How to finance a data center?

The datacenter industry is fizzing with energy, optimism, and opportunity.

Investor conversations tend to converge around the same themes: AI, power availability, network demand, cybersecurity, operational execution, and the growing complexity of serving customers in an increasingly digital world.

Major shifts are currently taking place in the market that shape investor perspective and how datacenters are financed. A capital “supercycle” in data centre capex, is set to escalate further in 2027. AI is the driver – it’s not a bubble. Bubbles are defined by a disconnect from underlying cash flows. AI is the opposite: capex → demand → revenue → margin expansion. But “AI” is not one market, and it does have several constraints: MW/ silicon/ latency grid /capacity. 

Energy to power AI is the decisive site-selection factor: “Energy is the new land”. Energy, ideally clean energy availability is now the primary consideration when selecting datacenter sites. The interesting part to come will be building AI models for grid optimization and planning. Across EMEA, the market is expected to add 13GW of new capacity by 2030, reaching 34GW in total. In that context, power is no longer just an input cost; it is a commercial asset.

So, a 100 MW AI campus may look like a compelling proposition in a demand forecast, but it only becomes truly feasible when powered land is fixed as well as customer commitments, construction delivery, and capital timing coincide. The binding constraint in infrastructure right now is lead time. Yet in a market moving this fast, it will not be possible to wait for contracted cash flows before deploying investment.

In financing datacenters, funders are not underwriting GPU demand in isolation. They are underwriting a real asset with unusually concentrated energy exposure, rapidly evolving technology requirements, and a narrow margin for execution error.

The opportunity is huge. AI workloads are changing the economics of capacity, density, and location selection. But higher revenue potential does not automatically translate into more leverage. The projects attracting institutional capital are those that convert AI demand into durable, contract-backed cash flow while keeping energy and delivery risk within a clearly defined capital structure.

How to Finance AI Data Centers: Is underwriting the cash flow the right approach?

The starting point is not the headline value of AI capacity. It is the quality of contracted revenue. Traditional datacenter financing has relied on leases with hyperscalers, cloud platforms, major enterprises, and investment-grade counterparties and increasingly, neoclouds. AI infrastructure raises the stakes because a smaller number of customers may account for a much larger share of revenue, while their requirements for power density, cooling, network performance, and delivery dates are more exacting.

A financeable project therefore needs a credible answer to several linked questions: Who will occupy the capacity? On what contractual terms? How much of the load is committed? What happens if deployment schedules move? And does the customer agreement pass through power costs, curtailment risk, and change-order exposure in a way lenders can accept?

Pre-leasing remains the strongest route to conventional project debt, particularly when the counterparty is creditworthy and the lease term extends well beyond debt maturity. That said, investors should resist treating a signed lease as a complete solution. A contract that permits broad termination rights, delays customer acceptance, or leaves power costs materially unhedged can weaken debt capacity even if a tenant name is impressive.

How will the capital be split?

There is a growing case for selective speculative development in supply-constrained markets. Developers with secured land, interconnection rights, and a proven delivery record may be able to capture a premium by building ahead of demand. Yet speculative AI capacity should generally be financed with equity, development capital, or flexible private credit rather than capital that assumes immediate stabilization.

The distinction matters. Contracted capacity can support lower-cost senior debt because cash flow is visible. Merchant or partially committed capacity may create greater upside, but it requires investors who are prepared to absorb lease-up risk, equipment repricing, and a potentially longer path to refinancing. Treating both models as if they carry the same risk is one of the fastest ways to create a strained capital stack.

Match the Capital Stack to the Build Curve

AI data center development is capital intensive before it produces meaningful revenue. Land control, transmission studies, interconnection deposits, substation works, design, long-lead electrical equipment, and civil construction all demand capital well before a facility reaches customer acceptance. The capital stack must reflect that sequence.

Sponsor equity typically carries the earliest risk. It funds site acquisition, diligence, permitting, utility negotiations, and the work required to make a project credible to larger capital providers. Development partners, infrastructure funds, or strategic joint venture investors can expand this capacity, particularly where a sponsor has a differentiated pipeline but wants to preserve balance sheet flexibility.

Once the project has credible customer commitments and a defined construction plan, construction debt can finance a significant portion of hard costs. The lender's willingness to advance will depend on fixed-price or appropriately protected construction arrangements, equipment procurement, contingency levels, contractor strength, and the status of the power connection. At stabilization, construction financing may be refinanced into term debt, private placement debt, or a longer-dated institutional structure.

Debt Capacity Follows Certainty, Not Ambition

Private credit has become increasingly relevant where conventional banks are cautious about concentration, development risk, or complex energy arrangements. It can move more quickly and tolerate tailored structures, including delayed-draw facilities, preferred equity, or holdco financing. The trade-off is cost, covenant pressure, and a greater need for a reliable exit plan.

Senior debt remains the most efficient capital for stabilized assets, but it is disciplined capital. Lenders will test debt service coverage under power-price stress, schedule delays, reduced utilization, customer concentration, and refinancing assumptions. An investor should model these cases before entering a financing process, not after a lender raises them in committee.

Treat Power as a Capital Markets Issue

For AI campuses, power is not simply an operating expense. It is a core investment thesis. The asset may have land, permits, and customer interest, yet remain unfinanceable if its path to delivered capacity is uncertain.

Capital providers will scrutinize the interconnection agreement, utility delivery dates, substation scope, transmission dependencies, demand charges, fuel exposure where on-site generation is involved, and the allocation of curtailment risk. They will also assess whether a project's power strategy can accommodate the customer's expected ramp rather than only its ultimate contracted load.

This is where regional context becomes decisive. In mature markets, grid congestion may push investors toward expensive network upgrades or introduce behind-the-meter solutions. In emerging markets, the opportunity may be stronger but country risk, currency exposure, utility credit quality, and regulatory stability can reshape financing terms. There is no universal structure. A project with a long-term renewable power purchase agreement may attract one set of investors; a campus dependent on new transmission infrastructure may require another.

Bring Strategic Capital in Before the Financing Process

The strongest AI data center financings are rarely arranged by a single party acting alone. They are built around aligned interests across the full stack: developers, operators, customers, utilities, equipment suppliers, and capital providers. Strategic investors can add more than equity: they may bring procurement leverage, operating expertise, customer relationships, or a credible pathway to future expansion.

For institutional investors, the central question is whether the platform can replicate its first asset. A single campus with a large customer can be attractive. A portfolio of sites with secured power, standardized design, proven delivery capability, and a repeatable leasing model is a different proposition entirely. Portfolio scale can support lower funding costs and broader investor participation, but only after the underlying assets demonstrate consistency.

Joint ventures are particularly useful when the parties contribute genuinely distinct strengths. A local developer may control land, permits, and utility relationships. An established operator may provide technical credibility and customer confidence. An infrastructure investor may supply long-duration capital and governance discipline. The agreement must clearly allocate development overruns, dilution, decision rights, and follow-on funding obligations before pressure arrives.

+   +  + 

Meet the Investors behind the Infrastructure at Platform Global 2026 – Antibes, 8-10 September 

Register today: www.platform-markets.com/global