US hyperscalers' euro thirst — lifeblood or vampire?: Mike Dolan
TLT•AI capex and borrowing needs continue to rise
What's certainly true is there's no sign the AI investment wave is cresting — with AI chip leader Nvidia's NVDA.O latest forecast of a 70% sales increase next year just the latest indication.
And many estimates put AI-related capital expenditure on data centers and other infrastructure at between $5 trillion and $7 trillion by 2030, after topping $1 trillion in 2026 alone. With free cash virtually gone for the hyperscalers, borrowing is set to balloon. Up to $400 billion of new debt is now expected for 2026 as a whole. AI-related debt now accounts for some 7% of U.S. investment-grade corporate bond indexes.
One concern about rising euro borrowing and reverse Yankees is that it could stymie or crowd out Europe's domestic funding for AI infrastructure.
As it stands, Morgan Stanley research last month on relative data-center development around the world showed Europe's buildout remains in its infancy. It estimated the region's data-center capacity grew 15% over the past year, compared with 26% in the U.S., and could accelerate at an annual rate of 20% through 2030. That would still be well below projected annual U.S. data-center growth of 30%, but on par with China.




