Comparing data center financing structures, side by side: which one is right here.

Corporate credit lends against the sponsor; project finance lends against the project's own contracted cash flows inside a ring-fenced entity. Corporate credit is faster, cheaper to document and far more flexible, but it consumes the sponsor's capacity and puts every other project behind it at risk. Project finance is slower and heavier, and it requires an offtake contract strong enough to stand alone — but it isolates the project, and it is usually the only route for a sponsor whose balance sheet is smaller than the asset. The fork is decided by two things: whether the contracted revenue can carry the structure without the sponsor, and whether the sponsor can afford to have this asset on its own credit.

Owning compute means holding the risk that the equipment is worth very little at the end; leasing means paying someone else to hold that risk. Ownership wins where the operator has genuine confidence in sustained utilisation past the current contract, a route to redeploy or remarket hardware, and the balance-sheet capacity to carry it. Leasing wins where demand is contracted for a finite term, where the sponsor cannot price obsolescence, or where the compute has to stay separable from a long-dated infrastructure structure. The decision is not about which is cheaper in nominal terms — it is about who is best placed to carry the residual.

An operating lease transfers the use of equipment for part of its life and leaves the risks and rewards of ownership with the lessor. A finance lease transfers substantially all of them to the user, which makes it financed ownership in lease form. The practical consequences are the residual, the term, and how easily the user can walk away. An operating lease fits a finite, contracted demand where the user does not want technology risk; a finance lease fits a user that intends to keep the equipment for its productive life, or that cannot find anyone willing to underwrite a residual on it.

The end-of-term option is where residual risk becomes explicit. Under a fair-market-value option the user may buy at whatever the equipment is then worth, so the lessor keeps both the downside and the upside and the user keeps a genuine choice to return, extend or buy. Under a dollar-out option the user buys for a token amount, which means the payments amortised the whole asset and the user held the residual all along. FMV fits finite, contracted demand and a user that cannot price obsolescence; dollar-out fits a user that always intended to keep the equipment and is willing to be paid for saying so.

A sale-leaseback releases capital by selling the asset to a third party and leasing it back, which transfers ownership and the residual along with it. A refinancing releases capital by borrowing against an asset the sponsor keeps, so the residual, the upside and the control stay put. Sale-leaseback typically releases more, does not depend on the sponsor's credit, and suits an asset whose future value the sponsor would rather not carry. Refinancing suits an asset the sponsor wants to keep, a sponsor with credit capacity to use, and a case where operating freedom matters more than the extra proceeds.

Front-of-meter generation connects to the grid and sells into it; the data center buys delivered power as a customer of a utility or a supplier. Behind-the-meter generation sits on the customer's side of the revenue meter and serves the facility directly, so its only customer is the facility. The choice decides who is the utility's counterparty, who holds the interconnection position, and whose credit stands behind reliability. Front-of-meter keeps the facility a purchaser of a commodity and leaves the obligation to serve with a regulated party; behind-the-meter buys schedule certainty and independence at the price of owning an operating business whose credit is now the facility's own. Most large projects end up holding both, and the structuring work is deciding which layer each one funds.

Under a construction loan the developer borrows against the project, builds it, and carries completion risk — usually backed by its own guarantees and a fixed-price contract — then refinances or sells once the facility is operating. Under forward funding a long-term buyer or investor commits during the build and releases capital against progress, taking the asset on completion at a price agreed in advance. A construction loan keeps the developer's upside and its exposure; forward funding transfers much of both, and prices certainty at the cost of the development margin. The fork turns on whether the developer can carry a cost overrun, and on whether the end value is worth more to it than the certainty of a committed exit.

A single-asset structure finances one project inside its own perimeter, so a problem in it stays in it and the sponsor keeps every other asset free to be financed independently. A portfolio structure pools several assets, which smooths the cash flow and can support terms no single asset would earn — but it links them, so one underperforming asset can trigger consequences across all of them. Single-asset wins where the assets differ from one another, where they will be sold or refinanced at different times, or where one carries a risk the others should not inherit. Portfolio wins where the assets are genuinely similar, held for the same horizon, and where diversification is worth more than isolation.

A data-center ABS pools cash flows from a portfolio of facilities — usually in a master trust that can issue more notes over time, amortising on a soft-bullet with an anticipated repayment date — so its diversification smooths any single loss and its refinancing risk is spread across successive issuances. A data-center CMBS is a single asset or campus in one fixed pool, more concentrated and often more balloon-heavy. ABS wins for a platform building and recycling a growing, diversified portfolio; CMBS wins for a one-off financing of a single stabilised campus where one strong tenant is the whole story. The choice is really portfolio-versus-single-asset, expressed in the securitisation market.

Identical capacity supports very different structures depending on who signed for it. A hyperscaler offtake brings established credit, long tenor and a contract most capital already understands, at the cost of price, specification control and negotiating position. A neocloud offtake typically pays better and moves faster, but the credit is younger, the term is shorter, and the counterparty's own ability to pay depends on contracts it has not necessarily signed yet. Hyperscaler offtake fits a long-dated, ring-fenced structure; neocloud offtake fits a sponsor with credit capacity of its own, shorter capital, or credit support that closes the gap.