After a generation of cheap government borrowing, bond markets are starting to price fiscal weakness again, just as AI, defence and energy demand the biggest infrastructure spend in modern history.
Sovereign bond markets are starting to discriminate again. After a generation in which governments could borrow at near-zero cost, investors are beginning to charge more to lend to sovereigns with weaker balance sheets. The repricing is arriving at the worst possible moment, with AI infrastructure, defence and the energy transition demanding the largest nation-building spend of the modern era.
Escala chief investment officer Tracey McNaughton makes that case in an opinion piece for Eco News, and her frame ties together a string of otherwise separate 2026 policy items. Australia now has a published National AI Plan, a joint ministerial statement on expectations for AI data centres, an APS AI rollout plan and a formal response to the Senate Select Committee on Adopting AI. Taken alone, each describes a new spending commitment. Taken together, they are entries on the same shopping list that bond markets are now beginning to price.
Treasury Secretary Jenny Wilkinson's Address to the Australian Business Economists 2026 framed fiscal headroom as the binding constraint, the same point McNaughton presses in market terms: that sovereign borrowing is no longer effectively unlimited, and that countries with stretched balance sheets will pay more for it than those still running surpluses.
McNaughton's claim runs deeper than any single budget. She frames the current wave as the largest, most expensive nation-building program in history because it stacks traditional categories (defence, transport, transmission) with the new AI stack of data centres, compute and the grid that feeds them. The capital ask, on her reading, is not something any one government's balance sheet can absorb.
That is where Australia's PwC-tracked superannuation pool of roughly A$4.5 trillion enters the argument. The implicit bargain is that funds supply the long-dated, low-correlation cash flows governments cannot raise from balance sheets alone, and governments in turn supply the regulatory and policy backdrop that makes those flows credible. The asymmetry, on McNaughton's analysis, is that accountability and pricing power flow to whichever side holds the scarcer resource: cheap, patient capital.
Two caveats. McNaughton is an industry CIO on a niche outlet, and her "sovereign risk premium" is a thesis rather than a measured outcome. The Prosper Australia submission on the Tax Reform No. 1 Bill 2026 views the same constraint from the tax-base side, arguing that the fiscal pressure already evident in revenue is what the bond market is starting to discount. Both readings land on the same fact: cheap government money is ending, and the spending list is not getting shorter.
What changes for the reader is the lens. The National AI Plan, the data-centre statement, the APS rollout and the Senate committee response are no longer separate policy moments. They are items on a balance sheet that bond markets are beginning to differentiate by fiscal posture, with strong creditors still borrowing cheaply and stretched ones watching their cost of capital rise. The watch item is whether the next round of AI capex announcements, any slip in the superannuation savings rate and any widening of the sovereign spread turn McNaughton's column into the market's working assumption.