AI Capex Drives Strongest EM Earnings in 20 Years

🔑 Enhanced Key Takeaways

• •Hyperscaler AI capex is projected to reach $667 billion in 2026 alone, representing 62% growth year-over-year and exceeding 90% of cash flows—surpassing dotcom-era ratios[1]. This unprecedented scale is concentrated among five mega-cap firms (Amazon, Microsoft, Alphabet, Meta, Oracle) that collectively account for the bulk of global AI infrastructure investment[4].
• •AI monetization is transitioning from narrative to tangible revenue contribution, with OpenAI achieving $20 billion in annual recurring revenue by end-2025 (3x growth from 2024) and Anthropic's revenue run rate surpassing $9 billion in January 2026, validating scalable business models[3]. However, these pure-play AI vendor revenues represent only ~3% of projected 2026 hyperscaler capex, creating a significant revenue-to-investment gap[3].
• •Goldman Sachs analysts predict AI infrastructure capex growth will decelerate sharply in late 2026, with quarterly growth rates expected to slow from 75% (Q3 2025) to 49% (Q4 2025) and further to 25% by end-2026[5]. This deceleration poses valuation risks for 'picks and shovels' AI infrastructure providers, particularly semiconductor companies facing intensifying competition[1].
• •EM asset class dynamics show divergent capital flows: EM sovereign net supply is expected to fall to $28.5 billion in 2026 from $92 billion in 2025, while EM corporates face modest net financing of -$20 billion[2]. This contrasts with the massive AI capex concentration in developed-market mega-cap tech firms, potentially widening the investment disparity.

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline