“We have a world economy that is standing on one leg on the AI leg. If this leg wobbles, that can lead to financial stability troubles.” – Kristalina Georgieva – International Monetary Fund CEO

The vulnerability lies in the financing structure surrounding artificial intelligence, not simply in the possibility that a model may disappoint users. Investment, corporate earnings, infrastructure spending and market valuations are becoming increasingly dependent on the assumption that AI will generate rapid and durable returns. The supplied report places this concern in the context of Kristalina Georgieva’s warning that a retreat from AI investment could weaken growth and create financial stability problems.1 That makes the issue broader than a technology-sector correction: it is a question of how many economic relationships now depend on one highly concentrated growth narrative.

Georgieva’s concern is sharpened by the scale and circularity of financing in the sector. The IMF figures cited in the report put circular financing among AI companies at slightly more than 1 trillion dollars, with around two-thirds funded through debt.1 In such arrangements, an investor may provide capital to an AI company, which then uses that money to purchase chips, cloud capacity or other services from businesses connected to the original investor. The transactions can be commercially rational, but they also create a network in which revenue, investment and credit exposure reinforce one another. If demand weakens, several balance sheets may deteriorate simultaneously rather than sequentially.

That mechanism matters because debt changes the consequences of a failed forecast. Equity investors can lose capital when valuations fall, but leveraged companies must still service loans, bonds and contractual commitments. The Bank for International Settlements has described the AI boom as shifting from reliance on internal cash flows towards debt financing, raising questions about credit standards, private credit opacity and systemic spillovers.5 A company that expects future computing demand to justify present borrowing may face pressure if customers delay purchases, model capabilities advance more slowly than expected or energy and infrastructure costs remain elevated. The result can be reduced investment, distressed asset sales and tighter lending conditions across related industries.

From technology enthusiasm to macroeconomic dependence

The phrase about an economy standing on one leg describes a concentration problem. Growth can benefit from productivity gains, new business formation and capital investment, but those benefits become fragile when one theme carries too much of the expansion burden. The reporting around Georgieva’s remarks says a sudden withdrawal of confidence from AI could undermine growth, while markets remain vulnerable because expectations may be running ahead of demonstrated capabilities.2,1 The danger is therefore not that AI has no value. It is that prices and investment plans may assume an unusually smooth path from technical progress to economy-wide cash generation.

Several feedback loops can intensify the adjustment. High valuations make it easier for companies to raise capital, and fresh capital supports more spending on processors, data centres and specialist staff. That spending produces orders for suppliers and helps sustain the appearance of broad demand. If investors revise their expectations, the process can reverse: falling share prices raise the cost of capital, weaker orders reduce supplier revenues, and lenders reassess borrowers whose assets or business models depend on continued expansion. Cross-border ownership can transmit those losses beyond the companies that originated the exposure, especially when investment vehicles, banks and private lenders hold similar positions.1

The risk is not limited to financing. The Financial Stability Board and BIS identify concentration among a small number of hardware, cloud and model providers as a separate vulnerability.4 Financial institutions may rely on common external services for fraud detection, credit assessment, trading support, customer interaction and operational processes. A disruption at a major provider, a cyberattack or a failure in a widely used model could therefore affect many institutions at once. Similar training data and model architectures may also encourage correlated decisions, creating a form of technological herd behaviour in which many firms respond to the same signals at the same time.7

Why a correction could spread quickly

AI can accelerate financial activity as well as automate it. Faster analysis and execution may improve liquidity in normal conditions, but the same speed can magnify feedback loops during stress. BIS analysis warns that common algorithms, limited private information and synchronised responses can rapidly affect asset prices and volatility.7 In practice, an unexpected earnings shortfall, model incident or cyber event could trigger simultaneous changes in risk limits, lending decisions and portfolio allocations. The system would not need every participant to make the same mistake; it would be enough for many participants to act on similar information using similar tools.

Opacity makes the problem harder to measure. Advanced models may be difficult to explain, validate or monitor, while exposure can sit outside traditional banks in private credit funds, technology suppliers and investment partnerships. The BIS has highlighted data gaps, third-party dependency, model risk, cyber vulnerabilities and market correlations as areas requiring closer surveillance.4 Regulators therefore face a practical dilemma: innovation moves quickly, but reliable information about adoption, contractual dependence and financial exposure often arrives later. By the time a concentration becomes visible, it may already be embedded in critical infrastructure.

There are substantial objections to a pessimistic reading. AI investment may produce genuine productivity improvements, lower operating costs and new sources of demand that justify current spending. Debt-funded infrastructure is not automatically reckless, particularly when borrowers have strong cash flows, long-term contracts and diversified customers. A correction in technology shares would also differ from a banking crisis if losses remained concentrated among well-capitalised investors. Treating every valuation decline as a systemic event could encourage unnecessary restrictions and deny economies the benefits of a major general-purpose technology.

Those qualifications do not remove the policy challenge. The sensible response is not to suppress investment but to test the links that could transmit a shock. Supervisors need better disclosure of AI-related borrowing, customer concentration, cloud dependencies and off-balance-sheet commitments. Banks and asset managers should conduct scenarios involving a sharp fall in AI valuations, interrupted computing supply, a major model failure and reduced demand for data-centre capacity. Authorities also need cross-border coordination because capital, infrastructure and service providers operate across jurisdictions. The FSB has recommended closing data gaps, reviewing supervisory frameworks and strengthening cooperation between regulators and industry.4

The deeper lesson is about resilience. AI may become a powerful source of productivity, but productivity gains cannot substitute for diversified growth, credible risk controls and sound balance sheets. Georgieva’s warning directs attention to the distance between technological promise and financial dependence: the more capital committed on the assumption of uninterrupted progress, the more damaging a modest disappointment can become. The objective is not to decide whether AI is transformative. It is to ensure that an economic system capable of benefiting from transformation can also absorb delays, failures and revisions to expectations without turning one sector’s wobble into a broader stability crisis.

 

References

1. https://www.aa.com.tr/en/economy/global-economy-standing-on-one-leg-as-ai-financing-risks-mount-imf-chief/4126372 – https://www.aa.com.tr/en/economy/global-economy-standing-on-one-leg-as-ai-financing-risks-mount-imf-chief/4126372

2. ‘Standing on one leg’: IMF chief warns about global economy as AI circular financing hits $1 trillion – 2026-11-16 – https://www.cnbc.com/2026/11/16/imf-georgieva-warning-ai-circular-financing.html

3. IMF’s Georgieva: World economy ‘standing on one leg’, and that leg is AI – 2026-11-16 – https://www.cnbc.com/video/2026/11/16/imfs-georgieva-world-economy-standing-on-one-leg-and-that-leg-is-ai.html

4. Financial stability implications of artificial intelligence – 2025-06-26 – https://www.bis.org/publications/fsi-summary-financial-stability-implications-artificial-intelligence-executive-summary

5. Financing the AI boom: from cash flows to debt – https://www.bis.org/publications/bulletin-120-financing-ai-boom-cash-flows-debt.pdf

6. AI and financial stability – https://www.bis.org/events/20251209/technological-innovations-financial-markets-risks-and-opportunities-banking-and-regulation/251209-tiifm-cfp-danielsson-presentation.pdf

7. Artificial intelligence and financial stability – navigating risks … – 2025-11-10 – https://www.bis.org/speeches/20251110-artificial-intelligence-and-financial-stability-navigating-risks-and-opportunities

8. The foundations of trustworthy AI in the financial sector – 2025-02-11 – https://www.bis.org/speeches/20250211-foundations-trustworthy-ai-financial-sector

9. tnmea2025016.pdf.md – https://www.imf.org/-/media/files/publications/tnm/2025/english/tnmea2025016.pdf.md

10. The World Can’t Afford to Get AI Wrong – 2026-10-15 – https://www.imf.org/en/blogs/articles/2026/10/15/the-world-cant-afford-to-get-ai-wrong

 

Global Advisors | Quantified Strategy Consulting
error: Content is protected !!