AI Bubble Could Trigger Global Crash, Warns Former Central Bank of Ireland Governor Honohan

AI exuberance: former Central Bank chief warns of a bubble with systemic risk

A former Central Bank of Ireland governor warns that AI enthusiasm is inflating into a bubble with the potential to trigger a wider financial shock. He argues that headline-grabbing tools can look like magic, but resemble the "little man" from The Wizard of Oz behind the spectacle.

The concern is simple: capital is rushing into AI faster than cash flows can validate. Even industry veterans are voicing doubt about the pace and scale of spend.

What's driving the bubble risk

• Record AI capex: hyperscalers, chipmakers, and data center operators are pulling forward spend on compute, power, and real estate.
• Narrow supply chains: dependence on a few GPU suppliers, fabs, and cloud platforms concentrates risk.
• Story-first valuations: revenue models lag hype; unit economics are unclear for many AI offerings.
• Private markets heat: venture and private credit are funding AI bets that may rely on future rounds or lofty exit marks.

How it transmits into finance

• Market drawdowns: AI-heavy indices and thematic ETFs can amplify a reversal.
• Credit stress: venture debt to pre-revenue AI firms, data center developers, and power projects faces refinance and covenant risk.
• Collateral fragility: equity, options, and even hardware-backed loans suffer if resale values or liquidity evaporate.
• Concentration exposures: lenders and asset managers often cluster around the same few names, vendors, and geographies.
• Derivatives and leverage: margin calls and volatility spikes can force procyclical selling.

Signals finance leaders should track

• Capex-to-revenue at hyperscalers and chip suppliers; watch for spending outrunning cash generation.
• Revenue per compute dollar: are customers paying for actual usage, or are credits/subsidies masking demand?
• GPU resale pricing and delivery lead times as a proxy for true scarcity vs. stockpiling.
• Data center lease rates, vacancy, and power constraints; delays can derail pro formas.
• Credit spread moves for AI-exposed issuers; look for early cracks in HY and private credit.
• Insider selling and lock-up expiries across AI-adjacent listings.

Practical risk actions for banks, asset managers, and corporates

• Run scenario tests: 40-60% equity drawdown in AI leaders, 200-400 bps HY spread widening, 20-30% drop in data center asset values, and capex deferrals by top cloud buyers.
• Map concentration: exposures by issuer, vendor, supply node (chips, power, cloud), region, and strategy (public equities, venture, private credit, real assets).
• Tighten terms: higher haircuts on equity/option collateral, shorter tenors, covenants linked to realized usage, not pipeline claims.
• Hedge selectively: skew, vol, and dispersion strategies; avoid crowded one-way bets.
• Liquidity buffers: model ETF outflows, prime broker margin changes, and LDI-style stress on collateral chains.

Underwriting guidance for lenders

• Underwrite to unit economics, not demos. Require proof of paid usage, churn data, and gross margin path after compute discounts end.
• Stress power pricing and grid delays for data center deals; include penalties for missed energization dates.
• Assess wrong-way risk with hyperscalers: counterparty, platform dependence, and contract termination clauses.
• Avoid residual value assumptions for specialized hardware unless backed by liquid secondary markets.

Capital allocation for corporates adopting AI

• Stage-gate investments: move from pilots to production only with measured ROI (cycle time cut, error rate, throughput, or working capital gains).
• Track full TCO: compute, orchestration, data cleaning, model drift, security, and talent.
• Vendor risk checks: portability across clouds/models, data ownership, and indemnities for content/IP.
• Governance: model validation, audit trails, and incident response playbooks before scaling.

If you need a curated view of practical AI software for finance teams, see this overview of tools and use cases: AI tools for finance.

Why the skepticism matters

Insiders flag that headline capabilities can hide limits. The metaphor is clear: impressive front-end, modest mechanics behind the curtain. That gap between perception and output is where bubbles form.

Context and references

Global regulators are mapping these risks. See the Financial Stability Board's view on systemic implications of AI and market structure, and the IMF's latest Global Financial Stability Report for macro risk context.

• FSB: Financial stability implications of AI
• IMF: Global Financial Stability Report

Bottom line for finance professionals

Price the story as optionality, not certainty. Tie funding to realized cash flows, cap concentrations, and prepare for a reversal that tests collateral and liquidity. If the air comes out quickly, you want exposure sized to survive the unwind-and the dry powder to buy quality at a discount.