving closer to the FPC’s aggregate limit of 15% shows the updated Recommendation working as intended. The FPC’s objective in relation to the LTI flow limit is to manage the build-up of vulnerabilities in the stock of highly-indebted households. The flow of high LTI lending in any one quarter will have a limited impact on the stock of lending. But if the aggregate high LTI share is persistently inconsistent with the 15% limit, then, as outlined in their consultation papers, the PRA and FCA will expect lenders above the limit to gradually reduce their high LTI lending until they reach 15%. The PRA and FCA have both consulted on changes to their rules and guidance in line with the updated FPC Recommendation, including how they would manage any instances where the aggregate flow looks persistently inconsistent with the 15% limit, and will publish responses in due course. The FPC will also continue to monitor the effect of the policy changes on the housing market.

Chart 5.4: The share of mortgage lending at high loan-to-income ratios has continued to increase in line with the FPC’s updated recommendation

The share of mortgage lending over the previous four quarters at ≥4.5 LTI ^{(a) (b) (c)}

• Sources: FCA Product Sales Data and Bank calculations.
• (a) The FPC’s flow limit applies on a four-quarter rolling average basis.
• (b) Range of lenders’ share of lending at greater than or equal to 4.5 LTI (four-quarter rolling average) constructed using the weighted 10th to 90th percentiles of firms’ use of their individual flow limits and shown from the introduction of the FPC’s flow limit in 2014 Q4.
• (c) The sample includes only all new mortgages for house purchases and external remortgages with a change in principal.

Overall, the FPC judges that, in aggregate, households remain resilient, even in a challenging external environment. The FPC judges that household debt vulnerabilities that could disproportionately amplify the impact of an adverse shock on the macroeconomy or the financial sector are not currently elevated.

UK household aggregate indebtedness remains low relative to historical averages and debt vulnerability metrics remain around their long-run averages, providing resilience against shocks. Although some vulnerable, low-income households remain more exposed, the FPC judges that aggregate household debt vulnerabilities are unlikely to lead to sharp reductions in consumption. The results of the 2025 Bank Capital Stress Test demonstrate the banking system’s capacity to continue supporting UK households, even under severe economic and financial conditions.

5.3: UK corporate debt vulnerabilities

UK corporate balance sheets are strong in aggregate, providing resilience against shocks.

Corporate balance sheets remain strong in aggregate with the gross debt to earnings ratio at around 275%, in line with the post-2000 average of 280%. The corporate net debt to earnings ratio (that accounts for readily available liquidity such as cash held by firms) is also relatively low by historical standards at close to 150% (versus a post-2000 average of 170%). In addition, debt vulnerability metrics – such as interest coverage ratios (Chart 5.5) – are around their long-run averages, providing resilience against shocks. Healthy aggregate corporate balance sheets provide resilience, allowing firms to draw down cash buffers or support additional borrowing to weather shocks.

While higher energy and borrowing costs could put additional pressure on debt servicing burdens, it would take a severe shock to energy prices and funding costs to return aggregate debt servicing burdens to previous peaks.

UK corporate debt-at-risk, calculated as the share of debt held by firms judged to be at higher risk of failure over the next three years as they breach three financial measures,^{footnote [36]} is currently in line with historical averages and has fallen post-Covid. Under Scenario C from the April MPR it is projected to increase modestly over the coming years as higher energy and other input costs weigh on corporate earnings and higher market interest rates increase debt servicing costs (Chart 5.5). Staff analysis finds that energy intensive sectors such as transport and manufacturing are most impacted by the rise in energy prices under this scenario.

One of the financial metrics within the debt-at-risk measure is interest coverage ratios (ICRs). Firms with an ICR below 1.5 (that is, a firm with earnings less than 1.5 times its annual interest expense) have been shown to be at higher risk of failure. The share of debt held by firms with an ICR below 1.5 is also projected to increase but remain within historical peaks. It would take a significant further shock to borrowing costs or earnings to increase debt servicing pressures close to historical peaks (Chart 5.5).

Chart 5.5: UK corporate debt-at-risk is projected to rise modestly in coming years. Interest coverage ratios are projected to increase but remain below historical peaks

Debt-weighted share of UK corporates with ICRs below 1.5 and companies at higher risk of default. Solid lines give historical data. Dashed purple and aqua line show projections under Scenario C from the April MPR. Dashed orange line shows a projection under Scenario C plus an additional 300 basis point increase in borrowing costs ^{(a) (b) (c) (d)}

• Sources: Moody’s Analytics, Bureau van Dijk, and Bank calculations.
• (a) These data refer to UK private non-financial corporations only.
• (b) Data for 2025 is estimated by keeping earnings constant and assuming full pass-through of funding cost changes over 2025 to end-2024 balance sheets. The central case ICR projection assumes full pass-through of the Bank Rate path to the interest rate paid on the stock of floating and maturing fixed-rate corporate debt. The scenario with higher borrowing costs incorporates a further 300 basis points increase in the average cost of debt servicing for UK corporates on this stock of debt.
• (c) ICR projections contain energy prices consistent with Scenario C from the April 2026 MPR that are applied to earnings on a sectoral basis. While some sectors will experience negative earnings shocks, the scenario assumes positive revenue growth in aggregate.
• (d) The aqua line represents the debt-weighted share of UK corporates that simultaneously breach the three thresholds associated with the highest likelihood of firm failure: whether a company’s ICR, calculated by dividing its earnings before interest and tax, is below 1.5; whether its liquidity ratio (current ratio) is below 1.1; and whether its return on assets is negative. The sample includes UK large and medium firms.

SMEs and companies in sectors most exposed to higher energy prices are more vulnerable to further shocks in energy prices or funding costs.

The corporate sector may come under increased pressure from higher energy and input prices. If firms are unable to pass on higher costs of production to consumers or suppliers, for example where competitive pressures and weak demand constrain their pricing power, this will reduce their earnings. These effects will not fall evenly across businesses.

Sectors with large exposures to energy or fuel prices (such as manufacturing and transport) could face significant cost pressures, although there is limited evidence of distress building to date. Businesses in more labour-intensive sectors, such as accommodation and food, have already been under pressure from recent increases in labour costs associated with a higher national minimum wage and employer National Insurance Contributions. This sector may come under further pressure as, while it is not particularly energy intensive, a high proportion of spending is discretionary, and it may face lower demand if real income growth slows.

Small and Medium sized Enterprises (SMEs) are typically more exposed to higher energy prices than larger corporates; they often have lower margins to absorb cost increases, and less pricing power to pass them on to consumers. SMEs now have smaller cash buffers than in recent years as Covid support has now unwound, making it more difficult to support a squeeze in margins. However, they are now less indebted as some of this debt has since been repaid. SME lending has limited implications for financial system losses in aggregate as the size of SMEs’ debt is small relative to larger firms. However, SMEs account for around 60% of UK employment, and therefore SMEs taking defensive actions, such as cutting investment or employment can act to amplify any macroeconomic downturn to the wider economy.

Some highly leveraged corporate borrowers reliant on riskier credit markets are particularly exposed to tighter financing conditions. Aggregate near-term refinancing needs remain limited, although refinancing walls for riskier debt are steeper.

Higher market interest rates have put upward pressure on corporate borrowing costs. A deterioration in investor risk appetite, or sharp correction in credit spreads, could significantly increase refinancing costs and limit the ability of some corporates to refinance maturing debt, particularly those with high leverage or near-term maturities. Businesses which struggle to roll over financing may take defensive actions by cutting back on investment and employment. If such pressures persist, they could weigh on companies’ cash flows and resilience, increasing the likelihood of distress, including defaults. However, near-term refinancing needs for corporates are relatively limited in aggregate. Refinancing needs are slightly higher for companies in energy intensive sectors which may experience additional challenges if higher energy prices raise investor concerns about their earnings resilience.

The proportion of corporate debt being provided by non-bank financial institutions in risky credit markets, including private markets, has grown in recent years. While these markets bring significant benefits for corporates, they also give rise to potential financial stability risks, including the extensive use of leverage, complex structures and interconnections with other parts of the financial system (Section 4). The refinancing walls are somewhat steeper in these markets, with around 20% of riskier debt needing refinancing by the end of next year, increasing these borrowers’ exposure to potentially higher rates and weaker investor demand. Spreads tend to widen more sharply in riskier markets during periods of stress, increasing the risk that some highly leveraged firms are unable to refinance on sustainable terms. Borrowers looking to refinance private credit loans may especially struggle given investor sentiment in this sector has weakened over the last year.

Chart 5.6: Refinancing needs for corporates over the next 18 months are limited, although refinancing walls for riskier debt are steeper

Cumulative share of debt maturing by type ^{(a) (b) (c)}

• Sources: PitchBook, a Morningstar company, LSEG and Bank calculations.
• (a) The chart captures market-based finance debt issued in all currencies by UK corporates, both UK issuers with a UK parent and UK issuers with a non-UK parent. Annual estimates.
• (b) 2026 figures show the share of debt that is due to mature in the remainder of 2026.
• (c) The private credit and leveraged loans series presented in this chart differ somewhat from Chart C in Box A as (a) the private credit line is the UK aggregate data sourced from Pitchbook rather than data from firms participating in the PM SWES only; and (b) the bonds and leveraged loans series are based on data from LSEG matched with corporate balance sheet data from Companies House, and so will incorporate different underlying assumptions and data coverage to align with the broader analysis presented in this section compared to Box A. In addition, this chart excludes financial corporates, whereas Chart C in Box A include them.

Current signs of distress are low and there is limited evidence of growing distress in recent months, but the impact of the shock will not have fully fed through, and data is lagging.

The proportion of businesses in arrears on bank loans is around historical norms for both SMEs and larger businesses, and the rate declined in the most recent data. Insolvencies have also remained broadly flat over recent months and are around long-term averages, after declining from a peak observed post-Covid.

However, these data are lagged and unlikely to show signs of distress if any had built in recent months. Market and supervisory intelligence, as well as intelligence from the Bank’s agency network across the UK, supports this picture. The Bank’s agents report that the Middle East conflict is not creating widespread stress, but is shaping behaviour, costs and financial conditions in a way that could amplify existing vulnerabilities in at-risk sectors, such as hospitality and consumer facing sectors, or where supply chain issues persist such as construction. While the agents report the overall picture as similar to pre-conflict, there is some evidence of clearer strain at the margin, particularly among smaller and more leveraged firms.

Overall, the FPC judges that, in aggregate, corporates remain resilient, even in a challenging external environment. The FPC judges that corporate debt vulnerabilities that could disproportionately amplify the impact of an adverse shock on the macroeconomy or the financial sector are not currently elevated.

UK corporate aggregate indebtedness remains low relative to historical averages and debt vulnerability metrics remain around their long-run averages, providing resilience against shocks. Smaller, and more leveraged corporates, particularly those which are financed by riskier credit markets like private credit and leveraged loans, remain more exposed. However, the FPC judges that aggregate corporate debt vulnerabilities are unlikely to cause companies to sharply reduce investment or employment. The results of the 2025 Bank Capital Stress Test demonstrate the banking system’s capacity to continue supporting UK corporates, even under severe economic and financial conditions.

6: UK banking sector resilience

6.1: Recent developments in UK banks’ resilience

The UK banking system remains appropriately capitalised with high levels of liquidity.

Capital and liquidity positions are broadly unchanged since the December 2025 FSR (Table 6.A). In December, the FPC set out its initial assessment of capital requirements, updating its benchmark for the system-wide level of Tier 1 capital requirements and identified areas where the framework may warrant adjustment. As set out in Box D and the accompanying FSIF, the FPC has agreed a package of proposed changes to the capital framework that will help ensure that it is simpler, more effective, more proportionate, and better calibrated to the risks in today’s financial system.

UK banks’ direct exposures to the Middle East are limited, but they remain exposed to the broader macroeconomic effects of the conflict.

The FPC maintains its judgement that the UK banking system has the capacity to support households and businesses, even if economic and financial conditions were to be substantially worse than expected. Major UK banks’ direct exposures to the Middle East are limited (around 3% of total global credit exposures), with banks likely to be more impacted by indirect spillovers from the conflict to the global economy, including supply chain disruption and higher energy and other commodity prices. Banks are responding to higher risks through additional provisions.

In the 2025 Bank Capital Stress Test, major UK banks were resilient to a scenario with a significant increase in oil and gas prices, alongside a deep global recession.

In the 2025 Bank Capital Stress Test (BCST), banks were resilient to a severe global aggregate supply shock. The scenario featured a sharp increase in commodity and energy prices as well as severe supply chain disruptions, leading to a rise in inflation across advanced economies as well as severe recessions in the UK and global economies. This scenario is substantially more severe than the current outlook and scenarios explored by the MPC in their April 2026 Monetary Policy Report (Table 6.B).

Major UK banks continue to report robust earnings and valuations have recovered following a decline in March which was driven by uncertainties around the conflict in the Middle East and concerns about private credit exposures.

Major UK banks reported an aggregate underlying return on tangible equity (RoTE) of 15.3% in 2026 Q1, and consensus expectations are for aggregate RoTE to rise a little above its current level over the coming years. Aggregate net interest margins (NIMs) increased year-on-year, with UK-focused banks seeing support from structural hedge income. Banks earn structural hedge income by investing non-interest bearing liabilities and equity in fixed-interest bearing assets, which often include interest rate swaps. This can stabilise income through the interest rate cycle, slowing the impact of falling interest rates and the benefits from rising interest rates. Loan growth has continued.

The aggregate impairment rate for the major UK banks rose in 2026 Q1, partially driven by losses some banks experienced as a result of their exposures to the default of Market Financial Services Limited (Section 4), a non-bank mortgage lender. The asset quality of banks’ wider credit portfolios has remained broadly stable, with the share of loans experiencing a significant increase in credit risk declining since the December FSR and provision coverage remaining flat (Table 6.A).

Major UK banks have continued to return capital to shareholders through buybacks and dividends, totalling £7.7 billion in Q1^{footnote [37]}, in line with their two-year average of £8.3 billion. UK banks’ average price to tangible book ratios (PtTB) declined in March, driven by uncertainties around the conflict in the Middle East and concerns around banks’ private credit exposures. These have since recovered to around 1.7x, similar to levels seen at the time of the December FSR and are around the highest seen since the Global Financial Crisis (GFC). As set out in Box A of the June 2024 FSR, PtTB ratios are an indicator of expected future shareholder returns relative to their cost of equity (the compensation investors require for the perceived riskiness of those returns).

In March, the FCA announced an industry-wide redress scheme to compensate motor finance customers who were treated unfairly between 2007 and 2024. This has since been subject to legal challenge and the Upper Tribunal has made an order suspending parts of the scheme. Banks have continued to provision for potential redress costs.

6.2: UK banks’ provision of credit to households and businesses

Aggregate lending volumes have increased since the December FSR, with UK banks continuing to support the provision of credit to households and businesses despite a deterioration in the macroeconomic outlook and heightened uncertainty. There is no evidence of banks restricting lending to defend capital positions.

Aggregate gross lending in Q1 rose 7% compared with the same period last year, and 4% quarter-on-quarter. The quarterly twelve-month growth rate for aggregate net lending also rose in 2026 Q1, to 4.7%. Aggregate increases in gross and net lending have mainly been driven by increases in large corporates’ borrowing.

In April, the annual growth rate of net borrowing by large businesses and SMEs increased to 12.3% and 4.2% respectively. There does not appear to have been a significant tightening in corporate credit supply linked to the broader uncertainty related to the Middle East conflict – with the Bank’s Agents continuing to report credit availability for corporates as better than normal, and more so for medium and larger businesses. They also note rising competition, particularly for higher quality borrowers. In the Q2 Credit Conditions Survey (CCS), the availability of credit was reported to be unchanged for large corporates, and to have slightly decreased for small and medium sized businesses. For all sizes of businesses, the availability of credit was expected to be unchanged over the next three months.

The Bank’s Agents have reported that large corporates continue to demand bank loans for working capital and refinancing needs, but that borrowing to invest remains subdued. Agents reported SME demand to be muted given heightened uncertainty around the economic outlook. In the Q2 CCS, large corporate lending demand was reported to be unchanged and demand from small and medium sized businesses was reported to have decreased. Over the next three months, demand was expected to be unchanged for all businesses.

Although bank lending to corporates has increased over recent quarters, the stock of corporate credit relative to GDP has continued to decline since 2021 and remains below its pre-Covid average, particularly for SMEs. Some of this reflects an increase in the provision of credit by non-bank financial institutions (NBFIs), which banks have played a key role in facilitating (Section 6.3). For SMEs, as set out in Box B of the December 2025 FSR, this likely reflects a combination of weaker demand and supply-side frictions. In conversations with the Bank’s Agents, structural supply side frictions cited by SMEs include long application wait times (some over six months) and reduced access to relationship-based banking advice. A range of initiatives across government, regulators and industry are underway to reduce these frictions and improve access to finance. Several of these were outlined in the December 2025 FSR.

Mortgage lending increased by 8% quarter-on-quarter in Q1, supported by strong supply conditions and a rise in mortgage refinancing activity, as some borrowers brought forward refinancing to lock in cheaper interest rates ahead of higher-priced products coming onto the market. The share of high loan-to-income lending also increased further, reaching 13.2% in 2026 Q1 up from 11.8% in 2025 Q4, following recent policy changes that have supported mortgage availability (Section 5). Following the outbreak of the conflict in the Middle East, some lenders withdrew and repriced mortgage products in response to movements in interest rate swap rates used in mortgage pricing. Product availability has since recovered to around December levels. In the Q2 CCS, the availability of credit for house purchases was reported to be unchanged, and availability was expected to increase over the next three months. Demand was reported to have increased in Q2 2026, potentially due to the continued effect of borrowers bringing forward activity in response to the Middle East conflict. It was expected to decrease over the next three months, likely reflecting broader macroeconomic uncertainty and the pass-through of higher swap rates into quoted mortgage rates.

Chart 6.1: Aggregate lending volumes have continued to rise, with signs of strength in large corporate lending

UK monetary financial institutions’ gross lending to UK households and businesses, seasonally adjusted ^{(a) (b)}

• Sources: Bank of England and Bank calculations.
• (a) Lending to UK households and businesses by banks and building societies that have permission to accept deposits in the UK.
• (b) SMEs are defined as businesses with annual debit account turnover on the main business account up to £25 million. Large businesses are those with annual debit account turnover on the main business account of over £25 million.

The FPC has maintained the UK CCyB rate at its neutral setting of 2%.

The FPC sets the UK CCyB rate to help ensure that the UK banking system is better able to absorb shocks without an unwarranted restriction in essential services, such as the supply of credit, to the UK real economy. In making this decision it takes into account a number of principles, as set out in FPC’s approach to setting the countercyclical capital buffer – Policy Statement. While the global risk environment remains elevated, UK household and corporate aggregate indebtedness remains low relative to historical averages, and it would take severe shocks to both funding and earnings to put debt servicing under pressure for both households and corporates in aggregate. Past stress test results indicate the banking system could withstand a scenario substantially more severe than the current outlook, and there are not signs of banks restricting lending in order to defend capital positions.

The FPC has decided to maintain the UK CCyB rate at its neutral setting of 2%. Maintaining a neutral setting of the UK CCyB in the region of 2% should help to ensure that banks continue to have capacity to absorb unexpected future shocks without restricting lending in a counterproductive way.

The FPC will continue to monitor the evolution of financial conditions closely to ensure the setting of the CCyB remains appropriate.

6.3: Interlinkages with non-bank financial institutions

Banks’ interlinkages with NBFIs can support credit provision to the real economy, while also helping banks manage their credit risk and capital positions. However, these interactions can also create channels through which risks can be transmitted back to banks.

Banks’ interactions with NBFIs fall into two broad categories. First, as facilitators of NBFI activities, including as providers of financing, leverage and services. Second, as issuers of risk transfer products, which NBFIs invest in. Both of these types of interactions can transfer risks from banks to elsewhere in the financial system, where they may still affect banks through their interconnections with NBFIs.

Banks provide financing to NBFIs, which in turn support lending to businesses.

Banks are increasingly supporting non-bank lending to the real economy through structures such as asset-backed and warehouse lending. This includes banks taking senior exposures in securitisations backed by NBFI-originated loans. Major UK banks’ total NBFI assets (excluding derivatives) have continued to grow, increasing by 5% year-on-year. This provision of financing and leverage to NBFIs has helped enable them to grow and increase their lending, with private credit and non-bank syndicated leveraged loans now representing an increasing proportion of lending to UK corporates.

For banks, lending to NBFIs can be an efficient way to support financing to the real economy while managing their own balance sheet risks. Compared with lending directly to individual corporates, these exposures are often structured with features that can reduce expected losses to banks, including senior ranking, diversified collateral pools, margining, and other credit protections.

Access to non-bank finance can bring benefits for corporates, including more diversified funding sources, tailored financing solutions and, in some cases, faster execution than through traditional bank lending. NBFIs may be better placed to finance some corporates where banks have less risk appetite or where borrowers need more flexible forms of term financing. Private credit funds can offer higher leverage, bespoke loan terms and faster underwriting processes, which may suit highly leveraged, lower-rated or operationally complex businesses. Private markets can also provide services that banks often do not, such as venture capital, making them a more suitable funding source for some higher-risk borrowers, while banks continue to provide working capital, trade finance and other support.

Increasingly, banks provide financing to NBFIs indirectly through securitisation structures, which can involve greater complexity and opacity, and may therefore carry higher risks.

Banks’ lending to NBFIs can expose them to direct and indirect losses. Banks holding senior tranches of securitisations may have less visibility of the underlying collateral, which can increase opacity and mask concentration risk, potentially leading to higher-than-expected losses. The complexity of securitisation structures and involvement of multiple third parties may constrain effective due diligence despite regulatory requirements. It may also leave banks dependent on the underwriting standards set by the NBFIs that originate the loans, including potential exposure to fraud. Recent bank losses related to the failures of Market Financial Services Limited and Tricolor have illustrated how these structures can result in direct and indirect losses for banks (Section 4).

In addition, NBFIs are significant investors in banks’ securitisations. This can support banks’ ability to lend and manage their credit risk and capital positions but can also further increase interconnections between banks and NBFIs.

In the UK, securitisations are an important mechanism used by banks to transfer credit risk. Significant risk transfers (SRTs) allow banks to transfer significant credit risk on loan portfolios to third parties through securitisation, and to obtain capital relief on the associated exposures, subject to regulatory requirements being met.^{footnote [38]} UK banks’ use of SRTs has risen significantly recently, with annual SRT issuance more than doubling in aggregate since 2017. This has increased interconnections between NBFIs and banks and could, in a stress, create additional transmission channels as well as heighten the risk of a procyclical reduction in banks’ lending capacity (Box C).

The PRA continues to highlight the need for banks to strengthen their monitoring and risk management of their NBFI exposures.

The PRA has previously highlighted the need for banks to manage appropriately counterparty credit risk relating to their NBFI exposures.^{footnote [39]} In response to recent losses related to NBFI exposures, banks have been reconsidering their risk appetite for certain types of lending and reviewing existing exposures. Banks should continue to monitor the risks from direct and indirect NBFI exposures, including under stress scenarios where correlations and losses may shift outside historical norms. In particular, banks should continue to consider the robustness of collateral verification and disclosure practices and consider investor concentration, refinancing and sudden market shifts that could affect transaction replacement. The PRA is continuing to engage with banks on the appropriate use of SRTs and is testing thoroughly any proposed structures that appear to be more complex or less robust.

Annex: Macroprudential policy decisions

Each Recommendation or Direction has been given an identifier to ensure consistent referencing over time. For example, the identifier 17/Q2/1 refers to the first Recommendation made at the 2017 Q2 Committee meeting.

Outstanding FPC Recommendations and Directions (as at the date of the FPC’s meeting on 26 June 2026)

On 23 March 2023, the FPC made the Recommendation (23/Q1/2) that:

• The Pensions Regulator (TPR) should have the remit to take into account financial stability considerations on a continuing basis. This might be achieved, for example, by including a requirement to have regard to financial stability in its objectives, which should be given equal weight alongside other factors to which TPR is required to have regard. The FPC noted that in order to achieve this, TPR would need appropriate capacity and capability.

On 27 June 2025, the FPC made the Recommendation (25/Q2/1) that:

• The PRA and FCA should together (i) aim to ensure that the aggregate flow of new residential mortgages from mortgage lenders at loan-to-income ratios (LTIs) at or greater than 4.5 does not exceed 15% of total new residential mortgages, and (ii) allow individual lenders to increase their share of lending at such high LTIs while aiming to ensure the aggregate flow remained consistent with the limit of 15%. The FPC recognises that, in doing so, such high LTI lending by individual lenders could exceed 15% of their total number of new residential mortgages while the aggregate flow remains consistent with the 15% limit. The aggregate flow is calculated based on new residential mortgages extended by lenders which extend residential mortgage lending in excess of £150 million per annum.

Other FPC policy decisions which remain in place

The following text sets out previous FPC decisions, which remain in force, on the setting of its policy tools. The calibration of these tools is kept under review.

Countercyclical capital buffer (CCyB) rate

The FPC agreed to maintain the UK CCyB rate at 2% on 26 June 2026, unchanged from its 27 March 2026 meeting. This rate is reviewed on a quarterly basis. The UK has also reciprocated a number of foreign CCyB rate decisions – more details are available at The countercyclical capital buffer. Under PRA rules, foreign CCyB rates applying from 2016 onwards will be automatically reciprocated up to 2.5%.

Leverage ratio

In September 2021, the FPC finalised its review of the UK leverage ratio framework, and issued a Direction and Recommendation to implement the outcome of the review as set out in its October 2021 Record.

In October 2022, in line with its statutory obligations, the FPC completed its annual review of its Direction to the PRA. The FPC revoked its existing Direction to the PRA in relation to the leverage ratio regime, and issued a new Direction on the same terms as in September 2021 with the addition of discretion for the PRA to set additional conditions to the central bank reserves exclusion.

The full text of the FPC’s Direction to the PRA on the leverage ratio is set out in the Annex of the October 2022 Record, together with the original Recommendation (now implemented).

The PRA has published its approach to implementing this Direction and Recommendation.

Other FPC activities since the December 2025 FSR

Other FPC activities since the December 2025 FSR not included elsewhere in this FSR include:

• Supported continuing policy engagement with industry and other authorities to enhance the resilience of the gilt repo market.
• Welcomed the FCA’s Consultation Paper ‘Enhancing fund liquidity risk management’, published in December 2025.
• Welcomed the PRA and FCA consultation on the implementation of the LTI flow limit Recommendation published on 1 April 2026.
• Welcomed the PRA’s consultation on modernising the liquidity framework published on 17 March 2026.
• Supported the Bank’s and FCA’s initiatives to continue to monitor the adoption of AI by regulated firms, including by re-running their joint AI survey in 2026, continuing market intelligence gathering exercises, and supporting the evolution of firms’ AI risk management practices, including via engagement with industry.
• Agreed the key features of the Bank’s private markets system wide exploratory scenario, which was published on 19 June 2026.
• Welcomed the publication, on 22 June 2026, of the Bank’s Policy Statement and draft Code of Practice for systemic sterling-denominated stablecoins.
• Welcomed recent statements by HM Treasury and the FCA on their plans to enhance MMF resilience, in line with the 2021 Financial Stability Board policy proposals.

The Financial Policy Committee Record – April 2026, and Financial Policy Committee Record – July 2026 contain more detail on the activities of the FPC since the July Report.

Glossary

• Abbreviations

  AAM – Alternative Asset Managers.

  ABS – asset-backed securities.

  AI – artificial intelligence.

  AIFMD – Alternative Investment Fund Managers Directive.

  AISI – Artificial Intelligence Security institute.

  ALRB – Additional Leverage Ratio Buffer.

  API – Application programming interface.

  BCST – Bank Capital Stress Test.

  BDC – Business Development Company.

  BIS – Bank of International Settlements.

  CAPE – cyclically-adjusted price-to-earnings.

  CBEST – Cyber security banking supervision and evaluation testing.

  CCLB – Countercyclical Leverage Buffer.

  CCPs – central counterparties.

  CCS – Credit Conditions Survey.

  CCyB – countercyclical capital buffer.

  CET1 – Common Equity Tier 1.

  CLOs – collateralised loan obligations.

  CMBS – Commercial mortgage-backed security.

  CMORG – Cross Market Operational Resilience Group.

  CNRF – Contingent NBFI Repo Facility.

  COLA DSR – Cost of Living adjusted debt-servicing ratio.

  CORST – Cyber and Operational Resilience Stress Test.

  CTP – Critical Third Parties.

  DB – defined benefit.

  DC – defined contribution.

  DSR – debt-servicing ratio.

  DV01 – dollar value of a basis point.

  EMIR – European Market Infrastructure Regulation.

  ETF – Exchange-traded fund.

  FCA – Financial Conduct Authority.

  FMSB – Financial Markets Standards Board.

  FPC – Financial Policy Committee.

  FSB – Financial Stability Board.

  FSR – Financial Stability Report.

  FTSE – Financial Times Stock Exchange.

  G7 – Canada, France, Germany, Italy, Japan, the United Kingdom and the United States.

  GCHQ – Government communications headquarters.

  GDP – gross domestic product.

  GEMM – Gilt-edged market maker.

  GFC – global financial crisis.

  HM Treasury – His Majesty’s Treasury.

  ICE – Intercontinental Exchange.

  ICR – interest coverage ratio.

  IFRS – International Financial Reporting Standard.

  IG – investment grade.

  ILTR – Indexed Long-Term Repo.

  IMF – International Monetary Fund.

  IOSCO – International Organization of Securities Commissions.

  IPO – Initial public offering.

  ISIN – International Securities Identification Number.

  KOSPI – Korea Composite Stock Price Index.

  LDI – liability-driven investment.

  LLM – Large Language Model.

  LSEG – London Stock Exchange Group.

  LTI – loan to income.

  LTV – loan to value.

  MoU – Memorandum of Understanding.

  MBF – market-based finance.

  MCOB – Mortgage Conduct of Business.

  METR – Model Evaluation and Threat Research.

  MiFID – Markets in Financial Instruments Directive.

  MMF – money market fund.

  MPR – Monetary Policy Report.

  NAV – net asset value.

  NBFI – non-bank financial institution.

  NCSC – National Cyber Security Centre.

  NIC – National Insurance Contributions.

  NIM – net interest margin.

  OECD – Organisation for Economic Co-operation and Development.

  OEF – open-ended fund.

  OIS – overnight index swap.

  ONS – Office for National Statistics.

  O-SII – other systemically important institution.

  PC – private credit.

  PD – probability of default.

  PE – private equity.

  PRA – Prudential Regulation Authority.

  PTFs – Principal Trading Firms.

  PtTB – price to tangible book.

  PV01 – present value of a basis point.

  RoTE – return on tangible equity.

  RWA – risk-weighted asset.

  S&P – Standard & Poor’s.

  SFTR – Securities Financing Transactions Regulation.

  SME – small and medium-sized enterprise.

  SMMD – sterling money market data.

  SRT – Significant Risk Transfer.

  STR – Short-Term Repo.

  SWES – system-wide exploratory scenario.

  TAIEX – Taiwan Stock Exchange Capitalisation Weighted Index.

  TOPIX – Tokyo Stock Price index.

  WAM – weighted average maturity.