Private Credit Has Tightened Terms for AI Infrastructure
Lenders are using stricter covenants to manage the risks of a projected 950 TWh data center power load by 2030.
Updated on Oct. 6, 2026 in Data Centers
Live Poll
Is now a good time for individual investors to increase their exposure to AI-related infrastructure?
Private credit lenders have begun implementing stringent loan protections to mitigate risks associated with the massive infrastructure buildout for AI. This shift in financing comes as the sector faces significant energy bottlenecks and potential technological obsolescence.
Why it matters
The rapid expansion of AI-driven data centers requires enormous capital, forcing lenders to rigorously assess electricity supply and asset longevity. Investors are now prioritizing cash flow resilience to navigate grid constraints and the high probability of borrower disruption.
Global electricity consumption by data centers reached 485 TWh in 2025 and is projected to climb to 950 TWh by 2030. Lenders are now increasingly reliant on asset-backed securities to pool cash flows from multiple data center leases to diversify their total portfolio exposure.
The players
Nuveen
An asset manager with significant involvement in global real estate, infrastructure, and private credit financing strategies.
Arcmont
A private debt investment firm focused on capital solutions and credit structures for large-scale corporate infrastructure projects.
International Energy Agency
An intergovernmental organization that provides analysis and data on global energy policies and market projections.
The details
Lenders are mitigating the risks of a $19 trillion infrastructure buildout by employing milestone-based disbursements, which release funds only as construction phases are completed. To overcome utility connection delays, financing is increasingly directed toward behind-the-meter assets—power generation sources like gas turbines or battery storage located at the site of the data center. These protections are designed to protect investors against the risk of overbuilding and the rapid pace of technological change in the compute hardware sector.
Timeline
2025: Global data centers consumed 485 TWh of electricity.
April 2026: The International Energy Agency updated its projections for data center power usage.
September 30, 2026: Executives from Nuveen and Arcmont discussed AI credit risks during an IPE webcast.
2030: Projected global data center electricity consumption reaches 950 TWh.
The Tech Race
The race to scale AI capacity is fundamentally constrained by the physical limits of the global power grid. This shift toward restrictive private credit terms reflects a maturation of the sector, moving from speculative funding to a disciplined focus on long-term energy viability.
Industry participants should expect more rigorous site requirements and longer approval timelines for new infrastructure projects as lenders prioritize grid connectivity. These financing hurdles may prioritize developers with pre-existing, behind-the-meter energy solutions over those reliant on traditional grid expansions.
The takeaway
The move toward milestone-based lending signals that the AI buildout is transitioning into a phase of disciplined capital management. Investors and developers should monitor whether energy availability, rather than raw compute demand, becomes the primary barrier to future project approvals.
Further reading
For more on how infrastructure scale influences sector stability, visit Data Centers.
Source note: This article includes information reported by Delano.
Live Poll
Is now a good time for individual investors to increase their exposure to AI-related infrastructure?
