Private Capital Targets AI Infrastructure as Power Bottleneck Becomes Primary Investment Thesis

Private capital managers overseeing $474 billion in global fundraising are positioning infrastructure assets to capture AI-driven electricity demand, with power generation and grid capacity emerging as the primary bottleneck constraining data center deployment through 2028.

The investment thesis centers on a structural mismatch: investment managers are seeking $474 billion worldwide, steering capital toward AI data centers, secondaries liquidity, and Energy mandates, while data-center developers concerned they could face power shortages, particularly in 2027 and 2028. This gap has elevated energy Infrastructure from a defensive allocation to a growth trade, with firms including Ares Management and Stonepeak Partners deploying tens of billions into power generation, transmission upgrades, and behind-the-meter solutions.

Ares Pivots to Digital Infrastructure, Flags Power as ‘Meaningful Opportunity’

Ares Management, with nearly $623 billion of assets under management, signaled the urgency of AI infrastructure at the Bank of America Financial Services Conference in February 2026. CEO Michael Arougheti outlined the firm’s acquisition of GCP last year, which added a data center development capability under Ada Infrastructure and a pipeline of large build-to-suit hyperscaler projects in markets including Tokyo, Osaka, London, São Paulo, and Northern Virginia. The firm closed a $2.4 billion Japanese data center fund in its first year.

Arougheti framed AI’s impact on private markets as an infrastructure play rather than a software disruption risk. According to Markets Daily, he said the AI revolution could drive “meaningful opportunities” in areas such as digital infrastructure, renewable energy, and transmission. The firm deployed a record roughly $46 billion of capital in the fourth quarter of 2025, with $150 billion in dry powder available for deployment.

Stonepeak Scales Data Center Platforms, Targets Hyperscale Behind-the-Meter Power

Stonepeak, ranked the sixth largest infrastructure investment firm based on total fundraising over the most recent five-year period, has committed $1.5 billion to AI-specific infrastructure through platforms including Montera Infrastructure. According to Data Center Frontier, Montera is designed to develop turnkey, build-to-suit facilities exceeding 100 megawatts in capacity, strategically located in Tier I and Tier II metropolitan areas with access to near-term power solutions.

The firm’s digital infrastructure portfolio spans data centers, fiber, residential broadband, small cell, and tower assets across the globe, operating more than 160,000 miles of fiber and over 100 facilities and 500 MW of capacity, with an additional 400 MW in development. Stonepeak has also invested in Longview Infrastructure, a transmission platform, and secured a 40% stake in a Louisiana LNG project valued at $5.7 billion, underscoring its vertical integration strategy across the energy-to-compute value chain.

Bloomberg Invest Signals LP Appetite for Long-Duration Infrastructure Cash Flows

At Bloomberg Invest New York 2026, institutional investors signaled clear allocation preferences: deployment discipline, secured power access, and contracted revenue streams. According to reporting compiled by The New York Report, limited partners are signaling clear preferences: they want deployment discipline, clearer liquidity paths, and evidence of operational expertise; they favor managers who can secure power, land, and grid access ahead of rivals.

The shift reflects a broader convergence: energy and technology are no longer separate asset classes in institutional portfolios. Bloomberg Intelligence projects that hyperscalers and Tier 2 cloud providers projected to invest more than $3.5 trillion in AI-related capital expenditures through 2030, with Microsoft on track to be the largest spender with projected capital expenditures exceeding $150 billion in 2026.

Power Bottleneck Drives Baseload Generation, Transmission, Behind-the-Meter Strategies

The fundamental constraint is not data center construction capacity but electrical infrastructure. According to Goldman Sachs Research, global power demand from data centers will increase 50% by 2027 and by as much as 165% by the end of the decade. In the United States, power demand from data centers could surge 20-40% in 2025 with strong double-digit growth likely to persist in 2026-30.

Transmission and interconnection queues represent the most acute bottleneck. Morgan Stanley analysis indicates that power spreads could rise by 15%, and this expansion in profit margins could lead to higher earnings forecasts for power generation companies and create $350 billion in value creation through the entire power supply chain. However, power companies are facing constraints that include more delays in getting grid equipment and an increase of 30% in costs since 2019.

3-5 Year Timeline Creates Urgency for Capital Deployment, Pre-Leasing Strategies

The investment window is compressed. Data center leasing hit record levels in 2025 — over 15 gigawatts according to industry trackers — but almost none of that capacity actually came online in 2025; it’s all scheduled to arrive in late 2026 and 2027, which means a massive surge in electricity consumption is about to hit the grid. According to Axios, up to 11 gigawatts of 2026 capacity “remains in the announced stage with no signs of construction,” with typical build times of 12 to 18 months meaning that capacity could still come online — but only with dramatic acceleration.

Private Equity firms have invested nearly $200 billion in data center deals since 2022, representing 80-90% of M&A activity in the sector, according to the Private Equity Stakeholder Project. Blackstone alone announced a $25 billion investment in data-center and energy infrastructure development in Northeast Pennsylvania. Brookfield launched a $100 billion AI infrastructure program in 2024, with BAIIF, together with additional capital from its co-investors and prudent financing, will acquire up to $100 billion of AI infrastructure assets, deploying investment across every stage of the value chain—from energy and land to data centers and compute.

Regulatory, Community Opposition Create Execution Risk, Delay Timelines

Political backlash is accelerating. Axios reports that with midterm elections heating up, communities are growing restless over rising power prices — which many blame on data centers that increasingly require city-scale electricity; Sightline has tracked more than 10 new moratorium proposals in the past month alone in U.S. states, including New York, Michigan, Virginia and Oklahoma.

In PJM, the grid operator serving the mid-Atlantic, at its energy capacity auction in December, PJM failed to procure all of the energy that it deemed necessary for the 2027-2028 delivery year — and the projected demand from data centers was largely to blame; PJM aims to have a 20% reserve margin for power system emergencies, and it only secured 14.8%. The shortage stemmed from an increase of nearly 5,100 megawatts of projected data center demand.

What to Watch

Deployment speed versus dry powder. With $150 billion at Ares and similar war chests at Stonepeak, Blackstone, and Brookfield, the critical metric is capital deployment velocity against power delivery timelines. Funds that secure interconnection agreements and PPAs in 2026 will hold structural advantages through 2030.

Secondary market liquidity. As exit markets remain constrained, secondaries volume in infrastructure could accelerate. Managers with platforms offering earlier liquidity — particularly in over-allocated hyperscale data center portfolios — may gain LP allocation share.

Policy shifts on self-generation. Maryland, Virginia, and Texas are testing frameworks requiring data centers to bring behind-the-meter power. If adopted federally or across PJM, this accelerates Bloom Energy-style fuel cell deployments and natural gas microgrid investments, shifting value from utilities to independent power producers.

2027-2028 capacity crunch realization. If even half the announced 2026 projects face delays, utilization rates for existing powered data center capacity could exceed 98%, pushing lease rates and acquisition multiples higher. Conversely, rapid efficiency gains from inference-optimized chips could reduce per-rack power draw 30-50%, extending runway and compressing returns.

Nuclear restart timelines. Small modular reactors remain speculative for 2026-2027 deployment but could reshape baseload economics by 2029-2030. Constellation Energy and Vistra contracts at $15-25/MWh premiums signal hyperscaler willingness to pay for firm capacity — a pricing dynamic that validates long-duration infrastructure allocations over merchant power exposure.