AI and the Next Growth Cycle: Follow the Money

Australia's business press is full of questions about whether AI will drive the next great growth cycle. It is a reasonable question. But it is also the wrong starting point for investors trying to position thoughtfully. The more useful question is not whether AI grows the economy. It is who finances that growth, who bears the risk when things go wrong, and where capital quietly accumulates as a result. The answers point to corners of the market that most commentary ignores entirely.

The Infrastructure Debt Behind the Hype

Training and running large AI models requires extraordinary amounts of physical infrastructure. Data centres are the obvious entry point, but the capital stack underneath them is what matters for investors. These facilities cost hundreds of millions to build, consume electricity at industrial scale, and are financed through a mix of corporate bonds, infrastructure debt, and increasingly, private credit. The growth in private credit as an asset class has been significant over the past decade, and AI-linked infrastructure is one of the cleaner drivers of continued demand for that capital.

For Australian investors with exposure to unlisted infrastructure funds or private debt managers, this is already part of the portfolio picture, even if the label on the fund says nothing about artificial intelligence. The connection is worth understanding.

Power Grids and the Energy Constraint

Data centres are, at their core, electricity-consuming machines. The International Energy Agency has flagged that data centre power demand could double globally by 2026. That pressure lands on electricity grids, and on the companies and governments that finance grid upgrades. In Australia, the national electricity market is already navigating a complex transition. Adding a new category of large, concentrated industrial load creates fresh demands on transmission infrastructure, dispatchable generation, and grid balancing services.

The second-order effect here runs through regulated utilities and network owners. Grid operators and transmission businesses typically earn returns tied to their regulated asset base. Significant new capital expenditure, driven by demand growth, tends to expand that base and the returns associated with it. Whether that translates to investor value depends heavily on regulatory decisions about how those costs are recovered, and from whom.

The AI story is not just about software margins. It is a capital expenditure supercycle, and the financing chain reaches into infrastructure debt, insurance, and sovereign bond markets.

Who Insures a $500 Million Data Centre?

Large physical assets require insurance, and data centres present a genuinely novel risk profile. They concentrate enormous economic value in single locations, depend on continuous power, and house proprietary technology that is difficult to value for replacement purposes. Cyber risk sits alongside physical risk. The insurance market for these assets is still maturing, and pricing reflects that uncertainty.

For investors in insurance-linked securities or in listed insurers with commercial property and technology exposure, the growth of AI infrastructure is a factor in the underlying risk pool. When a single facility represents hundreds of millions in insured value, concentration risk in the insurance book becomes a legitimate portfolio consideration. The broader question for the sector is whether premium income is growing fast enough to match the accumulation of exposure.

The Sovereign and Corporate Bond Connection

Governments are not sitting on the sidelines. The United States, European Union, Japan, and others have all announced industrial policy frameworks designed to secure domestic AI capability. That means public spending, which means bond issuance. Fiscal positions in several major economies were already stretched before AI infrastructure became a political priority. The additional borrowing requirement is not trivial, and it adds to the supply of sovereign debt hitting markets at a time when central banks are reducing their balance sheets.

For fixed income investors, this dynamic is part of the broader picture around term premia and long-end yields. More bond supply, meeting structurally higher inflation uncertainty, is a combination that bond markets are still digesting. Australian investors with international fixed income exposure are not immune to these pressures, even if local fiscal dynamics look different.

Labour, Productivity and the Consumer Stocks Question

The headline promise of AI is productivity. If it delivers, the chain of consequences runs through corporate margins, labour markets, and eventually consumer behaviour. Sectors with high labour intensity and routine task structures, think logistics, legal services, financial administration, and customer contact operations, face the most direct pressure. Whether that pressure manifests as margin expansion, workforce reduction, or re-skilling costs depends on regulation, union coverage, and the pace of adoption.

For investors in consumer-facing businesses, the question is not simply whether AI boosts efficiency. It is whether productivity gains flow to shareholders, workers, or customers through price competition. History suggests the answer varies significantly by industry structure and competitive intensity. Concentrated markets tend to retain more of the gain. Fragmented ones pass it through.

Concentration Risk in the Ecosystem Itself

One of the less-discussed features of the current AI build-out is how concentrated the supply chain is. A small number of chip designers, a handful of hyperscale cloud providers, and a limited pool of specialised engineers underpin much of the infrastructure. For global equity investors, particularly those holding broad technology index funds, this concentration is already embedded in portfolio weights. The top-heavy nature of major indices means that exposure to AI optimism, and AI disappointment, is higher than the diversification label might suggest.

• Semiconductor supply chains run through a very small number of manufacturers and one dominant designer of advanced chips.
• Cloud hyperscalers control the majority of the compute capacity available for AI workloads.
• Energy and water inputs for data centres create geographic concentration in specific grid and watershed regions.
• Regulatory risk is accumulating across multiple jurisdictions simultaneously, with outcomes that remain genuinely uncertain.

Risks Worth Naming

The investment case around AI infrastructure rests on assumptions that deserve scrutiny. Demand forecasts for compute power have been revised upward repeatedly, but they are still forecasts. Regulatory intervention, whether on data privacy, market concentration, or energy consumption, could reshape the economics of the build-out materially. The history of infrastructure investment cycles includes genuine booms and genuine busts, and the gap between physical asset investment and actual economic return has caught investors before.

There is also model risk in the AI systems themselves. If the productivity gains are slower or more uneven than expected, the capital deployed into supporting infrastructure will take longer to earn its return. Some of that capital is patient, held in long-duration infrastructure funds. Some of it is not.

PortLens Perspective

The AI conversation in Australian investment circles tends to focus on technology stocks and whether local companies can participate in the growth. That is a reasonable thing to think about. But the more durable portfolio questions sit further down the chain. How is the physical infrastructure financed, and does that flow through asset classes you already hold? How does the energy demand reshape regulated utility returns in your market? How does the insurance market price a new category of concentrated physical and cyber risk, and does that affect the commercial lines books of insurers in your portfolio? The growth cycle, if it materialises, will be financed, insured, and powered by ecosystems that most AI commentary does not reach. What is the second-order investment implication that most people aren't talking about: if AI's energy appetite forces a faster-than-expected repricing of electricity infrastructure assets, which parts of the Australian fixed income and unlisted infrastructure market absorb that shock first?