Updated
Updated · CNBC · May 13
AI Hyperscaler 2026 Capex Estimates Double to $725 Billion After Q1 Guidance
Updated
Updated · CNBC · May 13

AI Hyperscaler 2026 Capex Estimates Double to $725 Billion After Q1 Guidance

3 articles · Updated · CNBC · May 13

Summary

  • $725 billion is BNP Paribas' updated estimate for AI hyperscaler capital spending in 2026, nearly double the $365 billion consensus from a year earlier after first-quarter earnings guidance.
  • Q1 updates pushed forecasts higher because hyperscalers are accelerating AI buildouts faster than operating cash flow, increasing external funding needs as compute demand drives power demand.
  • Evercore ISI sees spending closer to $800 billion, led by Alphabet, Microsoft, Meta, Amazon and Oracle, while UBS estimates $511 billion by 2030 just for generation capacity additions.
  • That spending wave is reinforcing bullish views on energy suppliers and grid equipment makers, with UBS saying natural gas and solar order books could stay effectively sold out.
  • The latest revisions extend a broader trend from earlier reports that hyperscalers are pouring hundreds of billions into AI infrastructure and shedding their traditional asset-light model.

Insights

Can Chinese-made chips power ByteDance's AI ambitions to rival its US counterparts?
With profits plummeting, is ByteDance's massive AI spending a winning bet or a fast track to financial ruin?