Power Bottlenecks Reshape AI Infrastructure: Convergence of Financing and Rising Computing Scarcity

Overview

This report delves into the structural constraints facing AI infrastructure expansion, particularly how power supply bottlenecks are reshaping the industry landscape. The report argues that power is no longer a secondary consideration but a core constraint on par with data center development. This blurs the boundary between AI infrastructure financing and energy complex financing, giving rise to new investment models. Meanwhile, labor shortages, water resource challenges, and policy obstacles further exacerbate supply tensions, enabling companies with large-scale, reliable computing resources to gain stronger pricing power and emerge as 'computing merchants'.

Core views

Power supply bottlenecks are the primary constraint on AI infrastructure expansion. As AI data centers become one of the largest sources of incremental electricity demand, power generation capacity, transmission networks, and equipment supply chains face prolonged lead times. Data shows that the average delivery cycle for power transformers has reached 128 weeks, and for generator step-up transformers, it’s 144 weeks—compared to just 12-16 weeks before the pandemic. Moreover, grid connection backlogs at the beginning of 2025 exceeded twice the installed capacity in the U.S., and the situation could worsen further. This dynamic forces developers to prioritize sites with available power and explore solutions that integrate power generation more closely with computing. Financing models are undergoing profound changes. As access to power becomes an obstacle, financing needs for AI infrastructure and energy complexes tend to converge. Off-grid power solutions (such as fuel cells, gas turbines, and energy storage) and ‘fast-track power’ strategies (such as repurposing Bitcoin mining farms) have become central to construction. More and more AI players are directly acquiring, contracting, or financing these assets rather than treating them as independent utility investments. This effectively merges previously separate funding pools into a more integrated and interdependent financing model. Beyond power, multiple structural constraints are emerging. On the labor front, the U.S. is expected to face a shortage of about 300,000 electricians over the next decade, and more than one-fifth of the existing workforce is over 55 years old, facing an impending retirement wave. In terms of water resources, S&P analysis shows that 43% of global data centers are located in areas with high water pressure, raising sustainability concerns. On the policy side, bipartisan opposition to new data centers is growing; New York State proposes suspending new projects, Texas requires ensuring that incremental demand does not drive up costs for other users, and another 14 state legislatures are considering some form of moratorium. Supply-demand imbalance strengthens pricing power. These structural challenges make it difficult for the industry to deliver computing capacity according to current demand trajectories, intensifying the risk of structural supply-demand imbalances. In this environment, scarcity becomes a defining market characteristic, enhancing the pricing power of companies with large-scale, reliable capacity—the emerging ‘computing merchants’. Notably, enterprise-level applications currently show relatively low elasticity in their demand for computing power; higher prices are unlikely to substantially slow down adoption rates and may instead encourage shifts toward higher-value applications.

Analysis framework

The report adopts a top-down macro and industry cross-analysis framework. First, it starts from the perspective of capital expenditure (Capex) financing, identifying power as a core upstream constraint. Second, it quantifies the severity of bottlenecks through specific data (such as transformer delivery cycles and grid connection backlogs). Next, it expands the analysis dimensions to labor, natural resources (water), and the policy environment, building a comprehensive picture of supply-side constraints. Finally, combining demand-side characteristics (lack of elasticity), it derives the market structural changes and pricing power shift logic under supply-demand imbalances. This methodology emphasizes cross-industry (tech and energy) correlation analysis.

Methodology notes

• Industry/Industrial Analysis FrameworkSupply-demand framework

  Supply-Demand Framework Analysis

  By analyzing the rapid growth in AI computing demand alongside rigid supply-side constraints such as power and labor, the report concludes that the market will experience structural imbalances, driving up pricing power for those with supply advantages. This is a typical supply-demand analysis paradigm used to identify industry bottlenecks and value transfer directions.

• Industry/Industrial Analysis FrameworkUpstream-Midstream-Downstream Industry Chain Transmission

  Industry Chain Transmission and Bottleneck Identification

  The report focuses on the upstream constraints (power and equipment) affecting the midstream (data centers), pointing out that the bottleneck lies not in the data centers themselves but in their supporting infrastructure systems—a reflection of the approach to identifying critical bottleneck links in industry chain transmission analysis.

• Company Fundamentals and Financial Framework

  Capital Structure Integration Analysis

  The report observes the convergence of AI companies and energy assets at the financing level, where traditionally separate capital pools have become mutually dependent. This analytical perspective goes beyond single-industry financial analysis, focusing on cross-industry capital flows and the new investment logic brought about by blurred asset boundaries.

Key data

• Power Transformer Delivery Cycle128 weeksFar above the pre-pandemic level of 12-16 weeks
• Generator Step-Up Transformer Delivery Cycle144 weeksSevere supply chain congestion
• Grid Connection Backlog Size>2x U.S. Installed CapacityData from early 2025, with serious delays
• U.S. Electrician Shortage Forecast300,000 peopleExpected shortfall over the next decade
• Percentage of Data Centers in High-Water-Pressure Areas43%Globally, raising ecological sustainability concerns
• Number of State Legislatures Considering Moratoriums on Data Centers14Growing policy resistance

Impact & implications

The report suggests that the construction cycle for AI infrastructure will be longer, more sequential, and more capital-intensive than indicated by headline Capex forecasts. For investors, this means reevaluating the timing and risks of AI-related investments. Companies that own or can secure stable power supplies and computing resources will gain significant competitive advantages and pricing power, becoming 'computing merchants.' Meanwhile, the convergence of AI and energy financing offers new opportunities for energy infrastructure investors but also brings more complex cross-industry risks. Policy uncertainty could become a major variable impacting project implementation.

Risks

• Continued power supply shortages causing significant delays in AI infrastructure progress
• Tightening policy regulations (such as moratoriums) restricting new data center construction
• Unrelieved labor shortages driving up construction costs and operational risks
• Water resource pressures rendering operations unsustainable at some key hub data centers
• High electricity prices or excessively high computing prices curbing marginal demand

What to watch

• Legislative developments in various states regarding data center moratoriums
• Trends in delivery cycles for power equipment (transformers, etc.)
• Transaction cases and scales of AI companies directly investing in energy assets
• Updates on grid connection backlog data
• Actual performance of enterprise-level AI applications in terms of price sensitivity