The $1.7 billion project investment that Industrial Development Funding and Oaktree are committing to Bloom Energy fuel cells is not a utility contract. It is a real estate capital formation event. The power plant is becoming a building system, and the building system is becoming a financed asset.
IDF and Oaktree are not buying electricity. They are buying the right to convert natural gas into electrons at the point of consumption, inside a data center campus that Nebius will use to run AI cloud workloads. The fuel cells sit behind the meter. That location changes the capital stack.
Behind-the-meter power means the energy asset is physically and financially integrated with the real estate. It is not a grid connection. It is a building improvement with a 20-year useful life, a predictable fuel input, and a single tenant offtaker. That is a real estate investment, not an energy trade.
The reported structure confirms the shift. IDF is the lead developer. Oaktree is a minority equity partner. Morgan Stanley placed the tax equity. MUFG Bank provided the senior debt. This is a project finance capital stack applied to a real estate use case. The tax equity monetizes the investment tax credit. The senior debt funds construction. The equity earns a return on the spread between the cost of gas and the price Nebius pays for reliable, dispatchable power.
The market signal is clear. Institutional capital is now willing to underwrite power infrastructure as a real estate investment because the tenant is an AI cloud platform with a compute load that cannot tolerate grid intermittency. Nebius selected Bloom for speed to power, clean technology, and the ability to support the performance and availability demands of AI workloads. Speed to power is the critical phrase. The grid cannot deliver new capacity fast enough. Behind-the-meter fuel cells can.
That timing constraint is the economic tension driving the deal. AI data center operators need power now. The grid interconnection queue is years long. Natural gas turbines face permitting and emissions opposition. Solar and wind are intermittent. Fuel cells offer a path that is fast, clean relative to diesel, and financeable because the technology is proven and the fuel supply is contracted.
The capital implication is that power availability is becoming a site selection criterion as important as land price, tax incentives, and fiber connectivity. Owners who control sites with gas interconnection capacity and the ability to host behind-the-meter generation will command a premium. Owners who rely on the grid alone will face a growing discount.
The lender behavior in this deal is worth watching. MUFG provided the senior debt. That is a bank making a construction loan on a power asset inside a data center. The underwriting is not based on the building shell. It is based on the power purchase agreement with Nebius, the technology track record of Bloom, and the tax equity structure that reduces the effective cost of capital. The loan is secured by the cash flow from the energy asset, not by the real estate. That is a new lending product.
The tax equity component is equally revealing. Morgan Stanley placed the tax equity. That means the investment tax credit for fuel cells is being monetized by a bank that understands the technology risk and the offtake credit. Tax equity is typically reserved for solar and wind projects. Its presence here signals that fuel cells have crossed the threshold into institutional financeability.
The open question is how this structure scales. A $1.7 billion commitment is large but not infinite. The total capital required to power the AI build-out over the next five years is measured in hundreds of billions. The grid cannot deliver it. Behind-the-meter gas generation can, but only if the fuel supply, the technology, and the offtake credit all align. That alignment exists today for a narrow set of projects with strong sponsors and creditworthy tenants. It does not exist for every data center proposal.
The reader consequence is practical. Owners and developers of data center sites should be testing their gas interconnection capacity now. Lenders should be building underwriting models for behind-the-meter power assets. Investors should be watching whether the tax equity market expands to cover more fuel cell capacity or remains constrained by the supply of bank appetite for technology risk.
The deal is not proof that every data center can be financed this way. It is proof that the capital markets have found a template for the ones that can. The next phase of the AI infrastructure build-out will be defined not by who owns the most land, but by who controls the cheapest, fastest, and most financeable power.