Data Centers and Digital Infrastructure

Data centers and digital infrastructure have emerged as one of the most attractive real asset classes for institutional investors and UHNWI portfolios. The explosive growth in demand for computing power, driven by artificial intelligence (AI), cloud computing, and the digital transformation of business, has created a structural shortage of data center capacity, ensuring sustained growth in rental income streams and capital value. The global data center market is valued at $350B+ with an expected CAGR of 15–20% through 2030. For a large portfolio manager, investments in data centers represent a unique opportunity: a combination of stable cash flow from long-term contracts (Lease Terms 5–15 years) with exposure to the mega-trend of AI-driven compute demand.

Hyperscale Data Centers and AI-Driven Demand

Hyperscale data centers—facilities with a capacity of 100+ MW serving the largest cloud service providers (CSPs): Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform (GCP), Meta, Oracle, Apple. The hyperscale segment accounts for 40–50% of the global data center market and is growing faster than other segments—CAGR 25–30%. The key growth driver is the training and inference of large language models (LLMs): training a single model of the GPT-4 class requires 20,000–30,000 GPUs (NVIDIA H100/B200), consuming 50–100 MW of power over several months. Goldman Sachs and Morgan Stanley estimate that global data center power consumption will grow from 50 GW (2024) to 100–150 GW by 2030, equivalent to the electricity consumption of countries such as Japan or Germany.

AI compute demand is transforming data center economics: AI-ready data centers require higher power density—30–80 kW per rack versus 5–15 kW for traditional enterprise workloads; advanced cooling systems—liquid cooling and immersion cooling replace traditional air cooling; and significant investments in electrical infrastructure—transformer substations, backup generators, UPS systems (Uninterruptible Power Supply). This creates high barriers to entry and a price premium for AI-ready facilities: rental rates for AI workloads are 30–50% higher than for traditional enterprise workloads.

Market Leader Analysis: Equinix and Digital Realty

Equinix (NASDAQ: EQIX) is the world’s largest operator of colocation data centers with 260+ facilities in 72 cities and 33 countries. Market capitalization $75B+, Revenue $8B+, AFFO (Adjusted Funds from Operations) $3.5B+. Equinix is structured as a REIT (Real Estate Investment Trust), offering tax efficiency and mandatory distribution of 90%+ of taxable income as dividends. Key metrics for analyzing Equinix include: Interconnection Revenue—income from cross-connects (physical connections between clients within the data center)—a highly profitable (80%+ margin) and sticky revenue stream; MRR Churn Rate (Monthly Recurring Revenue Churn)—2–3% quarterly, one of the lowest figures in the industry; Same-Store Revenue Growth—increase of revenue at existing facilities excluding new openings; Utilization Rate—capacity occupancy, typically 78–85% (above 85% signals the need for expansion).

Digital Realty (NYSE: DLR) is the second largest operator with 300+ facilities in 50+ metro markets. Digital Realty differs from Equinix with a stronger position in the hyperscale segment: large-format lease deals with CSPs for 10–30 MW of power and lease terms of 10–15 years. Revenue $5.5B+, AFFO $2.5B+. The key metric is Backlog (order book, signed but not yet operational)—$3–5B, providing visibility of future growth for 2–3 years.

Other significant players: CyrusOne (acquired by KKR and GIP for $15B), QTS Realty (acquired by Blackstone for $10B), Vantage Data Centers (backed by DigitalBridge), EdgeConneX (backed by EQT Partners). Take-private trend: infrastructure private equity funds are aggressively acquiring public and private data centers, attracted by stable cash flows and the potential for AI-driven growth.

Co-location and Edge Computing

Colocation is a model in which the data center operator provides physical space (racks, cage, suite), power, and cooling, while the client installs its own server equipment. Colocation is the most common model for enterprise clients, financial institutions, and content providers.

Key benefits of colocation for clients:

Proximity (close to network-critical traffic exchange points—Internet Exchanges)
Redundancy (duplicate power supply and cooling—Tier III/IV standards of Uptime Institute)
Scalability (possibility to increase capacity without CAPEX into own infrastructure)
Ecosystem (access to hundreds of network operators and cloud providers within a single facility)

Edge Computing is a distributed architecture placing compute resources closer to end users to minimize latency. Edge data centers are compact facilities with 1–10 MW capacity, located in population centers (vs hyperscale facilities, often sited in remote locations with cheap energy).

Drivers of Edge Computing:

5G deployment (rollout of 5G networks requires edge infrastructure for data processing at base station level)
Autonomous vehicles (autonomous vehicles generate 4–30 TB of data per hour, requiring processing with low latency)

Power Infrastructure of Data Centers

Power infrastructure is a critical bottleneck for data center growth. Grid connection in major markets (Northern Virginia, London, Frankfurt, Singapore) takes 24–48 months due to shortage of transformer capacity and regulatory restrictions. Power infrastructure makes up 30–40% of total CAPEX for a data center.

Solutions include:

On-site power generation—natural gas, diesel generators, fuel cells
Power Purchase Agreements (PPAs) with renewable energy sources—solar and wind
Nuclear energy—Small Modular Reactors (SMRs) are considered a promising source of baseload power for hyperscale data centers (Microsoft signed a PPA with Constellation Energy to restart Three Mile Island)

Cooling infrastructure accounts for 30–40% of total data center energy consumption, measured by the PUE (Power Usage Effectiveness) index = Total Facility Power / IT Equipment Power. Ideal PUE = 1.0; the best hyperscale facilities reach PUE 1.1–1.2; industry average is 1.5–1.6.

Liquid Cooling is mandatory for AI workloads:

Direct-to-Chip (D2C) cooling delivers liquid directly to the processor
Rear Door Heat Exchangers (RDHx) are installed on the back of the rack
Immersion Cooling—servers are submerged in dielectric fluid, achieving PUE 1.03–1.05 and reducing cooling energy consumption by 90%

Strategy for UHNWI: invest in data centers through public REITs (Equinix, Digital Realty—Dividend Yield 2.5–4% + capital appreciation 8–12% = total return 10–16%); through infrastructure private equity funds (Brookfield, DigitalBridge, EQT—target IRR 15–20%); via direct co-investment in individual projects for large portfolios ($50M+).

Key Risk Factors:

Technology obsolescence (10-year lease may result in lock-in of obsolete asset)
Energy price volatility
Concentration risk (Northern Virginia accounts for 50%+ of US hyperscale capacity)