The Reserve Bank of Australia has released its October 2026 Financial Stability Review, warning that one of the key drivers of future financial instability relies on the future of the boom in artificial intelligence and the infrastructure that drives it.
The report recognises that AI promises much, and that “a small group of AI-exposed technology and semiconductor firms” has been driving a lot of positive movement in equity markets over the last 12 months.
There are, however, caveats.
“While advances in AI adoption may support earnings growth across the corporate sector over time, current equity prices reflect very strong expectations regarding future adoption and productivity rates, revenue growth and profitability,” the RBA said.
“If these expectations are not realised, or if increasing competition reduces returns on AI-related investment, highly valued AI-exposed firms could be susceptible to sharp repricing.”
The AI investment boom is referenced 13 times in the first chapter alone, and even has an entire boxout dedicated to it, titled “Funding the AI investment boom”, which outlines concerns that AI firms are increasingly relying upon debt to fund their work.
“While advances in AI adoption may support earnings growth across the corporate sector over time, current equity prices reflect very strong expectations regarding future adoption and productivity rates, revenue growth and profitability,” the RBA said.
“If these expectations are not realised, or if increasing competition reduces returns on AI-related investment, highly valued AI-exposed firms could be susceptible to sharp repricing.”
Almost 30 per cent of investment-grade corporate bond issuance in the United States in the first half of this year came from technology-related firms, the RBA said, and this is expected to grow. By 2030, debt markets could “help fund part of as much as US$7.7 trillion in AI capital outlays”.
Larger AI firms, with other revenue streams such as e-commerce (Amazon comes to mind) are considered relatively safe, as are hardware manufacturers. However, firms in higher risk categories, such as data centre construction, are on shakier ground.
“For example, to secure access to power, data centre developers are increasingly obtaining large bank-sponsored letters of credit that guarantee the costs of grid and generation upgrades required to connect new facilities,” the RBA said.
“Similarly, developers and neocloud providers, lacking the balance sheet capacity to deliver large-scale AI infrastructure projects, can seek support from larger firms, such as residual-value or lease-payment guarantees, to facilitate the necessary funding.”
This is all fine if the investments actually lead to a return, but right now, that’s the billion-dollar question when we’re talking about an industry where the only person actually making money is Jensen Huang at NVIDIA.
The RBA flagged four key areas of risk should the AI bubble finally burst:
- Circular financing arrangements: AI firms increasingly use vendor financing, with cloud providers and chipmakers funding customers that then buy their infrastructure or products, creating circular revenue flows.
- Potential for overinvestment: Unprecedented AI infrastructure spending could leave debt-funded projects unable to generate enough revenue if demand or adoption falls short of expectations.
- Debt maturity and asset-life mismatches: Long-term debt is funding hardware that could become obsolete or depreciate rapidly, potentially reducing collateral value before loans mature.
- Locked-in input commitments: Long-term contracts for hardware, water and electricity secure supply but could leave AI firms with inflexible commitments if demand, technology or energy needs change.
“If capital expenditures grow in line with current market expectations, considerable external financing of AI-related investment would increase credit exposures across banks, bond markets, private credit and other institutional investors,” the RBA said.
“While long-term earnings projections assume broad adoption of AI technologies and strong revenues, the timing, scale, and distribution of these benefits remain uncertain.
Want to see more stories from trusted news sources?Make Cyber Daily a preferred news source on Google.
