Infrastructure Funds and What Makes a Deal Financeable

Infrastructure funds are the other great pool of capital reshaping the sector, and they operate on the equity side rather than the debt side. These are enormous vehicles raised from pensions, sovereign wealth funds, and insurers to own long-lived physical assets, roads, ports, pipelines, power, and now data centers, for their steady, inflation-resistant, contracted cash flows. Their return hurdle is lower than a private-equity fund's because they are buying stability, not growth, and that low hurdle is precisely why they can pay full prices for stabilized data centers and still be satisfied. When an infrastructure fund supplies the equity, a project can carry more debt at a lower blended cost, because the equity partner is patient and cheap.

The pairing of infrastructure-fund equity with private-credit debt is the dominant capital structure of the current build-out. The fund puts up patient equity, a private lender supplies flexible debt, and together they fund a campus that neither would build alone. The fund gets its stable yield, the lender gets its secured return, the developer gets its capacity built, and the hyperscaler gets its compute. This is the machine assembling AI infrastructure at scale, and understanding it means understanding that most of the money is deliberately conservative, seeking contracted, bond-like returns rather than venture-style upside.

That conservatism is the discipline that keeps the sector sound, and its absence is where deals go wrong. The central danger is over-gearing: piling too much debt onto an asset relative to how secure its cash flow really is. An over-geared deal looks brilliant while everything goes right, because the thin equity earns an enormous return, but it has no margin for error. A modest shortfall in rent, a rise in refinancing rates, a tenant that fails to renew, and the coverage ratio breaks, the covenants trigger, and the sponsor can lose the asset. The failures in infrastructure finance almost always trace to the same root: debt sized for the optimistic case rather than the realistic one.

So the practical question that decides everything is what makes a deal financeable. Reduced to essentials, a data-center financing is sound when a few conditions hold together. The tenant is creditworthy and the lease is long, so the income is secure for the life of the debt. The power is contracted and the interconnection is real, because a building that cannot be energized cannot earn, and a signed power agreement is now as load-bearing as the lease. The debt is sized to the cash flow with genuine coverage and a real equity cushion, not stretched to the last dollar the appraisal allows. The construction plan is credible with fixed costs and completion protection. And the refinancing is de-risked, with the takeout arranged or the rate hedged before the risky phase begins. When those line up, the deal finances easily and cheaply. When one is missing, the structure is fragile no matter how attractive the headline returns.

The mirror image, the checklist of what makes a deal unfinanceable, is just as clear. A weak or unrated tenant, or several short-term tenants where the debt outlives their commitment. Power that is hoped for rather than contracted, or an interconnection date that keeps slipping. Debt stretched so high that coverage sits barely above one, leaving no room for a stumble. A speculative build with no anchor tenant, betting that demand and a takeout lender will both show up on schedule. A structure that assumes a long useful life for a specialized asset that may age faster than modeled. Any one of these turns a lender cautious; several together, and the deal simply does not get funded, which is why so many announced megawatts quietly never break ground.

The synthesis returns to the domain thesis. AI infrastructure is a capital-formation story, and capital formation is governed by financeability. The firms that build are the ones that can present lenders and infrastructure funds with a creditworthy tenant, contracted power, conservative gearing, a credible build, and a de-risked refinancing, because that package is what unlocks cheap capital at scale. The firms that merely announce are the ones whose deals fail one of those tests and never clear a credit committee. Cost of capital and financeability are the true competitive battleground of the AI build-out. The chips get the headlines, but the map of what actually gets built is drawn by who can finance capacity, and how soundly they structure the debt that pays for it.