What: Economists Philipp Carlsson-Szlezak and Paul Swartz argue the AI capex boom poses a manageable macroeconomic risk today, sized at roughly $315 billion, or 1% of US GDP, once import-heavy spending is stripped out.

Why it is important: Retail leaders navigating AI investment decisions and board pressure can use this framework to separate genuine systemic risk from investor losses, rather than reacting to headline-grabbing multi-trillion-dollar figures.

Estimates of AI data-center capex now range from hundreds of billions to several trillion dollars, fueling fears of overcapacity and recession. The authors argue the more useful question is not when the bubble will pop, but what it entails and how it could damage the economy. Combining hyperscaler and private AI-lab spending yields roughly $630 billion in 2026, but around half goes to imported semiconductors, leaving a domestic GDP impact closer to $315 billion, or 1% of GDP, rising to 1.5% by 2028.

Three channels could transmit damage: a direct activity stop if capex halts, a wealth effect from an equity sell-off given household equity holdings near 30% of wealth, and a credit crunch if debt losses spread. Lasting structural damage requires losses to hit the banking system specifically, as in 2008. Today, hyperscalers are funding capex mainly through cashflow and non-bank debt, and banks have limited direct exposure, containing the systemic risk for now.

IADS Notes: The capex boom's macroeconomic linkages are already visible in retail practice. Retailers' growing reliance on hyperscaler infrastructure for AI deployment is illustrated by THG Ingenuity's launch of an AI stylist built on Google Cloud's Gemini platform (Retail Week, July 2026), a partnership tying measurable commercial gains directly to continued hyperscaler investment. On the demand side, an analysis of Canadian consumer spending (BCG, July 2026) found that headline growth is increasingly propped up by savings drawdowns, asset gains and borrowing rather than income, leaving middle- and lower-income households most exposed to tightening credit conditions — a fragility that would amplify any equity-market wealth effect from an AI-driven correction. On the leadership side, guidance on balancing board pressure for AI adoption against organisational readiness (BCG, September 2026) reinforces the case for engaging deliberately with the AI boom rather than sidestepping it, framing success around disciplined experimentation rather than headcount reduction.

The questions you should be asking about the AI bubble