The money underwriting the artificial intelligence boom is no longer just equity from cash-rich tech giants. It is increasingly debt, and a very large amount of it. According to reporting in The Australian Financial Review, a wave of AI-linked borrowing now measured in the trillions is starting to reach Australian shores, and it is forcing a reckoning about who ultimately carries the risk when the bill for all those graphics chips and data halls comes due.
The headline number, roughly 5.5 trillion US dollars, captures the scale of financing that global markets are marshalling to build out the physical backbone of AI. Data centres, power connections, cooling systems and the specialised silicon inside them do not come cheap, and the hyperscalers driving the buildout have begun leaning on bond markets, syndicated loans and private credit rather than funding everything from operating cash. What began as a story about a handful of American technology companies spending their own money has broadened into something that touches insurers, pension funds and credit investors across the world, Australia included.
How a US spending spree became a global debt story
For most of the past two years the AI capital cycle looked self-financed. Companies such as Microsoft, Alphabet, Amazon and Meta were generating enough free cash flow to bankroll enormous capital expenditure without stretching their balance sheets. That has changed as the ambitions have grown faster than the cash. Special-purpose vehicles, off-balance-sheet structures and joint ventures with private credit funds have become the vehicles of choice for financing new campuses, and the debt attached to them has ballooned accordingly.
The concern is not that AI is a fad. Few serious investors dispute that demand for computing power is real and rising. The worry is about the shape of the financing, the assumptions baked into it, and what happens if the revenue from AI services arrives more slowly than the interest payments. Chips depreciate quickly, power costs are volatile, and a data centre built for one generation of hardware can look dated within a few years. When you fund long-lived buildings with fast-moving technology using leverage, the margin for error narrows.
Two ways to read the wave
Optimists argue the debt is a sign of maturity rather than fragility. Infrastructure has always been funded with borrowed money, from toll roads to power stations, and lenders are pricing these deals with covenants, contracted revenue and creditworthy tenants behind them. On this view the arrival of institutional debt is exactly what you would expect as AI moves from a speculative bet to a genuine utility, and spreading the funding across bond markets and private credit is healthier than concentrating it on a few tech balance sheets.
The sceptics see echoes of past cycles where cheap financing outran the underlying economics. Their fear is circularity: chipmakers investing in AI start-ups that then commit to buy chips, vendors offering financing to customers who use it to buy their products, and valuations that only hold if growth stays vertical. If any link weakens, the losses do not stay contained in Silicon Valley. They flow through the credit funds, insurers and pension portfolios that bought the paper, and that is where the story stops being purely American.
Why this matters for Australia
Australia is exposed on several fronts, some obvious and some less so. The most direct link runs through superannuation. The nation’s retirement pool, now well north of four trillion dollars, has been steadily increasing its allocations to private credit and unlisted infrastructure, precisely the asset classes being used to fund AI campuses offshore. When a large industry fund buys into a global private credit vehicle, it may be taking on data centre debt without members ever seeing the acronym. That is not inherently reckless, but it does mean the health of the AI debt market is now quietly relevant to ordinary Australians’ balances.
The second link is the domestic data centre boom itself. Australia has become a favoured destination for hyperscale construction, with major projects across Sydney, Melbourne and increasingly regional centres, and ASX-listed operators have ridden the enthusiasm. Companies such as NextDC, Macquarie Technology, Megaport and the digital infrastructure names have raised capital and taken on debt to expand ahead of demand. If global sentiment towards AI financing sours, the cost and availability of that capital could tighten quickly, and local operators would feel it in their funding lines even if their own tenancies remain solid.
There is a third, structural concern that policymakers in Canberra have started to voice. The energy and grid demands of large data centres are significant, and financing them with leverage adds a layer of fragility to assets that are also central to the country’s digital sovereignty. A heavily indebted operator under stress is a less reliable custodian of critical infrastructure, which is one reason the intersection of AI, capital markets and national resilience keeps surfacing in official commentary.
What the local players and regulators are watching
For now, the mood among Australian institutions is watchful rather than alarmed. The Reserve Bank and the prudential regulator have both flagged the rapid growth of private credit as an area worth monitoring, given how much of it sits outside the traditional banking system and how little visibility investors sometimes have into the underlying loans. AI infrastructure debt is a natural test case for those concerns, combining a fashionable asset, opaque structures and a lot of capital chasing a limited number of deals.
Fund managers and analysts locally are urging clients to look through the marketing to the fundamentals: who the tenants are, whether revenue is genuinely contracted, how quickly the hardware depreciates, and what happens to returns if AI adoption plateaus. Those are not questions unique to Australia, but the answers matter here because so much retirement money is now channelled into exactly these vehicles.
What happens next
The near-term question is whether the buildout keeps accelerating or pauses to let revenue catch up. If AI products continue to gain commercial traction, the debt will look prudent and the wave will be remembered as the moment infrastructure finance embraced a new asset class. If adoption stalls or funding costs jump, the leverage that made the boom possible will amplify the pain, and Australian investors will discover just how connected their portfolios have become to a spending cycle largely decided in the United States.
Either way, the era of treating AI as a distant American story is over. The financing that powers it has gone global, and a meaningful share of the risk now sits, however indirectly, in Australian hands.
