Finvest
DLR Data Center REITs · AI infrastructure · REIT · Global data centers · Thesis updated June 12, 2026

AI demand is real, execution is the test

01 Running thesis

AI bookings changed the story

Digital Realty is one of the clearest public ways to invest in the data center buildout. In Q1 2026, the company signed its second-highest bookings quarter ever. The standout was a 200 megawatt AI inference lease, the largest lease in its history. Inference means running an AI model for users after it has been trained, often closer to where people and businesses are.

The bull case is stronger because demand is not only coming from giant cloud customers. The smaller 0 to 1 megawatt plus interconnection category also set a record, with 21% of bookings tied to AI. That suggests more businesses are starting to use AI inside normal IT systems, not just in huge training clusters.

Digital Realty is turning that demand into new supply. Its development pipeline grew by more than 50% in one quarter to 1.2 gigawatts under construction. It was 61% pre-leased at an 11.4% average expected yield, and management raised 2026 Core FFO guidance to $8.00 to $8.10 per share.

The bear case is now simpler and sharper. The company has a $16.5 billion pipeline, and building that much data center capacity is hard. Power, labor, cooling gear, and construction timing all matter. The stock also does not look cheap, so strong leasing may not be enough if returns slip or projects run late.

May 2026The Q1 2026 Form 10-Q did not change the thesis. Management still said demand was strong in most key metro areas, and the company reported no change to its risk factors.
Apr 2026Q1 2026 results strengthened the story. Digital Realty signed its second-highest bookings quarter, landed a 200 megawatt AI inference lease, expanded its pipeline to 1.2 gigawatts, and raised 2026 Core FFO guidance.
Feb 2026The 2025 Form 10-K showed continued capital recycling and more than 300 data centers in more than 55 metro areas. It also made the execution risk clearer because 2026 capital spending plans stayed large.
Oct 2025The Q3 2025 Form 10-Q confirmed stronger revenue and progress on the U.S. hyperscale fund. That helped support the funding side of the growth plan.
Oct 2025Q3 2025 earnings showed record interconnection bookings and another guidance raise. Management also said AI had averaged more than 50% of quarterly bookings since mid-2023.
Aug 2025The Q2 2025 Form 10-Q confirmed more than $3 billion of equity commitments for the U.S. hyperscale data center fund. That reduced funding risk for new development.
Jul 2025Q2 2025 earnings showed record 0 to 1 megawatt plus interconnection bookings and higher Core FFO guidance. The quarter supported the idea that growth was broadening beyond the largest cloud deals.
May 2025The Q1 2025 Form 10-Q showed more joint venture activity, including Indonesia, and rising commitments to the U.S. hyperscale fund. This improved confidence in the capital plan.
02 Business model

Renting power, space, and connections

Digital Realty is a real estate investment trust, which means it owns real estate and pays out much of its taxable income to shareholders. Its real estate is data centers. Customers rent space, power, cooling, and network connections so their servers can run safely.

The largest customers are cloud service providers and other hyperscalers. A hyperscaler is a company that needs huge computing sites. Digital Realty also serves enterprises, which are regular companies that need smaller deployments tied to cloud, data, and network partners.

The moat comes from scale. Digital Realty says its global portfolio spans more than 300 data centers in more than 55 metro areas across 6 continents. The PlatformDIGITAL network gives customers places to connect to clouds, carriers, and partners. That matters more if AI inference workloads need to sit near users.

Funding is a key part of the model. Digital Realty uses joint ventures, asset sales, and funds with partners such as Blackstone and GI Partners to help pay for new data centers. That can protect the balance sheet, but it also means growth depends on outside capital staying available at fair terms.

03 Product portfolio

What customers actually rent

Growth engine

Hyperscale data centers

Large cloud and AI customers lease big blocks of power and space. This is where the 200 megawatt AI inference deal fits.

Growth engine

0 to 1 megawatt deployments

These smaller deals serve enterprises and mixed cloud users. This category set another quarterly record in Q1 2026.

Cash cow

Interconnection

Customers pay to connect their systems to carriers, clouds, and business partners inside Digital Realty sites. These links can make a data center stickier.

Option

High-density colocation

Digital Realty is adding liquid-cooled space for GPU servers used in AI. This gives the company a way to serve denser workloads than older data halls were built for.

Option

Development land and power bank

The company holds land and utility agreements that can become future capacity. This is valuable only if power arrives on time and construction stays within budget.

Steady

Capital partner vehicles

Joint ventures and funds let Digital Realty keep partial ownership while sharing the cost of large projects. This supports growth, but it adds partner and deal timing risk.

04 Business segments

Where the rent comes from

Northern Virginia21%growing fast
Chicago7%modest
Frankfurt6%modest
London4%modest
Singapore4%modest
Other metros56%modest

This mix uses annualized rent by top metro area as of December 31, 2025, from the company context tied to the 2025 Form 10-K. Digital Realty reports more detail by geography than by cloud versus enterprise customer type.

05 Risk factors

What could break the thesis

Power does not arrive

High impact · Medium odds

Data centers cannot lease what they cannot power. AI servers use more electricity than older workloads, which raises the burden on utilities and grid connections. If Digital Realty cannot secure enough power at the right cost, its land bank will be less valuable.

We watchWatch management comments on utility agreements, energization dates, and power availability in Northern Virginia, Atlanta, and other large campuses.

The pipeline runs late or over budget

High impact · Medium odds

Digital Realty's pipeline has scaled to $16.5 billion and 1.2 gigawatts under construction. That creates a large execution test. Delays, labor shortages, equipment shortages, or cost overruns could reduce the expected 11.4% yield and slow Core FFO growth.

We watchWatch the pre-leased share, expected yield, development cost updates, and any delayed delivery dates in quarterly reports.

Capital partners pull back

High impact · Medium odds

The company relies on joint ventures, funds, and asset sales to help fund growth while managing leverage. This has worked well, including more than $3 billion of equity commitments for its U.S. hyperscale fund. But if capital markets tighten, Digital Realty may have to slow development, issue equity, or accept weaker terms.

We watchWatch new fund closings, joint venture pricing, asset sale proceeds, and management's leverage targets.

AI demand shifts to lower-return sites

Medium impact · Medium odds

AI training can often run in very large campuses where power is cheap, while AI inference may need lower latency near users. Digital Realty's bull case is stronger if inference grows inside its core metro markets. If most demand stays in remote, lower-priced training campuses, returns may be less attractive.

We watchWatch the mix of AI bookings between inference and training, plus AI demand in the 0 to 1 megawatt plus interconnection category.

Price leaves no safety margin

Medium impact · High odds

The business is improving, but the market already knows data centers are tied to AI. Finn's valuation view is weak, which means the stock may need both strong leasing and clean execution to work well. A good company can still be a hard stock if expectations are too high.

We watchWatch Core FFO guidance, re-leasing spreads, development yields, and any reset in investor expectations for data center REITs.
06 Quick answers

In one breath

Is Digital Realty an AI stock?

Digital Realty is not an AI software company. It owns the data centers that AI systems need for power, cooling, and network access. Q1 2026 showed strong AI demand, including a 200 megawatt inference lease.

How does Digital Realty make money?

It leases data center space, power, cooling, and connections to cloud providers, AI users, and enterprises. It also uses joint ventures and funds to help pay for new developments while keeping some ownership.

What is Core FFO?

Core FFO is a real estate cash flow measure that adjusts earnings for items like depreciation and certain one-time effects. Management guided to $8.00 to $8.10 per share for 2026 after Q1 results.

What is the biggest risk for Digital Realty?

The biggest risk is execution. Demand is strong, but Digital Realty must deliver 1.2 gigawatts of new capacity and a $16.5 billion pipeline while power, labor, and equipment are tight.