JPMorgan has raised its estimates for both power capacity and AI-related capital expenditures, while flagging strong year-to-date performance in high-yield data center debt.
JPMorgan analysts now project global AI and data center spending will hit at least $5 trillion by 2030, with a ceiling that could stretch to $7 trillion.
The capex surge is accelerating
The major hyperscalers, Meta, Alphabet, Microsoft, Amazon, and Oracle, are expected to spend $342 billion on capital expenditures in 2025 alone. That represents a 62% increase year over year.
JPMorgan estimates 122 gigawatts of new data center capacity is slated for deployment between 2026 and 2030. One gigawatt can power roughly 750,000 homes.
Data center construction itself has hit a $40 billion annualized rate by mid-2025, up 30% year over year.
JPMorgan describes the demand for computational power as “astronomical.”
Debt markets are eating it up
JPMorgan projects roughly $150 billion in leveraged finance and $1.5 trillion in investment-grade bonds tied to data centers and AI over the next five years.
A high-yield AI data center bond linked to Nvidia, issued in March 2026, was set at $3.8 billion. It drew approximately $14 billion in orders.
JPMorgan also projects that annual data center securitizations could reach $30 billion to $40 billion during 2026 and 2027.
What this means for investors
Companies that build and finance data centers are increasingly tapping credit markets rather than relying solely on equity or internal cash flows. The projected $150 billion in leveraged finance creates a new ecosystem of yield-generating instruments that didn’t meaningfully exist three years ago.
When $14 billion chases a $3.8 billion deal, spreads compress. That’s great if you already own the paper. It’s less exciting if you’re trying to get in at attractive levels going forward.
There’s also the power question. Adding 122 gigawatts of new capacity requires either massive investment in generation or a rethinking of grid allocation. Permitting delays, utility pushback, and environmental concerns could all slow the buildout in ways that financial models don’t fully capture.