Digital Realty AI Demand Sets Records, but the Backlog Raises the Execution Stakes

Digital Realty AI demand produced another record in the second quarter of 2026, despite a sharp shift in the company’s leasing mix. Bookings from deployments of up to one megawatt and interconnection reached $108 million at Digital Realty’s share. That category crossed $100 million for the first time.

The larger story is not another forecast about AI consuming more computing capacity. Digital Realty now has signed commitments, a record backlog, and two substantial hyperscale leases completed after the quarter. The company also raised its full-year operating outlook for the second time this year.

That progress creates a tougher test. Signed leases must become operational data halls with available electricity, completed construction, connected networks, and paying customers. Digital Realty’s backlog offers unusual revenue visibility, but it also exposes the company to delays that bookings alone cannot resolve.

Equinix provides the clearest competitive reference. It also reported stronger bookings, record interconnection activity, and higher 2026 guidance. Both operators are benefiting from customers seeking infrastructure for cloud services, enterprise workloads, and artificial intelligence.

The emerging contest is therefore not simply about who can sign the largest AI lease. It is about which operator can convert demand into energized capacity without weakening returns or overextending its balance sheet.

Digital Realty Bookings Reveal a Broader Demand Shift

The quarter’s most important record came from smaller deployments and connectivity, not from one enormous AI campus.

Digital Realty signed second-quarter bookings expected to generate $307 million in annualized GAAP base rent at full ownership. Its economic share was $208 million. The company’s second-quarter results attributed $88 million to deployments between zero and one megawatt and another $20 million to interconnection.

Interconnection refers to direct links between companies, networks, and cloud platforms inside or between data centers. These links help customers move data without routing every exchange through the public internet.

Together, those two categories represented 52% of Digital Realty’s share of quarterly bookings. The company also added 142 new customer logos. This combination matters because smaller deployments and network connections can diversify revenue beyond a few giant hyperscale tenants.

The result followed a strong first quarter. Digital Realty had reported $98 million of bookings from the same zero-to-one-megawatt and interconnection categories. That earlier figure already represented a record before the second quarter surpassed it.

However, the total quarterly bookings figure fell from the exceptional first-quarter level. Digital Realty signed $707 million at full ownership during the first quarter, including the largest hyperscale lease in its history. Its share of those bookings was $423 million.

That comparison helps separate the signal from the headline. The second quarter did not set a company-wide record for all bookings. It set a record in the smaller-capacity and interconnection segment, while maintaining substantial leasing across the platform.

The distinction supports a more balanced reading of Digital Realty AI demand. Customers are not only reserving massive training clusters. They are also placing workloads closer to networks, cloud on-ramps, data sources, and other companies.

AI inference helps explain why this mix could matter. Inference is the process of using a trained model to generate an answer, classification, or prediction. Unlike training, inference can occur across many locations as applications reach more users.

Digital Realty has positioned its portfolio around three connected businesses: colocation and connectivity, hyperscale capacity, and private capital. The second-quarter figures suggest that demand is reaching all three rather than remaining isolated within giant, single-tenant facilities.

The company’s July leasing activity strengthened that interpretation. After the quarter ended, Digital Realty signed two hyperscale leases representing $410 million of annualized GAAP base rent at full ownership. Its share was $205 million.

Those agreements show that large-scale demand did not disappear when smaller bookings took the quarterly spotlight. Instead, the company entered the third quarter with evidence of demand at both ends of its capacity range.

Raised Digital Realty Guidance Reflects More Than AI Optimism

Higher guidance gives the bookings story financial weight, although several nonrecurring items complicate the headline earnings comparison.

Digital Realty raised its 2026 Core Funds From Operations outlook, excluding net promote income, to between $8.15 and $8.20 per share. Core FFO is a real estate cash-flow measure that adjusts net income for items such as property depreciation and selected noncore effects.

The company had started 2026 with guidance between $7.90 and $8.00 per share. It increased that range to $8.00 through $8.10 after the first quarter, then lifted it again in July.

Constant-currency Core FFO guidance, excluding net promote income, also increased. The new range was $8.10 to $8.15 per share, compared with the initial $7.90 to $8.00 outlook.

Revenue guidance excluding promote income rose to a range of $6.85 billion through $6.95 billion. Digital Realty had started the year expecting $6.60 billion through $6.70 billion.

The company also increased its adjusted EBITDA outlook to between $3.75 billion and $3.85 billion. Its February range had been $3.60 billion through $3.70 billion.

These revisions followed second-quarter revenue of $1.9 billion. Revenue increased 18% sequentially and 29% from the prior-year period. Adjusted EBITDA reached $978 million, up 6% from the previous quarter and 19% year over year.

Core FFO excluding net promote income was $2.13 per share. That compared with $2.04 in the first quarter and $1.87 one year earlier.

Investors should still distinguish underlying operations from transaction-related income. Digital Realty recognized $188 million of net promote income connected with the development and leasing of three joint-venture data centers. Promote income is an incentive allocation that a managing partner can earn after an investment meets agreed performance conditions.

The company also recognized a $94 million insurance settlement after tax concerning a previously disclosed 2024 matter. Approximately $27 million entered Core FFO as business-interruption recovery, while property-damage proceeds were excluded.

Digital Realty’s preferred measure removes net promote income when presenting its revised outlook. That exclusion makes the guidance more useful for evaluating recurring operations, but it does not eliminate every judgment involved in a non-GAAP metric.

Renewal pricing provides a separate operating signal. Cash rental rates on second-quarter renewals increased 25.4%, while GAAP renewal rates rose 32%. Management subsequently raised its full-year cash renewal-rate assumption to a range of 9% through 11%.

Those spreads suggest that constrained capacity and customer demand are supporting stronger economics on existing space. They also reduce the need to rely entirely on new construction for growth.

Still, higher renewal rates do not establish that AI generated every incremental dollar. Digital Realty serves cloud platforms, network providers, software companies, financial institutions, and other enterprises. Management combines cloud, digital transformation, and AI when describing the demand environment.

The most defensible conclusion is narrower. Digital Realty AI demand is contributing to stronger leasing, while broader cloud and connectivity needs are improving the economics of the existing portfolio.

Digital Realty AI Demand Is Becoming a Conversion Test

A record backlog turns customer interest into contracted visibility, but it does not become revenue until facilities are ready and leases commence.

Digital Realty ended June with $1.9 billion of annualized GAAP base rent in signed but uncommenced leases at full ownership. Its economic share was $1.4 billion, up from approximately $1.0 billion at the end of March.

A backlog contains contracts for capacity that customers have reserved but have not started using. The figure can provide visibility into future revenue, especially when customers sign leases before a facility is completed.

It also creates a delivery obligation. Digital Realty reported a nine-month weighted-average period between signing and contractual commencement for second-quarter leases. The first-quarter average had been 19 months, partly reflecting the scale of its largest hyperscale agreement.

The shorter second-quarter period should bring a larger portion of new bookings into revenue sooner. However, the overall backlog includes commitments scheduled across several years, and actual commencement dates can change.

Digital Realty had approximately 1,402 megawatts under construction at the end of June. That represented an 82% increase from December 2025. About 54% of this capacity was preleased, according to the company’s quarterly filing.

Preleasing lowers demand risk because customers commit before a project opens. Yet it increases the consequences of a construction or utility delay. A contracted tenant cannot begin paying for capacity that lacks power, cooling, equipment, or final approvals.

The company’s development pipeline extends beyond current construction. Its earnings materials identified about 9 gigawatts of future development capacity and approximately 3 gigawatts of installed capacity. More than 5 gigawatts of the future portfolio sat in locations with at least 100 megawatts of buildable capacity.

Digital Realty also added two gigawatts of potential capacity in the Kansas City area. The location offers another market for cloud and AI deployments, but raw development capacity is not equivalent to immediately available power.

That difference defines the central mechanism behind the growth story. Customer demand produces bookings. Bookings support construction and financing decisions. Completed facilities convert the backlog into rent, and connected customers can then generate additional interconnection revenue.

Failure at any stage delays the conversion. A utility can postpone energization. Equipment suppliers can miss delivery schedules. Construction costs can rise. Customers can alter deployment plans, even when contracts provide some protection.

Digital Realty expects to spend between $2.8 billion and $3.3 billion on capital projects during the remainder of 2026. It also reported approximately $4.1 billion in open construction commitments at June 30, including reimbursable amounts.

The company estimated an 11.5% stabilized cash yield on projects under construction. That estimate uses expected investment costs and projected operating income, including assumptions about future market conditions. Digital Realty explicitly cautioned that actual costs, completion timing, and yields can differ.

This is why record Digital Realty bookings should not be evaluated as a final result. They are the beginning of a multiyear conversion process with measurable operational checkpoints.

For customers, the process affects more than a landlord’s earnings. AI infrastructure buyers need confidence that reserved computing environments will receive sufficient power, network access, and cooling when their deployment schedules require them.

For knowledge workers evaluating infrastructure announcements, separating bookings from active capacity is essential. A searchable AI knowledge base can help teams preserve those distinctions across filings, technical plans, and vendor updates.

Equinix Faces the Same Race for Connected Capacity

Digital Realty’s results show strong execution, but they do not describe a market where one operator has captured AI infrastructure demand alone.

Equinix reported annualized gross bookings growth of 23% year over year in its second quarter. The company called the result its second-highest bookings volume and said it contributed to a record backlog.

Equinix also added a record 9,700 net interconnections during the quarter. Monthly recurring revenue increased 11% on both a reported and normalized constant-currency basis.

Like Digital Realty, Equinix raised its full-year guidance. Its quarterly release cited stronger demand, bookings, presales, and execution.

The parallel results make Equinix the most useful primary opponent for this story. Both companies operate large global platforms, sell colocation and connectivity, and support cloud and AI deployments. Both are presenting record backlog as evidence of future growth.

Their portfolios are not identical. Digital Realty has emphasized a continuum from smaller colocation deployments through large hyperscale campuses. It also uses joint ventures and private capital to expand while sharing development investment.

Equinix has built a particularly dense interconnection platform serving enterprises, networks, and cloud providers. Its record quarterly connections underscore the importance of proximity between computing systems, data, and users.

AI workloads can strengthen both approaches. Large model training supports demand for high-capacity facilities. Enterprise adoption and inference can increase demand for distributed infrastructure near existing networks and company data.

The competitive pressure therefore appears in several layers. Operators must secure power and land before customers need them. They must fund construction without allowing capital costs to overwhelm returns. They must also build locations where customers gain access to useful networks and counterparties.

Digital Realty’s record bookings in the zero-to-one-megawatt and interconnection category address that last requirement. The figure suggests its connected facilities can participate as AI moves from centralized training into production applications.

However, Equinix’s record interconnection additions show that this opportunity is contested. Customers can also deploy through cloud providers, specialized AI infrastructure companies, or facilities operated by other large data center landlords.

Another competitive question concerns capital efficiency. Hyperscale leases can produce long-term contracted revenue, but the facilities require substantial upfront investment. Smaller colocation deployments can carry different margins and may create more opportunities for network-related revenue.

Digital Realty’s strategic private capital business attempts to bridge these models. The company can develop large assets with partners, earn fees, and recycle capital while retaining exposure to operating performance.

During the second quarter, Digital Realty agreed to acquire Columbia Capital, an investment firm focused on digital infrastructure. The transaction was expected to expand assets under management to approximately $9 billion after completion, subject to regulatory approval and other conditions.

Digital Realty also increased its ownership in three fully leased hyperscale facilities in Northern Virginia. Those investments add operating exposure to existing capacity rather than waiting for every new project to reach completion.

None of these moves establishes a permanent lead. Equinix is also expanding, reporting stronger presales, and raising its longer-term outlook. The competition will be decided through conversion speed, network density, customer retention, and returns on invested capital.

What the Record Backlog Does Not Guarantee

The strongest challenge to Digital Realty’s outlook is not weak demand today, but the physical and financial difficulty of delivering capacity on schedule.

Electricity is the first constraint. AI servers concentrate substantial computing and cooling requirements inside each facility. A suitable parcel of land does not become usable data center capacity until utilities can provide dependable power.

The International Energy Agency expects electricity generation serving data centers to rise from 460 terawatt-hours in 2024 to more than 1,000 terawatt-hours in 2030. Its energy outlook identifies AI as a central demand driver, while emphasizing uncertainty around efficiency and adoption.

That growth does not distribute evenly. Data centers cluster in markets with network connectivity, customers, land, and skilled workers. Local grid limitations can become more important than the total electricity available across a country.

Digital Realty’s regulatory filings identify power availability and connectivity disruptions as material risks. The company also warns about development delays, cost overruns, increased competition, customer concentration, and dependence on external financing.

Those disclosures are standard for a public data center company, but they map directly onto the current backlog. Digital Realty must deliver 1,402 megawatts already under construction while preparing a much larger future pipeline.

Capital is the second constraint. Digital Realty reported $1.63 billion of capital expenditures, excluding indirect costs, during the first half of 2026. Development projects accounted for approximately $1.48 billion of that amount.

The remaining 2026 spending plan can change with demand, financing availability, and leasing results. Higher interest rates or weaker capital markets can alter returns, particularly when projects require years of investment before reaching stabilization.

The company has used joint ventures, asset sales, equity issuance, and debt to fund expansion. These options can extend its development runway, although each comes with tradeoffs involving ownership, dilution, leverage, or execution complexity.

Customer concentration creates another uncertainty. Large hyperscale agreements provide significant contracted revenue, but a small number of tenants can influence a project’s economics. Changes in a major customer’s AI strategy could affect future leasing, even if existing contracts remain enforceable.

There is also a measurement problem. Digital Realty does not disclose a complete division between AI-specific bookings and conventional cloud or enterprise demand. Its facilities often support several workload types, and customers can repurpose capacity over time.

That ambiguity does not invalidate the growth. It limits how confidently observers can attribute every record to generative AI. The broader evidence supports rising demand for digital infrastructure, with AI serving as an important contributor rather than the only cause.

Renewal spreads also deserve careful treatment. A 25.4% cash increase is favorable for the landlord, but one quarter can reflect the timing and composition of expiring leases. Investors need several reporting periods before assuming that pace represents a durable annual norm.

Backlog growth carries similar limits. A rising backlog strengthens revenue visibility only if commencement schedules remain credible. Delays can push expected rent into later periods without necessarily canceling a contract.

The company’s raised Digital Realty guidance indicates that management expects operating momentum to overcome these pressures during 2026. It does not remove the construction, power, financing, or attribution risks that extend beyond one fiscal year.

Three Signals Will Decide What Happens Next

The next stage of Digital Realty AI demand will be measured by backlog conversion, energized construction, and repeatable leasing across customer sizes.

The first signal is the movement of signed leases into commenced revenue. Digital Realty’s next quarterly report should show how much backlog entered service and whether estimated commencement dates changed.

A healthy conversion would reinforce management’s claim that the backlog supports several years of growth. A rising backlog paired with repeated delays would weaken the case, even if new bookings remained strong.

Readers should watch Digital Realty’s share of backlog rather than only the full-ownership number. Joint ventures allow the company to expand efficiently, but the share attributable to Digital Realty better reflects its direct economic participation.

The second signal is construction delivery. The company should disclose changes in megawatts under development, preleasing rates, expected stabilized yields, and remaining capital requirements.

Capacity under construction grew 82% between December 2025 and June 2026. Maintaining schedules across that larger portfolio will test procurement, contractors, utilities, and internal project management.

Power availability deserves particular attention. Announcements about land or buildable capacity matter less than firm energization schedules. Utility commitments and completed substations provide stronger evidence than a theoretical site pipeline.

The third signal is the leasing mix. Digital Realty should demonstrate that its record smaller-deployment and interconnection activity can persist alongside hyperscale demand.

Another quarter above $100 million in bookings from zero-to-one-megawatt deployments and interconnection would strengthen the inference story. It would suggest that AI-related infrastructure demand is spreading beyond isolated training campuses.

A return to heavily concentrated hyperscale leasing would not necessarily represent weakness. However, it would make the diversification argument less certain and increase attention on large-customer exposure.

Equinix’s results belong in the same monitoring framework. Continued record interconnection activity from both companies would support the view that enterprise AI adoption requires connected, distributed infrastructure.

The comparison will also reveal whether Digital Realty is gaining distinctive traction or participating in a broad market upswing. Strong numbers from every operator can confirm industry demand without identifying a clear winner.

Investors should treat the raised guidance as a near-term operating benchmark. Meeting the revised revenue and Core FFO ranges would support management’s execution narrative. Another increase would add evidence that renewal pricing and lease commencements are exceeding prior assumptions.

Missing the range would require a closer look at timing. The cause could be delayed capacity, slower customer deployment, financing pressure, or an operating issue unrelated to AI demand.

For enterprise technology teams, the practical question is availability. A large industry backlog means buyers should evaluate where and when capacity can actually be delivered, not merely which provider announces the largest global pipeline.

Teams can use a structured knowledge workflow to compare provider filings, utility milestones, capacity schedules, and changing deployment assumptions over time.

Digital Realty has already cleared the demand test. Customers are signing leases across hyperscale, colocation, and interconnection categories, and management has raised its 2026 outlook twice.

The harder test begins after the signature. Can Digital Realty energize capacity, preserve projected returns, and turn its record backlog into recurring rent on schedule?

That conversion will determine whether the current Digital Realty AI demand story represents durable infrastructure growth or a temporary gap between contracts and completed capacity. Watch the next backlog roll-forward, construction update, and leasing mix. Together, those three signals will provide a clearer answer than another record headline.