Data Center REIT Landscape: Equinix and Digital Realty

Introduction

Two public REITs sit at the center of investable data center real estate, and the most common mistake is to treat them as interchangeable bets on the same theme. Equinix and Digital Realty both ride the AI and cloud tailwind, but they monetize it through nearly opposite models. Equinix sells connection and ecosystem, packing thousands of customers into densely interconnected campuses and charging for the network effects that result. Digital Realty sells scale, building the large, power-dense blocks of capacity that hyperscalers consume by the dozens of megawatts. Understanding that divergence is the difference between a candidate who can name the two tickers and one who can explain why they trade differently.

The cleanest way to hold the comparison in mind is a side-by-side of the strategy, not just the size:

Both are genuine REITs, distributing taxable income to shareholders and benefiting from the pass-through structure detailed in the broader explanation of what a REIT is and why the structure exists. What separates them is where in the hyperscale, wholesale, and retail colocation stack each chooses to compete.

Equinix carries a market capitalization of roughly $101 billion, making it the most valuable data center REIT despite owning a smaller physical footprint than its rival. The reason is the business it actually runs. Equinix operates more than 270 data centers across 36 countries and serves over 10,000 customers, including roughly 60% of the Fortune 500. Its defining asset is not the buildings but the Fabric of interconnections among the tenants inside them: more than 499,000 metro interconnections that let customers privately exchange traffic with cloud platforms, networks, and each other.

Interconnection

A direct, private physical or virtual connection between two parties inside or across data centers, bypassing the public internet. Interconnection revenue is higher-margin and stickier than space-and-power rent, because once a customer wires into an ecosystem, the cost and risk of moving rise with every new connection.

That ecosystem is what gives Equinix pricing power and low churn. A customer who has built dozens of interconnections to partners, clouds, and carriers inside an Equinix campus faces real switching costs to leave, which shows up as durable pricing and a long runway of organic growth. Equinix's adjusted funds from operations per share compounded at roughly a 9% annual rate from 2020 through 2024, and management has guided to 8% to 11% growth, a profile that looks more like a compounding platform than a static landlord.

Build Bolder: From Compounder to Builder

For most of its history Equinix grew like a capital-light compounder, adding interconnections and raising prices inside campuses it already owned. The AI wave has pushed it to spend far more aggressively under a strategy it calls Build Bolder, accelerating capacity delivery and acquiring land for large new campuses. The scale of that pivot is visible in the guidance: Equinix expects 2025 AFFO of roughly $3.73 billion to $3.81 billion, up 11% to 14%, on adjusted EBITDA near $4.5 billion at a 49% margin, and has guided 2026 AFFO to about $4.16 billion to $4.24 billion on total capital expenditure of roughly $3.7 billion to $4.2 billion. When Equinix reported record bookings and a bullish 2026 outlook in February 2026, the stock jumped about 11% in a session, a sign the market rewarded the heavier spend as long as AI demand kept filling the new capacity. The risk in Build Bolder is the mirror image of its promise: more capital deployed into development is more capital exposed if hyperscaler demand cools, which is why investors now watch Equinix's capex and development yields the way they once watched only its interconnection adds.

The breadth of that customer base is also a structural hedge. Where a pure wholesale owner depends on a handful of enormous tenants, Equinix spreads its revenue across more than ten thousand customers, which softens the tenant-concentration risk that dominates the rest of the sector. That matters because the same hyperscalers driving the demand surge are also the entire counterparty universe for wholesale capacity, a concentration explored in the article on the hyperscaler counterparty universe. Equinix's enterprise and network customers give it a revenue base that does not rise and fall with five cloud budgets.

Digital Realty is the larger operator by footprint, with more than 300 data centers across 50 metros, and it is built for the opposite job: delivering the massive, capital-intensive blocks of capacity that the AI supercycle demands. Where Equinix optimizes for connection density, Digital Realty optimizes for deliverable megawatts, and its value capture comes from sheer physical scale rather than network effects.

Wholesale Colocation

Large blocks of data center space and power, typically a megawatt or more, leased to a single tenant on long terms, with the tenant operating its own equipment. It contrasts with retail colocation, where many smaller customers share a facility and pay largely for interconnection and modest footprints.

The distinction is not academic. A wholesale tenant signs for capacity, brings its own gear, and treats the building as conditioned power; a retail customer is buying proximity to an ecosystem. Digital Realty's portfolio skews heavily toward the former, which is what makes it the natural counterparty for hyperscalers building AI capacity at scale.

That positioning showed up directly in its 2025 results. Digital Realty signed a record 200-megawatt lease in Charlotte and reported a substantial leasing backlog that gives clear revenue visibility into 2026 and beyond, exactly the kind of large, lumpy commitment that defines wholesale. The company has also leaned into sustainability as a competitive feature, contracting roughly 1.5 gigawatts of renewable capacity and running its European and North American productized colocation portfolios on 100% renewable energy, which matters because hyperscaler tenants increasingly write clean-power requirements into their site selection.

Funding the Wholesale Build-Out

Funding that build-out is the central financial challenge of the model. Wholesale capacity is enormously capital-intensive, and the sums required to deliver hundreds of megawatts exceed what retained cash flow can cover, as the underlying build costs and time-to-power math make clear. Digital Realty has therefore turned to joint ventures, hyperscale development partnerships, and capital recycling (selling stabilized assets to fund new development) to keep its pipeline moving without overleveraging the balance sheet. The result is a business whose returns come from continuously developing and stabilizing new capacity, which is why investors watch its development yields and funding plan as closely as its in-place rents.

That funding strategy took concrete shape in 2025. Digital Realty closed its first U.S. hyperscale data center fund, raising roughly $3.25 billion of institutional equity commitments while retaining a 20% ownership interest and serving as manager, which lets the company develop large campuses alongside institutional partners while keeping leverage in check. The demand side cooperated: the company posted record bookings in 2025, including a quarterly record near $96 million in its smaller-deal and interconnection segment, and ended the year with a record $1.4 billion backlog of signed-but-not-commenced leases. That backlog underwrites 2026 guidance of $7.90 to $8.00 in core funds from operations per share, roughly 8% growth over 2025. A record backlog plus a fresh pool of third-party capital is exactly what lets a wholesale developer keep building into a supply-starved market without straining its own balance sheet.

Neither company can ignore the hyperscalers, who take the largest blocks of capacity, but they reach them differently. Digital Realty competes for hyperscale wholesale directly, both on balance sheet and through joint-venture capital. Equinix, whose core model is retail, built a separate vehicle to play in hyperscale without diluting its interconnection focus: the xScale program, run through joint ventures so that the heavy capital sits largely off Equinix's balance sheet.

The scale of that program has grown sharply. Equinix originally partnered with GIC, Singapore's sovereign wealth fund, on European xScale facilities, and in October 2025, as part of its Build Bolder push, expanded into a joint venture with GIC and CPP Investments intended to raise more than $15 billion of capital, nearly tripling the program and targeting multiple US campuses each with hundreds of megawatts of capacity. xScale has now leased roughly 430 megawatts across more than twenty operational facilities in over a dozen metros. The structure is deliberate: it lets Equinix serve hyperscale core workloads adjacent to its existing campuses while keeping the capital intensity, and the lower-margin wholesale economics, in a partnership rather than on its own books. The same sovereign and pension capital that funds these vehicles is profiled in the discussion of in-house teams at sovereigns and pensions, and it is no accident that the largest such investors anchor the largest data center JVs.

Training Favors Wholesale, Inference Favors Interconnection

The two models also map onto the two phases of the AI workload. Model training concentrates enormous, power-dense clusters into a few large facilities, which is squarely Digital Realty's wholesale game and the reason its biggest leases now run to hundreds of megawatts. Inference, the serving of trained models to users, is latency-sensitive and distributed, which plays to Equinix's dense, metro-edge interconnection footprint: an AI application that needs to sit close to users and wire privately into multiple clouds is exactly the deployment Equinix's Fabric was built for. As inference is widely expected to overtake training in volume later this decade, Equinix's bet is that the AI build-out will not only fill giant training campuses but also thicken the interconnection mesh in the metros where models are actually consumed. The same demand surge therefore reaches the two REITs through different doors, and an analyst has to underwrite each against the workload phase it is best positioned to capture.

That split also tempers the biggest fear hanging over both stocks. Through 2025 there were recurring reports of hyperscalers pausing or reshuffling some leasing commitments, raising the question of whether the AI build-out had run ahead of real demand. Both REITs answered with record results: leasing and backlog hit all-time highs even as individual tenants slowed at the margin, which suggests the demand base is broad enough to absorb the occasional pause from any single counterparty. Whether that holds through a genuine capex pullback is the open question, but so far the data has favored the builders over the skeptics.

The binding constraint on both companies has shifted from real estate to electricity. The buildings are relatively straightforward to construct; securing hundreds of megawatts of reliable power, on grids with multi-year interconnection queues, is not. Access to power now drives site selection more than land or fiber does, which is why both REITs land-bank sites with secured power capacity and increasingly pursue on-site generation and long-term clean-energy contracts. A campus with committed power is worth far more than an identical one still waiting in an interconnection queue, and the power constraint on data centers has become the variable that now dominates underwriting new supply.

For valuation, both trade on funds-from-operations and AFFO multiples rather than earnings, following the logic laid out in how REITs trade on FFO and AFFO multiples. The premium tends to track the durability and growth of cash flow: Equinix's interconnection stickiness and double-digit AFFO growth support a premium multiple, while Digital Realty's wholesale model carries more lease lumpiness and development risk but offers more direct leverage to the raw megawatt build-out. An investor choosing between them is really choosing between a compounding ecosystem and a scale-and-development play. The growth gap shows up in the guidance: Equinix pointed to low-double-digit AFFO-per-share growth for 2026 while Digital Realty guided to roughly 8% core FFO growth, which helps explain why Equinix carries the richer multiple of the two.

There is also an income dimension that the AI-growth story tends to overshadow. Both are investment-grade REITs that pay and grow dividends, but the blend of yield and growth differs: Digital Realty has historically offered a somewhat higher current yield to compensate for its lumpier, more capital-intensive model, while Equinix pairs a lower starting yield with faster per-share growth. Because both fund enormous development pipelines, the durability of those dividends rests on AFFO payout discipline and on raising development capital efficiently, which is why a credible total-return view weighs not just the growth rate but the funding plan behind it. A REIT that grows AFFO at a double-digit clip while issuing equity or selling assets at unattractive prices can hand shareholders far less than the headline growth implies, and that gap between reported growth and per-share value creation is exactly what separates a real coverage view from a momentum thesis.

It is also worth remembering that the two names are now nearly the entire pure-play public set. A wave of take-privates thinned the listed cohort:

What is left on the public market is Equinix, Digital Realty, a diversified owner like Iron Mountain with a fast-growing data center segment, and, newly, Blackstone's listed digital infrastructure trust. That privatization wave means public investors compete for the asset class against deep-pocketed private platforms that never appear in a public comp set, which both validates the sector and caps how much of it the public markets can actually own. The reopening of the public data center pipeline is the counter-current to that privatization, and the reason the listed cohort may finally start growing again.

In the end, "data center REIT" is not a single category. It contains at least two distinct business models with different cash-flow shapes, different capital structures, and different sensitivities to the AI demand surge, and any credible coverage view has to start by separating them.