Bank for International Settlements: speeches

1. The question and the scope

At the ECB's conference on money markets, Pablo Hernández de Cos asks how large the lasting footprint of a monetary authority's assets and liabilities should be once crises have passed. These portfolios swelled after the 2008 crisis and the pandemic, and have been shrinking in the euro area since 2022 and in many other economies. He sets aside emergency episodes and the lower-bound problem, concentrating on the reserves that commercial banks hold at the central bank. Two forces set their level: the operating framework, which fixes supply, and banks' own demand. Because financial systems, institutions and objectives differ, countries can reasonably draw different conclusions.

2. Operating frameworks

Frameworks differ by where supply meets the demand curve. In a corridor or scarce-reserves system, supply sits on the sloping part and banks forgo a positive return when they park reserves overnight, as in Brazil, India or Korea. Ample or abundant systems, such as the euro area's, push the market rate toward or below the deposit rate. Tiered designs pay a lower rate above a quota; Switzerland uses one to keep trading alive despite plentiful reserves. The ideal opportunity cost is debated: a textbook rule favours cheap, plentiful reserves, yet a large central bank footprint can distort allocation and crowd out private markets, while abundant liquidity can head off forced asset sales and scrambles for cash. Both effects coexist, so the net balance is hard to judge.

3. Monetary control and market functioning

Every framework can steer the policy stance, but with high reserves the operational target stays closer to the policy rate, since swings in currency demand or government deposits need no offsetting forecasts; scarce systems must intervene, though deviations are only a few basis points and do not seem to spread to other short rates. Plentiful reserves also let a central bank add liquidity without disturbing rate control. As reserves become abundant, trading incentives fade and interbank activity can wither; in Sweden around half of counterparties have taken part in no money market segment since 2020. Tiering sustains trading. Views differ on how much interbank turnover matters, since secured markets stay busy and supervision can supply discipline, and a recent ECB study finds unsecured lending carries higher regulatory costs than collateralised lending, though Japan's unsecured market recovered after tiering.

4. Financial stability

Abundant-reserve systems supply liquidity in advance, scarce ones provide it after stress appears, often at penalty rates against strict collateral. A large reserve cushion helps banks absorb shocks, lowers reliance on frozen interbank lending, forced sales of other liquid assets and emergency facilities, and trims banks' holdings of government bonds, weakening the sovereign-bank link. In stress, fragmented markets can leave some banks short at day's end. Yet the system may adapt to abundance in ways that raise fragility.

5. What drives reserve demand

Demand has risen over the past decade. Liquidity rules matter in different ways: intraday requirements or supervisory pressure behave like reserve requirements, whereas the standard coverage ratio can be met with other liquid assets, so it shows little link to reserve levels; reserves merely take a larger share as their opportunity cost falls. Frictions raise the insurance value of reserves: selling other liquid assets takes time, runs have become faster as the 2023 turmoil showed, borrowing from the central bank can carry stigma or lack operational readiness, and non-bank institutions may amplify swings in stress.

6. Endogenous effects and open questions

Demand may respond to the framework itself. When markets fragment, inactive banks act like autonomous factors whose payment flows move effective demand unexpectedly, and Swedish banks raised precautionary buffers as supply grew. High reserves may also encourage banks to take on more uninsured deposits and credit lines, leaving them more exposed if liquidity shrinks. Whether such effects can be undone as balance sheets fall is unresolved, though recent experience, including the euro area's, suggests a smooth reduction is achievable.

1. What deposit insurers do today

Speaking in Berlin to the international association of deposit insurers, Pablo Hernández de Cos recalls that the purpose of insurance is to protect less sophisticated savers and sustain public confidence, though a state backstop or central bank liquidity could also do part of that job. Industry funding remains central because it spreads failure costs among banks, and the old worry that insured banks take more risk has been answered by capital rules and resolution regimes that put losses on owners and creditors. Mandates have widened: over 80% of members now play some role in resolution that keeps deposits accessible rather than only paying out.

2. Funding failures

Banks' own loss-absorbing resources should be the first line, as international standards require for the largest banks. But resolution costs are unpredictable, loss-absorbing buffers can be depleted before failure and they do not suit every business model, notably smaller banks. Industry-funded schemes therefore stay vital, as the recent European reform widening their use in resolving small banks illustrates.

3. Three sources of new risk

Crisis tools were built around bank solvency and liquidity stress, so he sketches three areas to review. Risks can arise outside the regulated perimeter, as shown by the Archegos collapse and the 2022 UK gilt disruption, which calls for understanding how stress spreads and possibly access to central bank liquidity for some non-banks under proper oversight. Technology makes runs faster, as the Silicon Valley Bank failure showed, and raises the odds of severe cyber or third-party outages for which authorities need a joined-up plan. Finally, new forms of money could matter if adopted widely.

4. Stablecoins and tokenised deposits

Protection exists because deposits fund credit and carry run risk, and because they are redeemable at par and dominate payments. Asset-backed stablecoins lack stable value and the credit link, so the core rationale is missing; yet a run could force fire sales and bank withdrawals, so safeguards on reserves, capital and holders' legal claims matter, and a 2025 review found major implementation gaps. Tokenised deposits stay tied to a bank and credit provision, so logically they should receive the same protection. Whatever is decided must be explained clearly to avoid public confusion, a task where deposit insurers have proven skills in raising awareness.

5. Conclusion

Over a quarter century deposit insurance has become a core part of the safety net and is moving closer to failure management. Because technology is advancing fast, crisis frameworks must change to handle new risks. Yet proven success against contagion does not mean every product with deposit-like features deserves cover; extension should depend on whether the product delivers a comparable benefit to society. The association has already updated its core principles and published work on digital products, business models and players, which should feed the day's discussions.

1. The question

In the SUERF annual lecture in Vienna, BIS General Manager Pablo Hernández de Cos asked how changes in public finances, market structure and technology will alter what central banks can and should do when the next financial crisis hits.

2. What recent crises taught

From the global financial crisis, the euro debt crisis and the pandemic, he draws three lessons. Central banks were decisive: standing and emergency liquidity, emergency loans and bond buying stopped fire sales and restored key funding markets. Credibility mattered as much as money: in several cases simply announcing purchases, or offering lending that banks barely used, was enough to calm markets. But each success exposed limits. Bond-buying programmes were hard to size and explain when nobody knew how long a shock would last, and some meant to fix temporary dysfunction turned into lasting stimulus, making exit harder. And lending directly to firms, markets or non-banks during the pandemic, beyond the usual banking circle, exposed central banks to credit risk and blurred the line with government support.

3. Three forces that complicate the next crisis

High public debt: with debt near post-war peaks and large deficits, sovereign bond markets react more to fiscal doubts, and it becomes harder to tell whether rising yields reflect fiscal concerns or a market malfunction; even a well-designed intervention may then look like help for governments. Non-banks: funds and other non-bank institutions now own the largest share of rich countries' government bonds, and highly leveraged traders like hedge funds have become central go-betweens in those markets. Their short-term funding and leverage can turn fiscal news into sudden sales, as in the dash for cash of March 2020 and the British pension-fund episode of 2022, creating what the BIS calls a fiscal-financial stability nexus. Since they usually cannot borrow from the central bank, support has to take the form of asset purchases, and widening access without comparable regulation would invite moral hazard. Digital innovation: online banking and social media made the 2023 deposit runs unusually fast and let rumours spread before official messages; stablecoins could transmit stress to money markets if they had to dump reserve assets; and widespread use of similar AI models may make market reactions more synchronised.

4. Responses he proposes

He calls for a layered answer. Governments should put debt on a credible path and build buffers in good times, which reduces both the need for central bank action and the suspicion that it serves fiscal ends. Regulators should keep refining bank rules, for example by assuming faster deposit outflows and making sure banks are operationally ready to use central bank facilities, and should apply similar rules to institutions posing similar risks, whatever their legal form; options include more central clearing of government bond and repo trades, with its own risks, and well-targeted minimum haircuts to curb leverage. Central banks should design backstops that are cheap enough in stress to avoid stigma but unattractive in normal times, keep market-repair purchases clearly separate from monetary stimulus, as the Bank of England did with its time-limited 2022 operation, and set exit conditions in advance for quantitative easing. Finally, he stresses international cooperation on standards and data for non-banks, central bank swap lines for dollar funding stress, and coordinated communication, with the BIS acting as a hub.

1. A world of repeated supply shocks

In an interview during a visit to Mumbai, BIS General Manager Pablo Hernández de Cos said recent central-bank strategy reviews share one conclusion: geopolitical tension and supply disruptions have become more probable. Over six years the world has absorbed the pandemic, the war in Ukraine, tariffs and now conflict in the Middle East. Such shocks push growth down and inflation up, so central banks must look at medium-run effects, stay focused on price stability and watch second-round effects on core inflation and expectations, while recognising they cannot solve everything alone. Growth has so far held up, helped by AI investment and exports. For resilience he sees a three-step approach: inventories in the short run, diversification of suppliers in the medium run, and technology, such as alternative energy sources, in the long run.

2. An AI boom that may not last

He warned that AI investment shows signs of becoming unsustainable, for two reasons: a race for market share breeds overinvestment, and spending increasingly relies on debt and leverage, including murky private lending vehicles and circular financing. A bust matters because AI has been a main pillar of recent resilience, supporting investment, wealth-driven consumption (notably in the US) and exports in some Asian economies; a reversal could hit all three, amplified by financial channels. On jobs, AI can complement or replace work depending on the task and sector, so retraining and active labour-market policies are essential, especially in developing economies.

3. Cooperation, cyber and regulation

On access to frontier technology, scale economies point to concentration in a few countries and firms, so global cooperation is needed to spread the benefits, and because financial stability is a global public good. AI-enabled cyber capabilities help defenders and attackers alike, and an incident can spread across borders, so coordination is the only effective response. He judged the bigger regulatory danger to be divergence between countries rather than slowness, and held up India's work on fast payments and data privacy as a good example.

4. Stablecoins and tokenisation

Stablecoins fall short of what money needs: exchange one-for-one with central-bank money cannot be assured when crises hit, and worries about integrity, fragmentation and interoperability remain. Because they are mostly dollar-based, wide adoption could threaten emerging economies' monetary sovereignty; the answers are international coordination on rules and, at home, good policy that builds trust. He backs two paths: tighter stablecoin regulation, and tokenisation inside the traditional two-tier system, with tokenised bank deposits and central-bank reserves. Project Agorá, on wholesale cross-border payments, now enters its second phase.

1. Demand, investment and trade

Speaking at the Global Fintech Fest in Mumbai, BIS General Manager Pablo Hernández de Cos organised his remarks around three questions. AI is already reshaping the global economy. On demand, it has triggered an investment boom in data centres, chips, cloud and hardware, and its optimism keeps financial conditions loose, supporting consumption through wealth effects. The five biggest tech firms alone plan over a trillion dollars of AI spending across 2025-26, and industry expects global AI investment to rise from about 500 billion dollars to 3-4 trillion by 2030, increasingly financed by debt and private credit. Trade is affected too: prices of AI-related goods have risen while other export prices stayed flat, so economies high in the supply chain, such as Taiwan, Korea, Malaysia and Singapore, gain, though narrowly (five Korean firms made 43% of export earnings), while heavy buyers of equipment lose terms of trade.

2. Productivity and jobs

On supply, task-level studies show productivity gains of 10 to 65% from generative AI, strongest for less experienced workers, and published estimates centre on roughly half a point a year added to productivity growth, with early evidence that better-prepared economies grew productivity faster. For work, AI both complements and substitutes. Informal workers in developing economies, with least protection and training, are most exposed; displacement so far is limited but emerging in customer support, programming and administration, and nearly 80% of firms discuss automation plans. His lesson is to prepare people through reskilling, not to resist the technology.

3. Divergence between economies

Advanced economies are likely to gain more in the short run because finance and professional services weigh more there and AI preparedness is higher. But emerging economies differ widely: China and parts of the Middle East are very well prepared, many low-income countries lag, and India, with its digital public infrastructure, has a real chance to close the gap. AI production is a five-layer chain (chips, cloud, data, foundation models, applications) with high fixed costs and scale economies, favouring a few giants in the US, China, Taiwan, Korea and the Netherlands. Smaller economies face a trade-off between building national capacity for resilience and wasteful duplication; many can focus on local applications, models for local languages and domain data.

4. Central banks and financial stability

The mandate is unchanged, but AI makes the economy harder to read, making it harder to pin down how much the economy can produce and where neutral interest rates sit. The BIS sets out three long-run scenarios: a bounded productivity boost, explosive self-improving growth, and a demand bottleneck in which displaced workers, as lost consumers, eventually starve the spending that supports innovation; in the last, the natural rate first rises, then falls below pre-AI levels, with inflation following. Financial-stability worries include cyber risk, which tilts towards attackers, and the investment boom itself: valuations are concentrated, spending exceeds cash flow, and circular financing among chipmakers, cloud firms and AI developers is opaque. An arms race risks over-investment, as in the canal, railway, electrification and dot-com episodes, and households now hold more equity, so a correction could hit consumption harder and spread globally. He does not say a bust is inevitable but urges caution. Benefits will depend on infrastructure, skills, competition, data governance and institutions; for central banks, the premium rises on measurement, judgment and cooperation.

1. What makes money trustworthy

Opening a panel at the Jackson Hole symposium, BIS General Manager Pablo Hernández de Cos framed the debate around the foundations of money: a common unit of account; singleness, meaning every instrument in that unit can be exchanged at par for central-bank money with finality; and elastic liquidity, supported by interoperability and financial integrity. Today's two-tier system delivers these through supervised intermediaries but still suffers patchy interoperability, limited competition and inefficient cross-border payments, which tokenisation and distributed ledgers could ease. The test for any innovation is whether it serves rather than erodes those foundations.

2. Stablecoins versus tokenised deposits

On singleness, a stablecoin holder paying someone who accepts a different coin must trade through markets where deviations from par are normal and larger under stress, while tokenised deposits settle in central-bank money and keep par. On interoperability, stablecoins sit on fragmented public blockchains, whereas tokenised deposits, with tokenised reserves as the settlement asset, can be fungible across banks. On integrity, most stablecoin balances sit in self-held wallets and more transfers bypass identity checks, complicating anti-money-laundering rules, whereas supervised, account-based deposits make this manageable. Hence today's stablecoins do not meet the standard of money accepted without questions.

3. Macro-financial effects

If stablecoins scale, reserve choices matter: holding wholesale deposits raises bank funding costs and tightens lending, holding short government bills drains banks' liquid assets, and holding central-bank reserves drains system liquidity, with smaller banks hit hardest. Extra demand for bills would lower short yields and ease fiscal space, so BIS scenarios show a modest net effect. Issuers remain exposed to runs, with fire sales able to strain money markets. Tokenised deposits keep deposit-taking tied to lending, but siloed adoption could favour large banks, and round-the-clock transfers could speed outflows.

4. A policy path

For stablecoins to work as money they would need enforceable par redemption backed by reserves and backstops, par exchange and finality across chains, and consistent anti-money-laundering control, including peer-to-peer transfers; some gaps are fixable by policy, others structural. Digital dollarisation also threatens monetary sovereignty. Tokenised deposits are promising but do not yet exist as interoperable multi-bank systems; challenges include avoiding walled gardens, governance, legal clarity on finality and smart contracts, resilience and migration. He proposes coexistence: tokenised deposits for most payments and settlement, stablecoins in narrow roles under strict regimes or as investment products, with central banks anchoring singleness, promoting interoperability and integrity, and watching credit effects. Project Agorá shows the approach is feasible. The BIS's annual report this year backs the same pragmatic course: fold tokenisation into the two-tier system anchored in central-bank money, and agree consistent international rules for stablecoin arrangements. He added that, in emerging economies, coins tied to foreign money bring dollarisation risks, and the strongest shield is good macroeconomic policy together with effective local payment networks.

1. The Annual Economic Report

Presenting the BIS flagship report at the yearly shareholders' meeting of the BIS in Basel, its General Manager framed the past year as a mix of progress, from rapid advances in artificial intelligence, and peril, from renewed inflation, financial fragilities linked partly to AI enthusiasm and heavy public debt. A second chapter studies how high debt and changing bond markets complicate central banking, and a third, presented by a colleague, argues that new digital forms of money only work if built on proven institutional foundations. The common message is that each policy area gains room to act when the others are disciplined.

2. Why the world absorbed the tariffs

A year earlier, many expected US tariffs to hit growth and trade hard. Three things limited the damage. Effective US tariff rates ended close to 10% on average rather than the more than 25% announced, thanks to deals and exemptions, which explains about a third of the later upgrades to 2025 growth. Firms adapted by shipping early, rerouting goods and absorbing part of the cost in their margins. And enthusiasm for AI triggered a wave of spending on data centres and related equipment, lifting US investment and Asian exports and keeping financial conditions easy despite uncertainty. By the end of 2025, world growth was almost back to its pre-tariff forecast.

3. A historic oil shock

The year began strongly until the Middle East conflict. Shutting the Hormuz strait removed a larger share of world oil supply than the crises of the 1970s, and growth forecasts were cut across countries. Markets stayed calm, betting on a quick settlement, and signs of a lasting reopening brought relief. He warned against complacency: production halted across the region will take time to restart, damaged facilities and shipping routes need repair, and rebuilding depleted oil stocks could keep demand and prices high.

4. Four pressure points

The report identifies four sources of strain. First, inflation: global inflation has risen by about one point since the conflict began, and higher prices for inputs such as plastics and fertilisers can keep pushing consumer prices up after oil calms down, while recent memories of high inflation could unsettle expectations faster than before; more slack in labour markets and higher starting interest rates than in 2021 help. Second, AI: task-level studies show big efficiency gains but economy-wide effects are still small, and sectors more exposed to AI in the United States already show weaker job growth. If AI lifts productivity durably, the neutral interest rate rises; if automation squeezes wages and spending, a shortage of demand could reverse that, with matching consequences for inflation if policy misjudges it. Investment may also overshoot as firms race for market share, and past technology booms, from canals and railways to the dotcom era, often ended abruptly. Third, financial vulnerabilities: risk premiums are compressed, households hold more shares than in the past, AI firms and their suppliers increasingly borrow, some financing is circular and opaque, and private credit is growing, so a disappointment or a rate rise could tighten conditions sharply. Fourth, public finances: debt in many economies is near its highest since 1945, conflicts add spending demands, and interest costs and refinancing needs are rising.

5. Debt and new bond market players

Debt grew through repeated crises, but also because governments loosen policy strongly in bad times and tighten too little in good times, and since 2008 they have reacted even less to rising debt. Without consolidation, ageing, infrastructure and defence costs will push debt higher still, and it would be unwise to test the limits. Bond markets have also changed: pension funds and insurers remain steady buyers, but leveraged hedge funds financed by short-term repo, often with little or no margin, now play a much larger role in US and European government bonds. Liquidity that looks abundant can disappear under stress. The old link between weak sovereigns and weak banks has faded as banks are stronger, but turmoil can now spread more quickly via repo funding, from one country to another and from hedge funds to the banks tied to them. BIS estimates suggest that the chance of severe Treasury market stress is roughly ten times higher when debt is in its top tenth than in its bottom tenth, and also much higher when non-banks hold a large share of government bonds.

6. Three challenges for central banks

First, when investors reprice fiscal risk, bond yields jump and currencies fall; activity weakens, and prices first rise through expectations and import costs before falling later, a pattern long familiar in emerging markets that may spread to advanced ones. Second, calibrating policy gets harder: with high debt, rate rises hurt output more through debt costs, wider spreads, bondholder losses and forced deleveraging, but may cut inflation less because interest income supports spending and expectations react less; euro area evidence shows smaller price declines after tightening in high-debt countries. Third, market dysfunction can force central banks to intervene, yet repeated rescues encourage leverage, blur the policy stance, weaken fiscal discipline and raise doubts about independence.

7. What policymakers should do

Central banks should watch for second-round effects, keep expectations anchored and choose strategies that hold up across many AI scenarios, putting medium-term price stability first even when this burdens public finances, since credibility lowers long-term rates. Independence is vital but not sufficient. Governments should consolidate in good times within credible medium-term frameworks, protect growth-enhancing investment and pursue reforms that spread AI's gains. Regulators should apply equal rules to equal risks whatever the type of institution, curb leverage, liquidity mismatches and fragile funding, and keep up with AI and cyber risks. Liquidity backstops should be temporary, targeted and reversible, like the Bank of England's 2022 intervention; lending directly to non-banks could replace asset purchases if well regulated. Acting now costs less than acting later.