Power infrastructure intelligence for data center developers.

Utility interconnection queues in many U.S. markets now extend five to ten years. For hyperscale data center developers with aggressive build schedules, behind-the-meter generation has shifted from a contingency option to a primary power strategy. This piece examines the drivers, configurations and economics of behind-the-meter generation for AI data centers.

Battery energy storage systems are increasingly integrated into data center power architectures not as backup power replacements, but as active components of the power stack. This piece examines the primary use cases for utility-scale BESS in AI data center applications — resiliency, spinning reserve, load management and ramping support.

Institutional equipment financing through Triple Net Lease structures has emerged as a preferred capital solution for data center power infrastructure. This piece explains the structure, economics and credit requirements of Triple Net Lease financing for generation and storage equipment.

AI model training and inference workloads are driving electricity demand at a scale and pace that is straining existing grid infrastructure. This piece examines the demand projections, geographic concentration and infrastructure implications of the AI power surge for data center developers and power infrastructure providers.

Reciprocating engine generators and gas turbine generator sets each offer distinct advantages depending on project scale, fuel availability, site constraints and operational requirements. This piece provides a comparative framework for evaluating the two primary generation technologies for AI data center applications.

Special purpose vehicles are a standard structural element in power infrastructure financing, providing asset isolation, financing flexibility and tax efficiency. This piece explains the role of SPVs in data center power equipment financing and the key structural considerations for developers.

Behind-the-meter power generation has moved from a niche contingency strategy to the primary power solution for a growing share of AI data center projects. Utility interconnection queues in PJM, MISO, CAISO and other major U.S. markets now routinely extend five to ten years — timelines that are incompatible with the aggressive build schedules demanded by hyperscale operators and AI infrastructure investors. Behind-the-meter generation allows a data center developer to install generation capacity on the customer side of the utility meter, consuming electricity without routing it through the utility distribution system. The developer owns or leases the generation equipment, controls dispatch, and is not subject to interconnection queue timelines for the generation capacity itself. Utility service may still be used for supplemental power, backup or export, but the primary load is served by on-site generation. The economics of behind-the-meter generation have improved substantially as natural gas prices have stabilized and the cost of reciprocating engine generator sets and gas turbine generator sets has become more competitive relative to the opportunity cost of delayed data center commissioning. A data center that cannot be powered cannot generate revenue. When the alternative is a five-year wait for utility interconnection, the capital cost of behind-the-meter generation is frequently justified by the revenue acceleration alone. Reciprocating engine generators are the most common technology for behind-the-meter applications at AI data centers in the 10 MW to 200 MW range. They offer modular scalability — a developer can commission initial capacity and add units as load grows — along with fast-start capability, high efficiency at partial load, and broad availability from multiple OEMs including Wärtsilä, Caterpillar and Cummins. Natural gas fuel supply is typically available at data center sites in developed markets, and dual-fuel capability provides diesel backup for fuel supply disruptions. Gas turbine generator sets are preferred for larger-scale behind-the-meter applications above 100 MW, particularly at greenfield hyperscale campuses where a single large generation plant is more practical than a large array of reciprocating engine units. Aeroderivative turbines from GE and Siemens offer fast-start capability and high efficiency, while industrial frame turbines provide lower capital cost per MW at the expense of start time and part-load efficiency. Battery energy storage systems are increasingly paired with behind-the-meter generation to provide instantaneous response capability, load leveling and resiliency. A hybrid architecture combining reciprocating engines or gas turbines with a utility-scale BESS can deliver the continuous output of generation with the sub-cycle response of battery storage, eliminating the ramp-rate limitations of generation equipment and providing seamless ride-through during generation transitions. Permitting and emissions compliance are the primary regulatory considerations for behind-the-meter generation. Reciprocating engine and gas turbine installations above certain thresholds require air permits under the Clean Air Act, and developers must evaluate whether the project qualifies as a major source under Prevention of Significant Deterioration rules. Selective catalytic reduction and other emissions control technologies are available for both reciprocating engines and gas turbines and are frequently required in non-attainment areas. Equipment financing for behind-the-meter generation is available through institutional lenders and private equity investors using Triple Net Lease structures, sale-leaseback arrangements and project finance structures. The predictable cash flows of a data center operator with long-term hyperscale tenant commitments are well-suited to institutional equipment financing, and the equipment itself serves as collateral.

Triple Net Lease financing has become one of the most widely used capital structures for utility-scale power generation and battery energy storage equipment at AI data centers and hyperscale campuses. The structure aligns the interests of equipment owners, institutional investors and data center operators in a way that conventional equipment loans and operating leases do not. In a Triple Net Lease structure, a special purpose vehicle (SPV) is formed to own the power generation or storage equipment. The SPV acquires the equipment — typically reciprocating engine generators, gas turbine generator sets, battery energy storage systems or a combination — and leases it to the data center operator under a long-term Triple Net Lease agreement. Under the Triple Net Lease, the lessee (the data center operator) is responsible for all operating costs of the equipment, including maintenance, insurance and taxes, in addition to the base lease payment. The lessor (the SPV) receives a predictable, fixed lease payment stream with minimal ongoing obligations. The SPV is typically capitalized with a combination of institutional equity and senior debt. Institutional private equity investors provide equity capital in exchange for a preferred return on the lease payment stream. Senior debt is provided by commercial banks or institutional lenders against the equipment and the lease cash flows. The combination of equity and debt financing allows the SPV to acquire equipment at a lower all-in cost of capital than the data center operator could achieve through direct equipment purchase. For the data center operator, the Triple Net Lease structure provides several advantages over direct equipment ownership. The operator avoids the capital expenditure of equipment purchase, preserving balance sheet capacity for data center construction and IT infrastructure. The lease payments are typically structured as operating expenses rather than capital expenditures, which may improve financial ratios and covenant compliance. The operator also transfers equipment residual value risk to the SPV. For institutional investors, the Triple Net Lease structure provides exposure to the AI infrastructure buildout through a secured, income-producing asset with predictable cash flows. The equipment serves as collateral, and the creditworthiness of the data center operator — typically a hyperscale operator or well-capitalized developer with long-term tenant commitments — provides the primary credit support for the lease. Institutional investors with infrastructure mandates have shown strong appetite for Triple Net Lease structures backed by AI data center operators. Credit requirements for Triple Net Lease financing vary by investor and transaction size, but generally require the lessee to demonstrate sufficient creditworthiness to support the lease obligation. For large transactions, this typically means an investment-grade or near-investment-grade credit profile, demonstrated by audited financial statements, existing debt service coverage and evidence of long-term revenue commitments from hyperscale tenants. Smaller transactions may be structured with additional credit enhancements including letters of credit, parent guarantees or reserve accounts. Lease terms for power generation and storage equipment typically range from seven to fifteen years, with options to purchase the equipment at the end of the lease term at fair market value or a predetermined price. Longer lease terms reduce annual lease payments but increase the operator's long-term commitment. The appropriate lease term depends on the expected useful life of the equipment, the operator's long-term power strategy and the preferences of the institutional investor. Massive Technologies works with data center developers to structure Triple Net Lease financing for power generation and storage equipment, connecting qualified projects with institutional investors and banking relationships that have demonstrated appetite for AI infrastructure assets.

The U.S. electrical grid interconnection process has become one of the most significant constraints on AI data center development. In PJM Interconnection, which covers the Mid-Atlantic and Midwest regions and is home to the largest concentration of data center capacity in the world, the interconnection queue contains over 3,000 projects representing more than 300 gigawatts of requested capacity. The median time from interconnection application to commercial operation has extended to over five years in many regions, and studies have found that fewer than 20 percent of projects that enter the queue ultimately reach commercial operation. The interconnection queue problem is structural, not cyclical. The rapid growth of AI data center power demand has coincided with a surge in renewable energy interconnection requests, creating a queue backlog that cannot be resolved by incremental process improvements. FERC Order 2023, which took effect in 2024, introduced a cluster study process intended to improve queue efficiency, but the near-term effect has been to delay existing queue positions while the new process is implemented. For AI data center developers, the interconnection queue creates a fundamental tension between the speed of AI infrastructure demand and the pace of grid capacity expansion. Hyperscale operators and AI infrastructure investors are committing to data center campuses on timelines of 18 to 36 months from site selection to commissioning. A five-year interconnection timeline is not compatible with this development pace. The primary response among data center developers has been to pursue behind-the-meter generation as a substitute for or supplement to utility interconnection. Behind-the-meter generation allows the developer to install on-site generation capacity that serves the data center load without passing through the utility interconnection process. The developer still requires a gas distribution interconnection for natural gas-fired generation, but gas interconnection timelines are typically measured in months rather than years. A second response has been to pursue sites with existing utility interconnection capacity — brownfield industrial sites, retired power plant sites and locations with existing high-voltage transmission infrastructure. These sites command a premium, but the value of an existing interconnection position in a constrained market can be substantial. Some developers have acquired existing interconnection queue positions through assignment, though FERC rules on queue position transfers are complex and vary by regional transmission organization. A third response has been to engage directly with utilities and regional transmission organizations on dedicated interconnection agreements for large loads. Several utilities have developed expedited interconnection programs for large industrial customers, including data centers, that provide faster timelines in exchange for higher interconnection costs or load flexibility commitments. These programs are not universally available and typically require significant negotiation. The interconnection queue problem is also driving interest in grid-forming battery energy storage systems that can enable islanded data center operation independent of the utility grid. A data center powered entirely by behind-the-meter generation and battery storage, with no utility interconnection, eliminates the interconnection queue constraint entirely. The economics of fully islanded operation depend on fuel costs, equipment capital costs and the value of the avoided interconnection delay. For data center developers evaluating power strategy, the interconnection queue is a critical input to site selection, project timeline and capital structure decisions. Massive Technologies works with developers to evaluate behind-the-meter generation and storage options as alternatives to or complements of utility interconnection, and to structure equipment financing that supports rapid deployment.