Meta disclosed the financing architecture behind Hyperion, its Louisiana data center complex, through a partnership with Entergy that commits $200 billion in total project value and $27 billion in near-term power infrastructure. The transaction marks the largest private infrastructure project in American history by dollar commitment, structured to deliver 5 gigawatts of AI compute capacity without expanding Meta's reported capital expenditure line beyond 2026 guidance.

Entergy will construct 10 gas-fired generation plants dedicated to the Hyperion site, with financing routed through Blue Owl Capital vehicles that keep the power builds off Meta's consolidated balance sheet. The arrangement allows Meta to secure long-term capacity without the front-loaded capex hit that spooked investors during Microsoft's nuclear restart announcement in September. Entergy takes construction and regulatory risk; Meta locks a 20-year power purchase agreement at rates disclosed only to Louisiana Public Service Commission staff under sealed filings. The structure matters because hyperscale operators now compete for gigawatt-scale blocks, not marginal grid access, and the financing innovation here—using infrastructure credit vehicles to separate power builds from technology capex—will be copied.

The $200 billion figure includes land, data halls, cooling systems, and the Entergy generation assets amortized across contract life, but the relevant number for allocators is the $27 billion Entergy will deploy in the next 36 months. That capital does not appear in Meta's 2025 or 2026 capex guidance, which stands at $60-65 billion annually, because the power purchase agreement is structured as an operating expense once the plants go live. This matters in two directions: it shows Meta can add massive compute without breaching the capex ceiling that analysts use to model free cash flow, and it proves that off-balance-sheet financing for AI infrastructure has moved from theoretical to executable at $10 billion-plus scale.

Louisiana offered tax incentives worth roughly $1.2 billion over 10 years and exempted the project from certain emissions escalators that would have triggered EPA review, according to documents filed with the state. The site sits on 3,200 acres in Iberville Parish, with Mississippi River access for cooling and barge delivery of turbine components. Entergy's gas plants will use combined-cycle units delivering roughly 500 MW each, with first power expected in Q2 2027. Meta's timeline calls for initial compute deployment in late 2027, with full 5 GW capacity online by 2030.

The financing structure through Blue Owl—a $235 billion AUM alternative credit manager—creates a template for other hyperscalers boxed in by balance-sheet constraints. Google, Amazon, and Microsoft all face the same problem: AI training clusters require dedicated, always-on power that grids cannot supply on demand, and building captive generation inflates capex beyond what public equity markets will tolerate. The Blue Owl vehicles allow institutional allocators to own the power assets, collect steady contracted returns, and let the tech company pay for electrons as an operating cost. Entergy's credit rating absorbs construction risk; Blue Owl's infrastructure funds absorb long-duration rate risk; Meta gets compute capacity without the capex spike.

Operators should track Entergy's Q1 2025 earnings call for updates on construction timelines and whether the Louisiana Public Service Commission approves the cost recovery mechanism that lets Entergy pass through fuel costs without additional regulatory lag. The Iberville Parish site permits are final, but air quality activists have challenged the EPA's delegated authority to Louisiana, and any federal intervention would delay the 2027 first-power target.watch whether Meta discloses similar structures for its other announced builds in Iowa and Tennessee, both of which face grid capacity constraints that require new generation.

Hyperion's structure converts a $200 billion infrastructure problem into a $27 billion financing solution that keeps Meta's reported capex within investor tolerance while locking the power AI clusters actually need. The innovation is not the data center; it is the separation of electrons from silicon on the balance sheet.