The Context
The numbers are stark. Morgan Stanley forecasts global AI-related debt issuance will nearly double to USD 570 billion in 2026. Hyperscaler bond issuance exceeded USD 100 billion in 2025. This is not organic growth. It is one of the largest coordinated supply expansions in recent investment-grade credit history. The bank itself flags that the AI investment cycle is ‘modestly credit-negative in the near term’ until returns stabilise. USD 65 billion of corporate bond deals for data centres and AI investments have been led or co-led by Morgan Stanley since October 2025. This is a debt amplifier, not a profit engine. The market is being force-fed long-dated duration precisely when investors are nervous about rates and deficits. The audit trail shows a structural over-reliance on debt-funded AI capex by a small set of Big Tech hyperscalers. Their capital expenditure may consume nearly all operating cash flows by 2026, compared to a ten-year average of roughly 40%. The gap is being plugged by bonds.
The Risk
Why should a director care? The mathematics of liability is unforgiving. If your board is underwriting or issuing large volumes of AI-related bonds without robust disclosure of internal risk views, you may face securities-law claims for material omissions. The characterisations are now on record: ‘dangerously concentrated’ and ‘concentration without upside’. Directors have constructive notice that bond investors are being loaded into a single, high-risk theme with capped returns and asymmetric downside. Failing to adjust risk limits, portfolio guidelines, or underwriting standards could be attacked as a breach of the duty of care under corporate-law regimes. For New Zealand directors, the Companies Act 1993 imposes a duty of care, diligence, and skill. If your organisation is distributing or investing in these instruments, the governance gap is clear. The IOSCO report explicitly warns that AI use in capital markets can create concentration, outsourcing, and third-party dependency risks, potentially causing investor harm. Directors who fail to implement AI-specific risk-management frameworks, stress tests, and disclosure controls may be exposed to regulatory enforcement. The liability hook is not theoretical. It is numerical. It is auditable. It is now.
The Control
The strategic solution is not to stop AI investment. It is to stop funding it with blind leverage. Your board must demand a forensic audit of the debt structure. Map every dollar of AI-related bond issuance to projected cash flows. Stress-test the portfolio against a 200-basis-point spread widening. Require that internal risk views—including Morgan Stanley’s own warnings—are disclosed in offering documents. Establish a board-level AI risk committee with specific oversight of financing concentration and duration exposure. Ensure that any reliance on AI-driven models for pricing, allocation, or hedging is subject to human validation and documented governance. The control is a balance sheet that can survive a correction, not one that amplifies it.
The Challenge
These are the critical questions you should be raising at the board table:
| What is the total notional exposure of our balance sheet to AI-related debt instruments, and how does that figure compare to our tangible net worth? | |
| Have we independently stress-tested our AI-related bond portfolio against a 200-basis-point spread widening and a 30% decline in AI service demand, and what is the resulting capital impact? | |
| Do our offering documents for any AI-related bond issuance explicitly disclose the internal risk views—including the characterisation of this cycle as ‘modestly credit-negative’ and ‘dangerously concentrated’—and if not, why not? |