Machine Center: The AI-Driven Transformation of Data Center Investment

The sector recorded more than $70 billion in M&A transactions in 2024 alone, including Blackstone’s record-breaking $16 billion acquisition of AirTrunk, signaling unprecedented institutional appetite for digital infrastructure.[1] McKinsey projects companies will need to invest $5.2 trillion globally in data center infrastructure by 2030 to meet AI demand, with the US requiring $1 trillion in capital expenditures over the next five years.[2]

These estimates raise a central question for institutions: do data centers warrant strategic allocation, and do investors have the scale, expertise, and capital to execute in what now functions more like large-scale infrastructure than traditional real estate?

Rapid growth is undeniable, but it also brings material constraints and unknowns—particularly around power availability, technology pathways, and regulatory responses—that call for a candid, risk-aware approach from the outset.

The transformation reflects a fundamental shift in how society creates, processes, and stores information. US data center power consumption is projected to rise from roughly 4% of national electricity use today to 9–12% by 2030, driven by AI workloads that are far more power-intensive than traditional cloud applications.[3]

This surging electricity demand is colliding with grid bottlenecks and long interconnection queues, reshaping siting, valuation, and feasibility. In Northern Virginia—the world’s largest data center market with 3,000+ MW of capacity—developers face multi-year waits for power connections as utilities race to expand transmission and generation to match demand.[4] Power has become a binding constraint that increasingly defines which projects can proceed and which sponsors can reliably deliver capacity to tenants.

AI REVOLUTION DRIVES UNPRECEDENTED DEMAND ACCELERATION

The AI boom has altered data center economics, density, and design in ways that distinguish today’s cycle from prior cloud expansions. McKinsey estimates that AI will represent ~70% of total data center capacity demand by 2030, with global capacity potentially tripling from current levels.[5] Goldman Sachs forecasts data center power demand increasing 165% by 2030 versus 2023, with AI representing about 19% of data center power demand by 2028.[6] The core driver is computational intensity: training and serving modern AI models require large GPU clusters and high-density racks, pushing power per rack from ~10–20 kW in traditional designs to ~100–500 kW in specialized AI facilities.[7]

As density rises, purpose-built AI sites increasingly deploy liquid cooling, advanced power distribution, and new thermal management schemes that change both capital budgeting and operating practices.

Capital commitments reflect the scale of this shift. Hyperscale cloud providers—principally AWS, Microsoft Azure, and Google Cloud—collectively plan $300+ billion of data-center-related investment in 2025 alone, and industry analyses point to $1 trillion spanning data centers, chips, and utility upgrades.[8] Pricing has responded accordingly: according to JLL, rental rates increased 13–37% year-over-year depending on market and lease size, with the largest increases associated with high-density, power-secure deployments.[9]

Workloads also matter for footprints and contracts, because AI training facilities are few, extremely dense, and power-hungry, while inference and general cloud workloads are more distributed and latency-sensitive. As a result, hyperscalers are bifurcating their footprints into centralized campuses for training and regional capacity for inference and storage, and this architectural separation has implications for lease tenor, expansion options, and build sequencing for landlords and joint-venture partners.

The same momentum that accelerates demand also intensifies pressure on electricity infrastructure. In practical terms, power availability is the gating factor in development, and the reliability, timing, and cost of energization increasingly determine which sites move first and which proposals remain on paper. The net effect is that data center investing is converging with energy infrastructure investing, and power strategy is becoming as central to underwriting as tenant credit.

POWER SCARCITY RESHAPES LOCATION STRATEGIES

Power availability is now the top site-selection criterion, often superseding proximity to end users or fiber hubs.[10] Constraints vary sharply by region: some mature hubs face years-long interconnection timelines and expensive network upgrades, while select emerging markets offer faster access to capacity and land.

Northern Virginia illustrates both poles of the issue, with vacancy below 1% and peerless ecosystem advantages on the one hand, and severe power limitations on the other.[11] Dominion Energy projects demand that would require the equivalent of several large nuclear plants, underscoring structural scarcity and the need for long-cycle grid investments.[12] Where utility access is attainable, land values have appreciated materially, and adjacency to substations and transmission corridors has become a prized attribute that can materially compress time-to-revenue.

Development timelines are increasingly governed by the utility interconnection process rather than by shell construction. Queue position, required network upgrades, and study cycles often set the critical path, which means sponsors must underwrite multiple delivery scenarios, including phased energization and partial curtailment, and must build contingency for upgrade cost sharing.

Substation proximity, transmission headroom, and the ability to stage capacity in blocks can materially affect delivery risk and schedule reliability, and they therefore influence financing terms, tenant negotiations, and valuation. These grid mechanics have elevated power engineers and interconnection specialists to core members of underwriting teams, and they have also pushed many sponsors to treat power as a front-end asset rather than a downstream enabling input.

Energy sourcing is evolving in tandem with siting. Developers and operators are experimenting with on-site generation, dedicated substations, and long-dated contracts for nuclear and renewable resources, including examples such as Amazon’s acquisition of a nuclear-powered campus and Microsoft’s 20-year agreement related to the Three Mile Island site.[13]

Government analyses suggest that roughly 27% of facilities could incorporate on-site generation by 2030, signaling a partial shift from sole reliance on grid power to hybrid models that pair utility supply with behind-the-meter resources.[14] As primary markets approach grid limits, secondary markets with existing capacity, competitive wholesale power, and cooperative permitting regimes are attracting capital. Fiber diversity, water availability and discharge permitting, air-quality constraints on backup generation, and local land-use politics can still determine financeability once power is secured, which is why viability increasingly rests on an integrated plan for electricity, connectivity, and community impacts.

CONCENTRATED OWNERSHIP CREATES HIGH BARRIERS

Ownership and development capabilities have consolidated among large private equity and infrastructure sponsors, public REITs, and select strategics. Blackstone reports a $55+ billion portfolio and a $70+ billion prospective pipeline in data centers, positioning itself as a leading investor in AI infrastructure.[15]

Public REITs, notably Equinix and Digital Realty, continue to expand at scale, often via joint ventures and platform acquisitions that combine balance sheet capacity with operating expertise. Since 2015, the industry has recorded multiple $10B+ transactions and a steady cadence of platform acquisitions by infrastructure funds, including the $7 billion Digital Realty–Blackstone joint venture aimed at hyperscale development.[16] Sponsors in this cohort bring more than capital: they offer technical expertise, hyperscale relationships, and execution capabilities across design, procurement, and commissioning, which together constitute rising barriers to entry.

Operating models span retail colocation, wholesale or turnkey suites, and hyperscale build-to-suit shells with bespoke mechanical, electrical, and plumbing systems. Revenue stacks typically combine space and power rent with interconnection and managed services, and value creation tends to hinge on mechanical and electrical efficiency, change management during tenant fit-outs, and the ability to sustain strict uptime service-level agreements.

While hyperscalers drive the majority of leasing and development today, enterprise users, content platforms, and AI-native firms also contribute to demand, albeit at a smaller scale relative to the cloud majors. This concentration elevates credit exposure and re-tenanting risk, but it also affords longer lease tenors and clearer capacity roadmaps, which can strengthen the durability of cash flows.

Investor appetite remains strong, but barriers to entry continue to rise. CBRE’s survey indicates that 97% of investors plan to increase data center allocations in 2025, with 44% committing $500 million or more.[17] At the same time, development costs of roughly $10–14 million per megawatt, the need to secure power, and the scarcity of specialized operating talent constrain new entrants and tilt outcomes toward scaled platforms with repeatable delivery models.[18]

SOPHISTICATED FINANCING STRUCTURES EMERGE

The scale, technical complexity, and long lead times of data centers have catalyzed financing structures that prioritize capital efficiency and risk alignment. Sale-leasebacks, in which an operator sells an asset and leases it back on a triple-net basis that shifts taxes, insurance, and maintenance to the tenant, remain popular with hyperscalers and have often traded at 5–7% capitalization rates, a premium to many investment-grade corporate bonds of the same tenants.[19]

Joint ventures between developers and institutional capital, including the Digital Realty–Blackstone partnership, share development risk while leveraging specialist operating capabilities, and promote structures that align incentives when return hurdles are met.[20] Debt markets have adapted as well, with lenders such as commercial banks and insurance companies typically underwriting 60–75% loan-to-cost for pre-leased, credit-tenant construction, and with structured facilities that bridge construction and permanent phases to optimize all-in cost of capital.[21]

Underwriting mechanics reflect construction realities and tenant expectations. Milestone-based draws, completion guarantees or EPC backstops, and step-in rights for lenders or joint-venture partners address delivery risk, while owner-furnished equipment schedules and working-capital lines mitigate the impact of long equipment lead times on commissioning. Hedging strategies—ranging from forward rate locks to collars—must be sized to lease timing and commercial operation dates to avoid basis risk between funding and rent commencement.

Tax and incentive layering is equally consequential, since many jurisdictions offer sales and use tax exemptions on equipment, property-tax abatements or payment-in-lieu-of-tax agreements, and infrastructure credits for substations or road improvements that can move projects from marginal to financeable. Long-dated power purchase agreements and virtual PPAs can de-risk power costs and support customer sustainability requirements, which increasingly factor into procurement decisions and revenue visibility. As a result, the financing toolkit now resembles infrastructure finance, with long-dated cash flows and credit-anchored underwriting central to execution.

NEAR-TERM RISKS DEMAND CAREFUL NAVIGATION

Supply risk is the most visible near-term concern. Primary markets have 6,000+ MW under construction—more than double the pipeline at year-end 2023—and while vacancy remains below 2% in key hubs, the sheer volume under way raises the possibility of localized oversupply or absorption timing mismatches if AI adoption slows or efficiency gains reduce required capacity.[22] Investors should therefore examine not only headline megawatts but also the phasing of delivery, the degree of pre-leasing, and the mix of training versus inference workloads that will ultimately drive power draw and revenue ramp.

Power and delivery timelines remain stubborn bottlenecks. Even with accelerated investment, major transmission upgrades often require multi-year timelines, and CBRE and JLL indicate that 4–7 years is typical for power delivery to new sites, which creates a persistent gap between demand and available capacity.[23]

Some jurisdictions have imposed development restrictions or shifted infrastructure costs to developers, adding complexity and risk that must be explicitly modeled in budgets and schedules. In this environment, contractual mechanisms that protect against energization delays and curtailment events can be as important as economics on base rent.

Rates and valuations also warrant caution. Goldman Sachs notes that while demand remains strong, valuation multiples for data center assets embed optimistic growth assumptions, and public comparables trade at elevated multiples relative to broader infrastructure.[24] That premium could compress if financing costs remain high, if growth expectations moderate, or if supply catches up faster than anticipated.

At the project level, equipment lead times for transformers and switchgear and construction inflation continue to pressure budgets and schedules, which places a premium on capital structure discipline, contingency planning, and covenant flexibility to manage execution risk.

Regulatory and social-license considerations are growing more salient. Local moratoria, heightened scrutiny of water use and discharge, diesel-generator permitting constraints, and concerns about noise, traffic, and visual impact have delayed or resized projects in several jurisdictions. Disclosure requirements around carbon intensity and renewable sourcing are also tightening, and these policies influence siting, power procurement, and ultimately the depth of the tenant pool. Sponsors should assume that community engagement and environmental disclosures will remain core to feasibility rather than peripheral compliance items.

Tenant and contract structure risks must be underwritten explicitly. Heavy reliance on a small number of hyperscale tenants concentrates credit and renewal risk, and contract structures vary widely, with take-or-pay commitments differing materially from usage-based billing and variable power pass-throughs.

Investors should diligence termination rights, expansion options, and performance credits under service-level breaches, since these clauses determine revenue resilience when operating or grid conditions deviate from plan. Concentration can be an advantage when it yields long tenors and predictable expansions, but it can also magnify downside if a single counterparty reprioritizes a region or a technology stack.

CLIMATE AND TECHNOLOGY DISRUPTIONS LOOM LARGER

Physical climate risk is no longer a theoretical concern for the sector. Many top markets exhibit rising exposure to extreme weather, heat, and water stress, each with direct implications for cooling resiliency, backup power operations, and operating costs. Industry surveys indicate that nearly half of facilities report weather-related disruptions, and water scarcity constrains evaporative cooling in several regions where that method has historically been favored.

Site selection and design now routinely incorporate elevation and floodplain analysis, heat-tolerant cooling solutions, and water-light designs, and operators are deploying on-site energy storage and re-dispatchable resources to navigate grid events and curtailments. These adaptations improve resilience, but they also raise capital and operating costs that must be captured in underwriting.

Technology pathways introduce a different set of uncertainties. Edge computing is distributing processing closer to users, complementing hyperscale cores but changing latency, network, and capital-spending considerations for certain workloads. Quantum computing, photonic processors, and neuromorphic architectures could materially alter power density, cooling requirements, and facility layouts, even if their commercial timelines are uncertain.³³ The immediate shift to liquid cooling for AI workloads is already creating functional obsolescence risk for facilities designed solely for air cooling, which means investors should favor shells and central plants designed with modularity and multiple cooling configurations in mind. Emissions and reporting add another dimension, as tenants and regulators place greater emphasis on Scope 2 emissions, hourly matching, and grid carbon intensity; long-term competitiveness may hinge on access to low-carbon megawatt-hours and credible disclosure practices that withstand increasing scrutiny.

Even experienced real-assets investors may lack the technical depth to underwrite these transitions confidently. In practice, this argues for in-house technical expertise, specialist partners, and design optionality in electrical and mechanical systems, all aimed at reducing the risk that equipment choices or footprints are stranded by changes in computing and cooling. These capabilities are not optional add-ons but part of the central skill set for investing in an asset class whose performance is tied to technology and power rather than to building envelopes alone.

STRATEGIC POSITIONING FOR A TRANSFORMING SECTOR

The investment thesis is compelling but must be framed with guardrails. The sector sits at an inflection point where explosive demand meets binding power constraints, and McKinsey’s $5.2 trillion global investment need underscores the magnitude of capital formation required to support the AI economy. Contractual cash flows, inflation pass-through characteristics, and mission-critical use cases can generate attractive risk-adjusted returns, yet execution risk remains significant because deliverability, schedule reliability, and technology choices are inseparable from value realization. In this context, strategy should be built around a small set of durable advantages: power access, design adaptability, operating depth, disciplined capital, and institutionalized risk governance.

Power strategy must become a core competence rather than a procurement exercise. Sponsors who secure capacity early through utility agreements and substation adjacency, who pursue on-site generation where feasible, and who participate in nuclear or long-dated renewable partnerships are more likely to deliver on time and on budget. Design should assume high density and liquid cooling from the outset, with modular electrical and mechanical systems that can scale and retrofit as computing paradigms evolve, and with floor loading, white-space modularity, and thermal pathways engineered to accommodate multiple configurations without structural rework. Partnerships with operators who bring hyperscaler relationships, delivery track records, and resiliency engineering can balance risk and control through well-structured joint ventures, while capital stacks that match tenor to lease profiles and preserve flexibility for refinancing and market shifts can absorb shocks in equipment delivery and pricing.

Risk governance should be explicit and continuous. Investors can apply a simple but effective set of underwriting checkpoints—deliverability, durability, and adaptability—to triage opportunities and to ensure that power certainty, tenant credit and contract mechanics, interconnection ecosystems, and resiliency design are evaluated with equal rigor. The sector has shifted from commodity storage to critical AI infrastructure, and while power constraints are challenging, they also create moats for platforms that solve them consistently.

If current trends sustain, data centers may become a defining infrastructure asset of the digital economy; equally, the sector’s evolution from energy sourcing to compute architectures means that the only constant is change.

MORE FROM SUMMIT #19

PLATFORM SPONSOR

ASSOCIATE SPONSOR

Note from the Editor: Issue #19

Mid-Year Pulse: Global Investors on the Economy, Energy, and Housing

Benjamin van Loon | AFIRE

Bend, Not Break: Investing in Real Estate Amid Economic Uncertainty

John Murray + François Trausch + Russell Gannaway + Kirill Zavodov | PIMCO

Test Of Time: How US Real Estate Withstands an Uncertain Investment Market

Riaz Cassum | JLL

Oh Canada: Why International Investors Might Want a Second Look at Canadian Real Estate Markets

Amy Erixon + Long Tang + Daniel Goldberg + Marie-France Benoit | Avison Young

Beyond Oil: Why Gulf Family Offices Are Doubling Down on US Logistics, Data, and Housing

Abbas Hashmi | Saudi Family Holdings

South of the Border: Higher Yields and Growth in Mexican Industrial

Shaun Libou | Raymond James

Migration Myth: NIMBYism and Why Coastal Movers Aren’t Affecting Sunbelt Housing Supply

Donal Warde | Consultant + Ron Bekkerman | Constellation Data Labs

Machine Center: The AI-Driven Transformation of Data Center Investment

Sam Chandan | Chen Institute for Global Real Estate, NYU Stern School of Business

AI in Commercial Real Estate: A Practical Guide for Industry Leaders

Armel Traore Dit Nignan + Shaarvani Kavula | Principal Real Estate

Artificial Control: Are You Ready for AI Management and Oversight?

Marie-Noelle Brisson + Michael Savoie | CyberReady, LLC

Capitalizing on Dynamics: Demographic Mega-Trends Impacting CRE

Stewart Rubin | New York Life Real Estate Investors

Mid-Cap Assets: An Under-Examined Segment in CRE

Asaf Rosenheim | Profimex

The Lifestyle Renter: A Growing Opportunity for Strategic Investment in High-Quality Apartment Communities

Hannah Waldman | The Dermot Company

The Climate Is Speaking: CRE Underwriting for a Future That’s Listening

Ines Diez + Thomas Stanchak | Stoneweg

+ EDITOR’S NOTE

+ ALL ARTICLES

+ PAST ISSUES

+ LEADERSHIP

+ POLICIES

+ GUIDELINES

+ MEDIA KIT

+ CONTACT