How Digital Realty Built a $60 Billion Empire on the One Thing Every Digital Business Needs

Key Takeaways
- Digital Realty grew from acquired telecom switching facilities in 2004 to a $60B REIT with over 300 data centres across 50 global metros.
- Data centres behave more like power plants than traditional real estate, featuring 10-15 year leases and near-zero tenant churn.
- The AI inflection has created twin demand curves: location-insensitive training campuses and latency-sensitive urban-adjacent inference nodes.
From a handful of acquired telephone switching centres in 2004 to the largest data centre REIT on the planet, Digital Realty's growth is one of the most consequential case studies in commercial real estate — and a blueprint for the next generation of infrastructure investors.
The Insight That Started Everything
In 2004, Digital Realty Trust began acquiring former telephone company buildings and industrial sites across major U.S. cities, converting them into facilities that could house the servers of the earliest cloud-era internet companies. The founding insight wasn't really a real estate insight — it was a technology insight wrapped in one: as computing infrastructure migrated from company premises to third-party facilities, demand for the physical space, power, and cooling to house it would grow continuously, in ways existing real estate frameworks weren't built to capture.
Roughly 100 GW of new data centre capacity is expected between 2026 and 2030, equating to an estimated $1.2 trillion in real estate value creation. Digital Realty positioned itself early to capture a disproportionate share — not by being first to spot the opportunity, but by being the most disciplined in building the platform to serve it at scale.
Why Data Centres Behave Differently From Every Other Asset Class
A conventional office building is a container for human activity, valued on location, fit-out, and occupier demand. A data centre is closer to a power plant than a building — its value hinges on electrical capacity, cooling infrastructure, fibre connectivity, and resilience against outages. Tenants aren't moving in desks; they're installing hundreds of millions of dollars in computing equipment they intend to run continuously for years.
That specialisation produces a tenancy structure unlike anything else in commercial real estate: leases running 10 to 15 years with built-in escalations, and churn rates that are extremely low once a hyperscale tenant has deployed at scale — the cost of relocating computing infrastructure is simply too high.
The Platform Strategy
Digital Realty's growth has consistently prioritised platform scale over individual asset quality — a reversal of the usual real estate logic. The reason is connectivity: hyperscale cloud infrastructure functions as a single interconnected network, not a collection of independent buildings. Digital Realty's footprint of more than 300 data centres across 50 metros and 25 countries, unified under its PlatformDIGITAL architecture, gives it a connectivity advantage that single-market operators can't replicate. Every facility added increases the value of the rest of the network — and every hyperscale customer deployed across multiple sites becomes progressively more expensive to move.
Navigating the Grid Constraint
Digital Realty faces the same power constraint as every other operator, and it's managing it on three tracks: geographic diversification into markets with available power (Columbus, Indianapolis, Kansas City, Reno, and away from saturated hubs like Northern Virginia and Silicon Valley); behind-the-meter generation, including on-site solar, battery storage, and fuel cells; and long-term power purchase agreements with renewable and nuclear generators that lock in price certainty and the clean-energy credentials hyperscale tenants increasingly require.
Its ServiceFabric joint ventures with sovereign wealth funds, pension investors, and infrastructure capital partners allow it to co-invest in development pipelines — sharing capital requirements and risk while retaining management fees and long-term economic upside.
The AI Inflection
Cloud migration drove a decade of consistent demand growth for Digital Realty. AI has layered a second, distinct demand curve on top of it. AI training workloads are location-insensitive and have driven hyperscale campuses to power-rich regions regardless of proximity to population centres. AI inference workloads are the opposite: latency-sensitive, continuous, and geographically distributed — because users everywhere are making requests in real time. That shift is creating demand for a new category of smaller, urban-adjacent, high-density data centres — a category Digital Realty's existing portfolio is well positioned to serve.
The Legacy
Digital Realty's twenty-year thesis — that computing infrastructure migration would create durable, growing demand for specialised real estate — has been vindicated well beyond what even its earliest investors projected. It didn't build the AI models or cloud platforms. It built the buildings those platforms run in, at a scale and global reach that made it indispensable to the organisations defining the digital economy.
The 2026 outlook for U.S. data centre real estate rests on four interdependent pillars: power, capital, connectivity, and community. Digital Realty has treated them as a single equation for two decades. The market is still catching up.
Frequently Asked Questions
How did Digital Realty scale to become a $60 billion data center REIT?
By recognizing early that computing migration would require massive specialized physical infrastructure, prioritizing platform network connectivity across 300+ facilities globally over isolated real estate assets.
What is the difference between AI training and AI inference data center requirements?
AI training workloads require massive power and are location-insensitive, while AI inference workloads require low latency and must be geographically distributed near population centers.



