New data centers are orders of magnitude larger than those built a decade ago and require vast sums of capital to construct. Debt markets are evolving to meet the accelerating demand for financing
In October 2025, Meta and Blue Owl Capital announced a US$30 billion deal to fund Hyperion, a colossal 5GW data center in Louisiana.
Five years ago, a data center financing package of this scale would have been unprecedented—now, it is just one of many similar mega-construction projects announced over the past 12 months. Hyperscalers (technology companies that operate huge cloud computing and data center facilities) are deploying enormous amounts of capital to build the digital infrastructure required to meet the computing power demands of generative artificial intelligence (AI) technology.
According to Goldman Sachs, capital spending by AI hyperscalers will exceed US$525 billion in 2026—approaching the US$540 billion annual spend that the International Energy Agency (IEA) says is needed to maintain current oil & gas output. And this is just the start. McKinsey projects that nearly US$7 trillion in capital spending on data centers will be required by 2030 to meet demand. OpenAI alone plans to invest up to US$1.4 trillion in computing infrastructure over the next eight years.
The sector is now operating at a magnitude unrecognizable compared to 10 years ago. The number of hyperscale data centers since 2018 has tripled to 1,297 facilities globally, according to Synergy Research. There are not just more of these facilities in operation; they are also much bigger. Data centers requiring tens of megawatts of power used to suffice, but advances in AI have precipitated a steep change from megawatt-level projects to gigawatt-scale developments.
Financing evolves
The size of new data center campuses and the capital expenditure required to build them have spurred debt capital markets activity to fund construction.
In the US alone, technology companies issued US$157 billion of public bonds through the first nine months of 2025, up 70% year-on-year according to Bloomberg, with investment in AI infrastructure a primary driver of the activity. Investment banks expect technology issuers to tap debt financing even more in the coming years as data center construction ramps up.
The sheer scale of capital required has already reshaped how financiers pull packages together.
When deal sizes were smaller, dealmakers would frequently finance data centers using either project finance-style or real estate construction-style financing, depending on the lender groups involved. With financing requirements now regularly climbing into the tens of billions, hybrid structures (which blend project finance principles with real estate terms to attract a broader range of lenders, including insurance companies) are becoming more common.
Project finance and real estate finance have always shared some similarities, although the former typically includes ongoing financial covenants, whereas the latter often features cash-flow sweeps. Data center financings are now incorporating elements from both approaches to ensure that real estate and project finance lenders can capitalize on familiar terms and concepts.
Private credit players have also stepped into the market. Even though spreads for private credit would typically be around 150-200 basis points above the cost of traditional construction financing debt provided by banks, private credit lenders have gained traction by adapting capital structures used to finance large infrastructure projects in other capital-intensive industries to unlock meaningful sums of cash for data centers.
For example, hyperscalers and private credit providers have partnered to form joint venture (JV) holding companies where the hyperscaler takes a minority equity stake and the private credit firm takes a majority stake. The JV holding company then takes 100% ownership of a special-purpose-vehicle project company that owns the physical data center assets, including land, servers and all other components.
The structure is complex but offers advantages. Frequently, private credit providers make an equity investment at the JV holding company and are treated as equity rather than debt, allowing hyperscalers to raise large amounts of financing without adding debt to their balance sheets. In many cases, this means that such equity will not be consolidated as debt for accounting purposes.
Private credit providers often use debt and equity syndications to raise funds to support their equity investment at the JV holding company. The JV holding company typically receives additional protections from hyperscalers, who usually assume risk for cost overruns or construction delays. Hyperscalers also frequently commit to triple-net leases where, as tenants, they will cover property taxes, building insurance and maintenance costs in addition to base rents.
Dealmakers have used this model in other contexts, such as liquefied natural gas (LNG) projects, but variations of the structure are now being used to fund data centers as well.
Moving parts
The rise of new financing structures is a consequence not only of mounting data center debt requirements, but also the increasing operational complexity of these projects. As data centers have become bigger and more ambitious, there are more moving parts and a greater focus on project risk among lender groups.
For example, the power required by data centers to operate has become just as important to smooth operations as access to chips and servers. The IEA forecasts that data center electricity consumption will double by 2030.
Access to power is essential. Data center developers must either negotiate power purchase agreements with electricity utilities to fund the addition of power generation capacity to electricity grids, or contract directly with power generation companies to build out “behind the meter” energy generation exclusively for data center use.
This brings more counterparties and complexity into the value chain. Data center developers are not just installing racks and cables, but also energy infrastructure and grid connections. Lender scrutiny of supplier, contractor and customer creditworthiness has become more important as a result.
Refinancings and renewals move into the frame
In 2026, lenders and hyperscalers will be preparing for an upcoming round of lease renewals and refinancings.
Although the demand for exposure to data centers has been strong through the past 12 months, refinancings and renewals of debt issued to support the construction of earlier-generation data centers will not necessarily be straightforward. For example, rapid development in chip technology may require older data centers to upgrade power supply and cooling systems. There is thus a risk of existing tenants opting to move to newer, cheaper facilities with more up-to-date design, rather than renewing leases on existing sites. Stakeholders will be watching closely to see how refinancings on older sites progress.
Lender groups are likely to adopt a more cautious approach to deploying capital into the sector generally, with more scrutiny of whether the substantial upfront investment in data centers will translate into profitability. Technology shifts, such as more efficient chips, may impact future demand for computing capacity.
[View source.]
