Apollo Warns Prolonged AI Capex Without Returns Risks US Recession as China Competition Intensifies

Apollo's analysis highlights that AI-related capital expenditures have exceeded $200 billion annually among major US firms without corresponding revenue growth, creating a mismatch that echoes prior technology cycles. Token prices for key models have declined over 40 percent since mid-2024, pressuring margins for infrastructure providers. This dynamic diverts resources from sectors with faster payback periods, reducing aggregate demand if returns fail to materialize within 18-24 months.

US policy incentives, including CHIPS Act subsidies and export controls on advanced semiconductors, aim to secure AI dominance against China but raise input costs for domestic developers. Beijing's state-backed model training programs have accelerated catch-up efforts, documented in recent Ministry of Industry and Information Technology plans. The result is a bifurcated market where Western firms face higher financing costs while Chinese capacity expands under different capital allocation rules.

Market pricing currently embeds assumptions of rapid AI-driven productivity surges that historical data from automation waves do not support. Federal Reserve minutes from late 2024 note elevated uncertainty around technology diffusion timelines. If capex continues without measurable GDP contribution above 0.5 percent annually, corporate defaults in tech-exposed sectors could tighten lending standards and amplify downturn risks.

Next indicators to monitor include quarterly earnings guidance from hyperscalers and revisions to private nonresidential fixed investment figures from the Bureau of Economic Analysis. Sustained divergence between stated AI timelines and realized output would force portfolio reallocation away from growth narratives.