The need to develop the digital infrastructure for tomorrow’s economy has never been greater. Where previously the widespread adoption of cloud services underpinned development, the transformational power of artificial intelligence is now driving the need for unprecedented levels of investment. However, recent news in the sector requires investors to rethink their approach to investments in AI. In particular, the evolving capital needs of the sector, driven by expanding demand, larger asset builds, as well as the changing macroeconomic/geopolitical environment and other factors, could open up further opportunities across asset classes — perhaps none more so than private credit.
Data centers — particularly AI-driven hyperscale facilities, which now account for over 40% of worldwide capacity — are expected to continue being built at a record pace. Recent estimates peg the global spend on hyperscale data center builds in 2025 to be at roughly $220 billion among major cloud providers. Moreover, AI expansion is expected to steer further investor interest in related sectors, such as power generation and transmission infrastructure. Also complicating the picture are developments such as DeepSeek’s R-1, which is expected to require far less computational power compared to prior AI models and has led to a reassessment of long-term infrastructure needs.
What this means is that investors of all types should look to re-evaluate the growth assumptions that would have underpinned their previous investments in large-scale data centers. A key consideration in making this assessment is which data centers will continue to retain their value once the critical mass of generative AI models have been trained and are primarily being used for inference, or being applied to actual use cases. While latency may not be a major consideration for the training of AI models, proximity to population centers will likely become a critical factor in the models’ future usefulness to end users.
The capacity requirement of newer data center builds continues to increase to accommodate AI use cases, which include both AI-training and inference functions. Developers and sponsors must therefore expand the pool of capital available to fund the construction of larger-scale projects that cater to these capacity requirements. In recent years, developers had mostly looked to the asset-backed security market to take out construction financings for their developed assets. The ABS market, in which data centers currently account for approximately 5%, or roughly $25 billion, of total issuances in 2024, allowed developers to add completed and cash-flowing assets under an existing master trust structure, raising additional funds in a more time-efficient manner. The typical ABS structures worked well for developers with multi-tenanted facilities, as ABS investors saw these as a hedge against lease renewal risk. The ABS market was therefore a logical place for the refinancing of construction loans that began to reach their maturities between 2020-2022, resulting in issuances which now face a pending wall of maturities.
The newest generation of hyperscale data centers geared to run AI models, which can cost up to $2 billion to build, are not as easily suited to ABS financings due to their sheer size. Construction loans for these types of data centers have now also reached the stage where they must be refinanced, in addition to the looming maturity wall brought on by the ABS issuances mentioned above. In recent years, these types of larger, hyperscale-focused assets have been developed as single-tenant facilities under longer-term contracts with initial lease terms of up to 15 or even 20 years. Characteristics like these have allowed developers to tap and gain the attention of another large source of capital: private credit.
From the developers’ perspective, private credit offers a diversification of funding sources, access to institutional investors that can provide significant upfront capital for longer time horizons, and expedited processes without formal ratings. Certain institutions also bring teams with considerable experience in project finance and the ability to structure bespoke credit solutions. The benefits to lenders include credits based on leases with highly rated counterparties, long-term contracted cash flows and premiums for illiquidity. Since over $10 billion of U.S. data center debt was refinanced in the last three years alone, it is likely that private credit will emerge as one of the prominent asset classes that address the refinancings of incoming ABS maturities.
As the share of stabilized data center refinancings begins to attract investors with longer horizons, investors will have to shape their views on how long-term data center demand can shift based not only on newer technologies and existing infrastructure bottlenecks, but also how the development of AI is to be addressed by regulatory bodies.
The debate on regulation vs. innovation has become the latest headline-grabbing phase of AI’s transformative power. At the recent Artificial Intelligence Action Summit in Paris, representatives from around the world gathered to discuss AI-related issues. As expected, a divergence between the nations on how to regulate AI quickly took shape, with some leaders pushing for an international agreement that pledges accessibility and inclusivity while others worry such a declaration could stifle innovation. The summit underscores how nations across the globe, many of which have yet to pass comprehensive regulations of their own, are choosing to approach this topic. Regulation, or the lack thereof, represents yet another important consideration for investors to take a long-term view on, as it will undoubtedly shape the future development of digital infrastructure.
Considering the dynamic nature of the current AI landscape, investors of all types must have the flexibility to assess several impactful credit considerations. As future fundraising continues to step away from ABS and other quasi-public styles of issuances, a continued increase in private credit opportunities is likely to arise.
Andrew Kleeman is co-head of private fixed income at SLC Management. He is based in Wellesley, Mass. This content represents the views of the author. It was submitted and edited under Pensions & Investments guidelines but is not a product of P&I’s editorial team.
