The scale and growth of digital infrastructure finance and investment activity in North America in 2025 has little precedent in the asset class. Infrastructure investors and lenders have embraced the challenge of building out the computing power and ancillary infrastructure required to support AI workloads alongside the already burgeoning trends of growing demand for cloud computing and digitalization.

With several weeks remaining in 2025, US digital infrastructure transaction volumes already clock in at USD 202.4bn—nearly double the USD 117.4bn of transactions that reached financial close in all of 2024, according to Infralogic data. The capex cost for the average greenfield data center project has expanded to over USD 3bn, dwarfing the USD 882m average from 2024, driven in part by megadeals such as Meta and Blue Owl’s USD 35bn, 2 GW “Hyperion” artificial intelligence data center campus in Richland Parish, northeast Louisiana.

Unprecedented scale and capital demand

The digital infrastructure market is being defined by a growing number of megadeals and an intensive appetite for capital. Transactions for individual greenfield data center campuses now regularly exceed USD 1bn, with some projects reaching USD 10bn–USD 30bn and landmark valuations—such as the signed but not yet closed USD 40bn acquisition of Macquarie Asset Management-backed Aligned Data Centers by Artificial Intelligence Infrastructure Partnership (AIP). AIP comprises BlackRock’s Global Infrastructure Partners (GIP), MGX, and Microsoft—signaling unprecedented investor confidence.

“This year alone, the size of deals has consistently been over a billion dollars,” notes Dorina Yessios, partner and co-head of Energy, Infrastructure and Natural Resources at A&O Shearman. “We have seen developers starting to refocus on smaller deals as well and recognizing there is a market there in the mid-market, smaller market, whether they’re edge or co-location opportunities, and I think we’ll see more of that activity going into 2026.”

The investor base, consisting of an increasing array of infrastructure funds, pension capital, sovereign wealth funds, and new entrants from the worlds of tech and private equity, remains intensely eager to deploy capital. Yet a paradoxical constraint persists: only a handful of institutional investors can write multi-billion-dollar equity checks. This has led infrastructure investors to adopt new ways of sourcing sufficient resources for the still increasing demands for capital. Many investors in the trophy-asset tier are structuring consortium-style deals, as in the case of the AIP deal for Aligned, and there’s an emerging mid-market tier where capital formation requires more creative partnership structures.

“A typical hyperscale data center platform seeking growth capital to pursue its pipeline is looking at equity raises in the billions,” observes JPMorgan Managing Director and Head of North America Media and Communications Scott Wilcoxen. “When you take a step back and consider which institutions globally that can write a USD 2bn equity check it’s not a lot of firms.”

Adding to capital pressure, limited LP distributions over recent years have created an acute need for monetization for some. Developers who assumed they would build at a higher cost of capital and sell stabilized assets to lower-cost buyers face a structural shortfall: the low-teens-cost buyer base simply does not exist in sufficient number, forcing expected capital recycling models like stablecos or yieldcos to stall.

“The expectation was that developers with a high teens cost of capital would build assets and then sell the assets to the buyers with the low teens cost of capital. The problem is there’s not a lot of investors with a low teens cost of capital now so the capital developers assumed would recycle, isn’t being recycled,” Wilcoxen, who also serves as JPMorgan’s Head of Global Digital Infrastructure Investment Banking, added.

Some have managed to crack the code, however. Brookfield- and OTPP-backed Compass Data Centers recently signed a multi-billion dollar deal with KKR, the owner of 50% of CyrusOne alongside GIP, to monetize a portfolio of operating and future data centers in what appears to be a yieldco structure.

Financing structures and innovation

The most significant structural evolution in 2025 has been the convergence of project finance and real estate finance disciplines, the experts consulted for this piece told this publication. Historically siloed, these two approaches are now blending to accommodate the scale and complexity of modern data center financing.

“What we’re seeing in the project finance space is a convergence on structures between the more traditional project finance structures with real estate financings,” notes Gregory Tan, partner and co-chair of the Energy and Infrastructure practice at Paul Hastings. ”The amount of financing that’s needed is requiring a lot of these lead banks to tap not just the traditional project finance lenders but also the real estate lenders.”

Mini-perm structures—typically 5- to 7-year debt facilities—have become dominant, with lenders deliberately avoiding long-term commitments due to lingering uncertainty around technology obsolescence and tenant concentration risk.

Securitization issuance has surged dramatically with data center securitizations reaching over USD 11bn and CMBS issuance surpassing USD 12bn through mid-November 2025, according to a Fall 2025 report from commercial real estate finance trade association the CRE Finance Council.

The data center securitization market has more than doubled full-year 2024 volumes, according to the report’s author Raj Aidasani, Managing Director for Research at CREFC. “Average deal size has grown to approximately USD 1.1bn, up from USD 630m in 2024 and USD 320m during the 2022 trough.”

Insurance capital has emerged as an intriguing but still largely untapped capital source. While insurance companies have historically been active in securitizations and private placements, the equity side remains largely unexplored at scale, the sources consulted for this report said. Insurance companies are engineered to take credit risk, not equity risk, and the sector has yet to develop templates that fit insurance investment requirements.

“[Insurance companies investing in equity] hasn’t yet been done at size and so the assumption right now is that the market will have to evolve a little bit into the template that insurance companies require,” explains Adam Lewis, a managing director at Citizens and head of the firm’s Capital Markets’ Digital Infrastructure practice. “We’re still going to have to wait and see as to how these structures look that work for insurance companies.”

While yieldcos, hampered by rising rate environments, “are still not yet yielding,” in the words of one investment banker, joint venture structures have partially filled the void. They allow developers and sponsors to share risk and accelerate time-to-market. Private credit funds have also expanded their footprint significantly, moving beyond niche mezzanine roles into mainstream debt provision and, in select cases, entire deal execution.

“We have not yet seen the way people were initially thinking about yieldcos come to fruition,” observes Yessios. “But yieldcos are going to free up capital, and it’s going to give investment opportunities to new players.”

Power constraints and energy solutions

Go to any energy-focused conference and you will hear plenty of talk about data centers and vice versa. Power availability has emerged as the binding constraint on data center deployment, overshadowing even site acquisition and zoning challenges. This reality has fundamentally reshaped project design, financing structures, and sponsor partnerships.

“There will be a constraint in certain geographical areas because not everyone can access power. Everyone was talking about behind-the-meter solutions, so you can use it as a bridge to the longer-term solution. You just have to find the right partnerships to do that,” states Yessios.

Behind-the-meter generation using gas-fired generators has become a tactical bridge in capacity-constrained markets like parts of Texas and the Northeast. However, developers and lenders recognize this approach as temporary, a stopgap pending grid upgrades and renewable power development. The long-term solution invariably involves negotiation with utilities and, in many cases, developer-funded grid retrofit and upgrade projects.

“The only way that we’re ever going to get upgrades to the electrical grid, which needs to happen, is through these projects. But with these big projects, you’re seeing complete retrofits of the electrical system of the grid which should, over time, reduce costs,” notes Lewis. “Most of the data center operators that are putting in large load requests are actually paying for all those upgrades, directly, dollar for dollar.”

An important structural development has been the rise of integrated projects combining data center infrastructure with dedicated power generation—a shift from treating these as separate, sequential financings. Tan observes that sponsors are increasingly comfortable with integrated risk structures, though operational dependencies introduce complexities. “Integrated projects have both the data center and the power generation. This is addressing what in the past was the concern many lenders would raise about project-on-project risk.”

However, Wilcoxen cautions against overconcentration, arguing that comparative advantage suggests partnerships rather than vertical integration, where data center companies become power developers.

Location trends and community dynamics

The geographic footprint of digital infrastructure is expanding, yet traditional hotspots remain dominant. Virginia, Northern California (especially Santa Clara), Chicago, and Texas continue to attract outsized capital flows. However, the Midwest is emerging as a secondary growth market, offering lower land costs, better power availability, and more supportive regulatory environments.

“Texas is the area where we are seeing a lot of activity, both because of the size of the state and the access to gas and to power solutions. In the Midwest we’re seeing more and more activity,” explains Yessios.

Not everyone welcomes data center development on their doorstep, though, and community opposition is rising across markets, as highlighted by the high-profile public rejection of Blue Owl Capital real assets company Beale Infrastructure’s Project Blue in Tucson, Arizona (although the project has not halted and now appears likely to continue). Water consumption concerns and anxiety about energy prices are being amplified by the reality that most data center jobs are temporary with few permanent employment gains. Some jurisdictions have rejected projects despite zoning compliance, signaling a new political economy of digital infrastructure siting.

Citizens’ Lewis was more upbeat about the ability for investors to overcome the growing clamor against data centers.

“Most modern data centers don’t use any water and are configured in a closed loop,” he stated. “To the extent that there’s skilled labor on the construction side, there’s always an attempt to use local workforce. These things get sourced at 100 megawatts a year, so you could be employing people for 10 years.”

While data centers made up the bulk of the activity in the infrastructure finance world, fiber still deserves an honorable mention.

“For fiber and towers, inference AI will create more demand for bandwidth as the inference then has to get to our phones, cars, drones, etc.,” said Pat Lynch, executive managing director at CBRE Data Center Solutions. “The network component is opportunistic and without connectivity, a data center is just an expensive refrigerator. Building these factories in remote places means that the networks will race to get to them. Fiber and towers will remain a key component of network growth as all companies have data center space, with a lot of them residing in someone else’s facility, and so everyone touches the network in a broader sense.”

Large telcos remain highly relevant players among the field of infrastructure-owned fiber and broadband operators. One such incumbent player that has figured out the partnership approach is T-Mobile that has been very acquisitive in recent years.In July, T-Mobile completed the USD 9.8bn acquisition of Metronet from Oak Hill and the Cinelli family. As part of a consolidated fiber strategy, T-Mobile bought the Indiana-based broadband infrastructure company through its joint venture with KKR, which the pair had formed to acquire Metronet in July 2024.

Investor landscape and partnerships

As mentioned above, the investor base has expanded dramatically to meet rising demand. Traditional infrastructure funds remain dominant, but they are now joined by large PE firms, tech companies seeking direct compute ownership, real estate investors, international pension funds, and emerging-market sovereign wealth funds. “New entrants recognize that they need to put together the different pieces of the puzzle, and they’re coming from all around the globe,” said Yessios. “In the last year we have Israeli clients developing a new platform, Australian clients developing a platform, and Spanish as well.”

Joint venture structures have proliferated, serving multiple purposes: accelerating time-to-market, spreading risk, and providing new entrants with expertise. Financing banks have adapted real estate finance tools to digital infrastructure contexts.

“Traditionally, real estate financing has had guarantees, completion guarantees, bad boy guarantees,” explains Brian D. Hirsch, partner at Davis Polk, referring to clauses that hold a borrower personally liable in a non-recourse loan if they engage in specified bad acts. “Those were not normal in the infra world. Real estate historically is done through joint ventures, and few buildings are 100% owned by one owner. You’re starting to see that more in the data center space.”

The lender base has expanded dramatically. Once dominated by a handful of global banks and infrastructure lenders, it now includes substantial numbers of banks beyond the traditional bulge bracket, and major US banks are re-focusing on the project finance opportunities in the sector.

“We’re seeing lots of Asian banks participating in this space, more European banks than just the traditional French and Germans and the big US banks are stepping into the frame again,” notes Tan.

Capital formation and innovation

The financing ecosystem has expanded beyond traditional project finance into commercial real estate and capital markets frameworks. Available capital for deployment exceeds USD 170bn, according to Haig Bezian, managing director for the Infrastructure Credit Investment Practice at Cordiant Capital.

“Investor appetite remains robust, and a diversified financing ecosystem has emerged. Capital sources include traditional banks, infrastructure investors, real estate debt providers, private credit funds, asset-backed securitization, and ESG-linked financing instruments,” he notes.

A fundamental mismatch persists though. Debt supply vastly exceeds equity demand. Developers can readily access debt at attractive terms with long-dated tenant contracts, but equity capital formation—particularly for greenfield development—faces headwinds. This imbalance has driven leverage extension and creative capital stack innovation.

Outlook and key risks

As 2025 closes, valuations remain robust and sentiment bullish, yet headwinds merit close monitoring. “Capital is one constraint, but labor and parts are the biggest constraint,” warns Hirsch. “If you need a generator, it’s going to take six months. There’s still a shortage but if you take [data center developers] at their word, then this is going to continue.”

Tenant risk has evolved. While hyperscale tech platforms like Microsoft, Google, Meta, and Amazon provide quality anchors, the emergence of neo-cloud providers and generative AI startups has diversified the base. The central question is whether the broader economy will successfully monetize AI investments.

“There is a real risk that if companies don’t find ways to start monetizing all their AI investments, capex could be cut quickly. While I don’t think this will result in bankrupt companies and mothballed data centers, it would significantly slow the pace of development as existing capacity is absorbed,” cautions Wilcoxen.

Public market access remains another central unresolved question as investors consider their available exit strategies. While infrastructure funds are still early in 10-year holds, the lack of a deep, liquid public market creates a monetization squeeze that will likely dominate 2026. Multiple sources point to public markets as the only venue with sufficient depth.

“At the end of the day, it’s a question of market depth. I see an eventual need to get back to the public markets. That depth just doesn’t exist anywhere else,” notes Wilcoxen.

Many frame 2026 as a year of validation—execution on announced pipelines, fiber connectivity, and renewable power projects will test 2025 valuations. Transaction volume may moderate. Tan suggests activity could taper toward the second half of 2026, but the sweet spot for deals has shifted. “We might taper from having 25 deals in the market at the same time to only having 15. The sweet spot for the majority of the deals will be in the USD 2bn to USD 8bn range,” Tan explains.

Lynch takes an upbeat position on the state of the market when thinking of the state of artificial intelligence more broadly. “The digital infrastructure buildout is a global opportunity that is just getting started, especially in terms of building AI and raw compute,” he said. “To use a sports analogy: the US is in the bottom of the third inning, Europe is at the top of the first, and Asia and Latin America are still throwing out the first pitch. In practical terms, we’re beginning to see inference AI use cases in the US, while in many other regions AI is still in its early stages with significant opportunities ahead.”

Hirsch, however, provided a note of caution: “A lot of the low hanging fruit is gone. It is only getting harder to get deals done because now the easy sites have been gobbled up, and everybody’s aware of the value of what they have. The sites that are left are either harder to acquire or worth more.”

Digital infrastructure has matured from niche investment to core, yield-plus-growth allocation. 2025 was proof-of-concept; 2026 will be operational validation. The market will be defined less by transaction volume and more by underwriting quality, lease structure resilience, and resolution of the public market question.