Page references are to the report's own page numbers. Chapter 1 pages are numbered from 1; Chapters 2 and 3 use the report's running page numbers.
Where markets stand
Markets have held up. Growth has been steady and company earnings strong, especially among AI-related firms (Ch. 1, p. 1). Interest rates have crept up rather than jumped: 14 of 33 major central banks have hiked since the second quarter, because inflation, though higher after the war's energy shock, has been "more contained than initially feared". Financial conditions tightened somewhat early in the year, but rising company valuations more than offset the effect of higher rates. The exception is emerging markets outside China, where conditions have tightened this year.
The IMF stresses that calm markets sit on rising risks and unresolved weaknesses. It says fiscal and monetary room to deal with a crisis is more limited than during most of the past two decades (p. 2).
Risk 1: a correction in AI-related assets
AI has become a main support for the global economy and for markets: spending on data centres, chips and infrastructure has helped offset the war's drag (Ch. 1, p. 3). Valuation pressure has eased in several markets as earnings have caught up. But weaker or shorter-lived AI investment and profits could trigger a correction, and there is growing evidence AI can be delivered with less investment than assumed. The IMF's World Economic Outlook scenario of an AI unwinding cuts global output by 0.4%, investment by 1.3% and consumption by 0.6% by 2028.
Leverage could magnify it, through three channels (pp. 3-4). Hyperscalers, the largest cloud and AI firms, now supply about 10% of global investment-grade bond issuance, up from under 1%, as their free cash flow falls. Hedge funds' gross equity positions plus equity margin loans have risen from about 10% to about 15% of advanced-economy output. And single stocks make up 28% of leveraged exchange-traded funds, up from 16%, as retail investors use them to get geared exposure to AI shares.
There is also "circular financing". Fourteen disclosed deals since 2023 involve about $300 billion of equity and $650 billion of debt from large AI firms to financially weaker ones, in return for $1.4 trillion of future revenue commitments such as chip orders or services (p. 4). If a key investee firm struggles, both the investment and the revenue it was meant to generate lose value at once: a "wrong-way risk". Advanced Asian equity markets are the most sensitive to global AI returns, while the euro area, UK and most emerging markets are much less so.
Policy. The IMF recommends limits on equity margin financing and leveraged ETFs, tighter control of short-term options, sector capital requirements, concentration limits (including in prime brokerage) and capital buffers where AI exposures are material (p. 15).
Risk 2: a sharp rise in government bond yields
Ten-year yields have risen because investors expect higher short-term rates and because they demand more compensation for holding long bonds (the term premium) amid worries over persistent deficits, high debt and a heavy supply of new bonds. Term premiums are at the upper end of their historical range, especially for advanced economies. Large AI-related debt issues may also compete with government bonds for investor money (p. 5).
Some weak spots have eased: borrowing by hedge funds against government bonds, a sign of leveraged "basis trades", has partly unwound without stress, and liquidity has improved in the euro area and UK. But liquidity in Japan's government bond market has worsened, and rollover risk is high (p. 6). The IMF rates this risk as worse than in April, and says the impact could be severe because government bonds underpin so much of the financial system.
Policy. Credible fiscal consolidation comes first; then longer-dated debt to cut rollover risk, wider central clearing of bonds and repos, central bank liquidity facilities ready to use, and clear central bank communication that protects independence (p. 15).
Risk 3: cyber threats
Cyber risk is growing as finance digitises (p. 7). Three trends drive it: a wider attack surface from cloud services, third parties and real-time payments; reliance on a small number of critical technology providers, which creates common points of failure; and more capable attackers, including state-backed groups and ransomware gangs using frontier AI tools. In more than 70% of cases in 2026 so far, software providers had almost no time to fix a flaw before it was exploited, 20 points higher than a year ago. Regions such as Asia face shortages of cybersecurity staff. AI can help defend, but raises the urgency of basics: cyber hygiene, incident reporting, crisis preparation and oversight of critical third parties.
Banks, and the global stress test
The IMF's 2026 stress test uses a scenario that combines weaker-than-expected AI gains, an energy shock and sharply higher yields (short rates up 150 basis points and long rates up 300) (p. 8). Global growth falls about 3 points below baseline over the first two years, unemployment rises 3.5 points and inflation is over 4 points higher by 2027. Banks' aggregate CET1 capital ratio falls about 150 basis points to about 11.5%, "well above regulatory minimums". But banks representing 22% of global banking assets end up below a 7% CET1 ratio, against 18% in the 2025 test. In North America the rise is driven by market losses at a small number of large banks when bond yields jump; in Asia it reflects capital damage in oil- and gas-importing economies. Banks in most other regions, particularly emerging markets, lose less (p. 8-9). Banks look able to absorb a stand-alone shock to private assets, but losses at banks with large non-bank exposures could build if a severe downturn happened at the same time.
Households, emerging and frontier markets
Household debt-service ratios have improved and remain well below their post-2022 peak, though lower-income households could be hit hardest by higher rates, and a fall in house or share prices would reduce spending (p. 14). In emerging markets, risk premiums have returned to prewar levels, and frontier sovereigns accounted for nearly a fifth of emerging market sovereign bond supply in the first half of 2026. But spreads are unusually tight next to a 10-year US Treasury yield above 4.5%, a combination last seen in 2007 (p. 10). Several frontier borrowers rely on official lenders or on newer structured deals such as sovereign total return swaps, which can trigger margin calls just when markets are stressed and are poorly captured by standard debt statistics (p. 11).
Institutional investors and private credit
Pension funds and insurers hold more than a quarter of publicly traded corporate bonds in the US and euro area, so forced sales by them could spread stress widely. Their growing use of derivatives and private assets raises embedded leverage and creates contingent calls for cash, and the investment chain is concentrating around a few consultants and managers (p. 12). In private credit direct lending, default rates have fallen, partly because lenders amended and extended loans, but borrowers have added debt, often to pay dividends to their private equity owners. Median debt relative to earnings has risen since 2024, especially in software and services, leaving many reliant on sponsor support. The IMF says the channels linking banks, funds and borrowers "have yet to be tested" in a sharp tightening (p. 13).
Chapter 2: hedge funds and financial stability
Hedge funds hold about $13 trillion of gross assets, against $4 trillion in 2013, yet are only about 5% of non-bank financial intermediation (Ch. 2, pp. 22-23). Growth has been financed mainly by borrowing: the gap between gross and net assets has risen by nearly $6 trillion, and gross notional exposure including derivatives has tripled to more than $40 trillion. Prime brokerage and repo financing have grown by more than $5 trillion, and hedge funds' government bond exposures have more than doubled since 2022, about two-thirds of it in US Treasuries (p. 24). In calm markets they support liquidity and efficiency. In stress, leverage, fragile investor funding and crowded trades can amplify price falls, and losses can spill over to the large dealer banks that act as their prime brokers. Monitoring is hard because of data gaps on leverage, derivatives, counterparties and cross-border activity (Ch. 2, at a glance).
Policy. Close data gaps and run stress tests; use targeted rules for crowded trades and synchronised selling, including stronger collateral management; where risks are concentrated, apply risk-based leverage limits, capital or margin add-ons and exposure limits; strengthen counterparty risk management; keep credible crisis backstops and coordinate internationally.
Chapter 3: tokenization
Tokenization means representing financial assets on shared programmable ledgers. It could cut reconciliation work, improve transparency and enable automation, and its benefits and risks both grow with the number of assets and participants (Ch. 3, at a glance). Public markets for tokenized real-world assets, excluding repos, have reached about $65 billion, with at least $8 billion of tokenized bonds in private markets and tokenized repos trading about $303 billion a day (p. 46). The IMF identifies four constraints on growth: legal certainty, regulatory clarity, interoperability and the availability of settlement assets. Tokenized equity markets show 24/7 trading and fractional ownership, but are so far less liquid, more volatile and more fragmented than traditional markets. Tokenization can also add technology and infrastructure risks and new channels for leverage and liquidity mismatches. For now, the IMF judges systemic risks limited.
Policy. Adopt a technology-neutral approach. Give tokenized assets legal certainty, apply "same activity, same risk, same regulatory outcome", use policy sandboxes, promote interoperability and settle in safe forms of money. As the market grows, strengthen monitoring of operational and governance risks, manage the liquidity effects of instant settlement and collateral reuse, design circuit breakers, and watch links with traditional markets.