Digital Realty: Data Center REIT and the AI Infrastructure Build-Out Tailwind

Executive Summary

Digital Realty Trust (DLR) is the second-largest data center REIT globally, operating more than 300 data centers across 50 metropolitan areas in 25 countries, with approximately $5.5 billion in annual revenue in 2024. The company's PlatformDIGITAL framework — positioning its facilities as a connected data hub strategy rather than standalone colocation boxes — differentiates it from traditional colocation providers and positions it specifically for enterprise hybrid AI deployments. Digital Realty is a near-pure beneficiary of the AI infrastructure supercycle, competing directly with Equinix and a growing set of hyperscaler-owned campuses for AI workload hosting.

Digital Realty's AI margin pressure score is 2/10 — a strongly AI-positive business where the primary investment questions are execution quality and competitive positioning rather than disruption risk.

Business Through an AI Lens

Digital Realty's strategic evolution over the past five years has positioned it specifically for the AI infrastructure era. The company's acquisition of Interxion in 2020 for approximately $8.4 billion significantly expanded its European interconnection-dense campus footprint. Its joint venture with Blackstone — which acquired a $7.3 billion stake in a portfolio of U.S. and European data centers — provided capital for accelerated growth without proportional leverage increase. The ServiceDigital platform offers managed services on top of colocation infrastructure.

For AI workloads, Digital Realty's most relevant differentiator is its PlatformDIGITAL connected data hub architecture. The thesis is that enterprise AI requires not just compute co-located in a data center, but a connected fabric — low-latency links to cloud on-ramps, private networks, and data exchange partners in the same facility. A retail company building an AI-powered demand forecasting system needs co-location for its GPU servers, connectivity to its cloud AI training environment, and data exchange access to its retail data partners. Digital Realty's campus architecture attempts to provide all of these in a single venue.

The company's customer base includes a broad cross-section of AI adopters — cloud providers using DLR facilities as regional points of presence, enterprises building private AI infrastructure, and AI-native companies that need global data center presence before they can afford to build proprietary facilities. The cloud on-ramp density in DLR facilities — more than 20 cloud provider direct connections per major campus — is a compelling value proposition for enterprises running hybrid AI architectures.

Revenue Exposure

Digital Realty's revenue derives primarily from colocation and interconnection, with a mix of retail colocation (smaller footprints for enterprise customers), wholesale colocation (large footprints for hyperscalers), and interconnection services.

Digital Realty's revenue quality has improved as the company has focused more explicitly on retail colocation and interconnection versus pure wholesale hyperscale leasing. Wholesale leases generate lower margins and are more susceptible to hyperscaler build-versus-buy decisions. The company's shift toward retail colocation — smaller deployments for enterprise AI customers — improves revenue predictability and margin quality.

The geographic diversification of Digital Realty's portfolio — with meaningful revenue in EMEA (approximately 35% of total) and Asia-Pacific (approximately 12%) — provides exposure to AI infrastructure buildout in markets where data sovereignty regulations limit enterprises from using U.S.-headquartered hyperscalers. European data localization requirements are creating strong demand for European-operated AI infrastructure that Digital Realty is well positioned to serve.

Cost Exposure

Digital Realty's most significant cost challenge is the same as Equinix's: power. The AI compute buildout is dramatically increasing power density requirements in data center facilities. AI GPU clusters require 30-80 kilowatts per rack versus the industry average of 8-12 kilowatts for traditional enterprise workloads. This density gap requires significant investment in power infrastructure, cooling systems (particularly liquid cooling), and backup power capacity.

The company has committed to capital expenditure of approximately $3.0-3.5 billion annually through 2026 to build new capacity and upgrade existing facilities for high-density AI workloads. This level of investment is necessary to remain competitive but requires careful financing management given the company's approximately $16 billion in long-term debt.

Energy costs run approximately $800 million annually and represent a meaningful operating cost pressure as power prices have risen in key markets. Digital Realty's renewable energy commitments — targeting 100% renewable energy by 2030 — provide some protection through long-term power purchase agreements but also require upfront investment in renewable procurement and certificates.

AI is also being deployed within Digital Realty's operations. The company uses machine learning for data center infrastructure management — predicting cooling equipment failures, optimizing power distribution, and managing cooling efficiency dynamically based on workload characteristics. These tools reduce both operating costs and downtime risk, though the dollar impact is measured in tens of millions rather than hundreds of millions annually.

Moat Test

Digital Realty's competitive moat is strong but slightly less deep than Equinix's due to its heavier mix of wholesale leasing, which is more commoditizable than interconnection-dense retail colocation.

The location moat is shared with Equinix — DLR's campuses in Northern Virginia (the world's largest data center market), Silicon Valley, Amsterdam, Singapore, and other prime markets are difficult to replicate due to power availability constraints and permitting complexity. In Northern Virginia specifically, Digital Realty is one of the largest landowners in the data center corridor and has significant land bank for future development.

The PlatformDIGITAL connected hub architecture creates customer switching costs through the integration of interconnection, cloud on-ramps, and colocation services. An enterprise that has built its AI architecture using Digital Realty's cloud connectivity and data exchange services faces meaningful migration costs to move to an alternative provider.

The Blackstone joint venture structure is a financial innovation that functions as a competitive moat by allowing Digital Realty to fund growth at a lower cost of capital than standalone debt financing would permit, enabling faster capacity expansion than competitors with more constrained balance sheets.

Timeline Scenarios

1-3 Years (Near Term)

Near-term dynamics are dominated by the extraordinary pace of hyperscaler AI infrastructure investment. Microsoft, Google, Amazon, and Meta are collectively committing more than $250 billion in capital expenditure in 2025-2026, with a significant portion going to data center capacity. Digital Realty is well positioned to capture a share of this investment through its existing relationships and campus positions. The primary near-term execution challenge is power — securing utility commitments for new capacity in constrained markets. The company's land bank in Northern Virginia and its relationships with utilities provide some assurance, but power procurement timelines are measured in years, not months.

3-7 Years (Medium Term)

The medium-term scenario involves Digital Realty's success or failure in capturing enterprise AI colocation demand. The most attractive long-term opportunity is not serving hyperscalers directly — they are increasingly building their own facilities — but serving the wave of enterprise companies deploying private AI infrastructure at significant scale. A Fortune 500 manufacturer deploying AI-driven predictive maintenance and supply chain optimization may need 2-5 megawatts of dedicated compute capacity; a major financial institution deploying AI risk management systems may need 10-20 megawatts. These enterprise deployments are precisely what Digital Realty's retail colocation and interconnection platform is designed to serve.

7+ Years (Long Term)

The long-term scenario involves the potential emergence of AI-specific data center architectures that diverge meaningfully from current colocation models. Liquid immersion cooling, photonic computing, and neuromorphic processor architectures could require facility designs that current data centers do not support. Digital Realty's significant investment in existing campus infrastructure creates some option value for adaptation, but the company will need to ensure its new development pipeline incorporates next-generation architectural capabilities from the start.

Bull Case

In the bull case, Digital Realty successfully executes its PlatformDIGITAL strategy and captures a disproportionate share of enterprise AI infrastructure spending. Revenue grows at 10-12% annually through 2030, reaching $9-10 billion. The Blackstone JV structure is extended and expanded, enabling capital-light growth in additional markets. European data sovereignty requirements drive strong demand for DLR's European portfolio, with EMEA revenue growing faster than the Americas segment. Interconnection revenue — the highest-margin component — doubles as AI data exchange volumes compound across the platform. The stock re-rates toward Equinix's valuation multiple as investors recognize DLR's differentiated connected hub positioning.

Bear Case

In the bear case, hyperscaler vertical integration of data center infrastructure accelerates, with Microsoft, Amazon, and Google building more proprietary campuses and reducing their reliance on third-party colocation. Wholesale leasing revenue faces pricing pressure as hyperscalers gain negotiating leverage. Power procurement delays in key markets force Digital Realty to defer new capacity builds, losing market share to more power-secure competitors. The company's European portfolio faces margin compression from energy cost inflation and regulatory compliance costs. Leverage at approximately 6x EBITDA becomes a constraint as interest rates remain elevated, limiting the company's ability to invest in high-density AI upgrades.

Verdict: AI Margin Pressure Score 2/10

Digital Realty scores 2 out of 10 on the AI margin pressure scale — a strongly AI-positive business where the disruption risk is essentially zero and the growth opportunity is substantial. The company's global data center portfolio, connected hub architecture, and established customer relationships position it as a core beneficiary of the AI infrastructure supercycle. Investors should focus on execution — specifically on power procurement, enterprise colocation market share, and capital allocation discipline — rather than disruption risk.

Takeaways for Investors

• Digital Realty is a direct beneficiary of the AI infrastructure supercycle — every hyperscaler and enterprise AI deployment generates demand for data center capacity that DLR is positioned to serve.
• The PlatformDIGITAL connected hub strategy differentiates DLR from pure colocation competitors by integrating interconnection and cloud on-ramps — this is the right architecture for enterprise hybrid AI deployments.
• Power procurement is the binding constraint — monitor utility partnership announcements and power availability disclosures in Northern Virginia, Silicon Valley, and key European markets.
• The wholesale-to-retail mix shift is strategically important; retail colocation carries higher margins and lower hyperscaler concentration risk — track retail leasing as a percentage of new bookings.
• The Blackstone JV structure reduces balance sheet leverage while maintaining economic exposure to portfolio growth — evaluate JV asset performance as a proxy for non-consolidated value creation.
• At current valuations that discount DLR at a meaningful premium to book value, the AI infrastructure growth thesis is partially priced in — margin of safety depends on execution quality over the next 2-3 years.