AI infrastructure is rewriting the cost of capital: 2 Expert Foresights

Forecasts for AI infrastructure capex keep climbing. Financial Times cites forecasts of global spending associated with AI data centres of approximately $470 billion in 2025 and approximately $620 billion in 2026 as investor attention remains fixed on compute, power and data centre construction.

For AI and fintech analysts, this aggregate means more for capital allocation than it does for technological storytelling. When markets favour long-lived infrastructure assets, it reprices duration, broadens the balance-sheet constraints and pushes second-order considerations into household financing and corporate funding thinking. To take that market thinking and apply it to actual decision-making, I consulted two finance professionals in the fintech space: mortgage expert Luke Patterson at Koalify and mortgage expert, Sen FP&A Analyst at Parikh Financial, Daniel Cañizares.

Mortgage will follow term premium, not policy headlines

The AI infrastructure wave acts like a duration trade. Data centres involve heavy up-front investment, long payback periods and consistent refinancing conditions more stable than the tempo of central bank policy. Thus, even if short-term rates move down at some point, the consistent need for long-dated funding keeps the long end of the curve solid which then flows through mortgage pricing via swap curves, banks’ wholesale funding and securitisation spreads.

According to Luke Patterson, a mortgage expert at Koalify: “People look at central bank moves and think mortgages move the next day. They don’t. Mortgages price in the entire funding stack. If long-dated capital remains in demand as the market funds data centre construction, the term premium remains sticky and mortgages have that stickiness too.”

For analysts, there’s an implication of behaviour seen in the gap between cuts expected versus actual behaviour of term structure. When the curve is steep or long yields do not want to fall, it suggests that lenders want to keep their margins wide. Says Patterson: “You need to operate on the assumption that the cost of money over five years will be higher than your base case. Borrowers must stress test their affordability based on a tougher rate, then choose a structure based on whether they want to keep their options open should the number crunches go against them.”

FP&A will treat AI like a balance-sheet cycle, with unit economics as the control knob

AI may get to the budget as a technology line item but it poses more risk as a cycle in cost of capital and cash conversion. Model computation and model use shift COGS, vendor pricing adjusts rapidly, and governance ripple adds recurring expense that ups the ante on exposure. Similarly, debt markets respond to the same macro conditions that bankroll AI development, tightening refinancing windows at inopportune times.

Daniel Cañizares, Senior FP&A at Parikh Financial: “AI belongs in capital allocation not in a catch-all opex bucket. It can come back to haunt you fast through compute, vendor concentration and rollout timing. The variance is pressure on liquidity. Finance has to model it just like it would any other cycle that puts pressure on capital.

Such a mentality shifts the mechanics of planning. Cañizares’ FP&A team becomes empowered by a clearly defined threshold of cost per unit of value produced, cash payback windows, and downside developments that assume wider spreads. Cañizares notes: “It’s a discipline because deployment ties to unit economics and expands gates on growth. If the use case does not pass the margin test or pushes cash conversion the wrong way, the plan must force redesign before scaling. It keeps the balance sheet safe when conditions tighten for funding.”

The analytical advantage for 2026

The AI surge is a critical theme for the market going into 2026 as it changes the status quo for capital—where it sits and how long it stays locked up. It’s no longer a question of the next headline for the mortgage market but rather a question of term premium and spread behavior. For corporates, advantage in FP&A comes from scenario building that treats AI like a funding sensitive cycle and uses unit economics as a brake. Analysts who connect the dots will read Phase 2 better: momentum in AI infrastructure keeps long-dated pricing firm which firms cascades to both household borrowing and corporate assumptions.