More than $5 trillion in hyperscaler capital spending is projected between 2025 and 2030, and the overwhelming majority of that money will not come from corporate balance sheets. Instead, it will flow through off-balance-sheet structures such as joint ventures, project bonds, and private credit deals — a shift that gives institutional lenders like Pimco, Apollo, and BlackRock decisive influence over which energy sources power the AI buildout. Sustainability executives who track only megawatts and water gallons are missing the real decision point: the term sheet.
The numbers are staggering. The four largest hyperscalers — Amazon, Microsoft, Alphabet, and Meta — plan roughly $725 billion in capital investments for 2026 alone, up 77 percent from last year. Amazon’s free cash flow is expected to turn negative. When Meta raised its capex guidance, its stock dropped nearly 10 percent in a day. No balance sheet, however deep, can absorb that scale without strain. So the financing is moving to a project-finance model, where lenders fund the asset and the hyperscaler leases it back. Meta’s two-gigawatt Hyperion campus in Louisiana — larger in power demand than many American cities — was built through a joint venture with Blue Owl, financed by roughly $27 billion in bonds from Pimco, Apollo, and BlackRock. The debt does not appear on Meta’s books.
Within months, nearly $125 billion moved into similar structures, and total data-center debt issuance nearly doubled to $182 billion. This is now the template.
What does off-balance-sheet financing mean for the energy mix?
Speed is everything in this market. Lenders financing AI data centers prioritize “speed-to-power” — the fastest route to electrifying a campus. That structural bias favors on-site natural gas over slower clean-energy interconnections, which can take years for transmission upgrades and permitting. Short-term project-finance terms can discourage investment in renewable PPAs or long-duration storage. If the sustainability community continues to evaluate AI energy impacts only after the data center is operating, it will find that the most carbon-intensive choices were already locked in by the financing terms set months or years earlier.
- Meta’s Hyperion campus: 2 GW of demand, funded via $27 billion in off-balance-sheet bonds
- Total data-center debt issuance hit $182 billion, nearly doubling in under a year
- Hyperscaler 2026 capex expected at $725 billion, up 77% year-over-year
- Goldman Sachs projects over $5 trillion in combined capex from 2025 to 2030
Why should sustainability leaders care about term sheets?
The financial plumbing determines the environmental outcomes — not the procurement meeting. When a data center is financed through a joint venture with private credit, the lender’s return requirements and timeline directly shape decisions about power supply, water use, and location. A 20-year power-purchase agreement for wind or solar may not fit a 5-year project-finance horizon. The same logic that accelerated LNG export terminals and toll roads is now being applied to AI infrastructure: build fast, secure the cash flow, and let environmental compliance follow later. Sustainability leaders need to insert themselves into the finance discussion — interrogating not just energy suppliers but the capital structures that will dictate which electrons flow.
This also has implications for energy markets more broadly. Data centers financed off-balance-sheet can bypass traditional utility procurement processes, potentially disrupting grid planning and renewable portfolio standards. Regulators and utilities accustomed to negotiating with a single corporate credit are now facing special-purpose vehicles backed by Wall Street’s largest asset managers. The energy transition risks being defined by the speed of capital, not the depth of strategy.
Read the full report at GreenBiz.