Digital Realty Trust

Phase A — Understand the business

Lens 1 · Company Overview

Digital Realty is the world's second-largest data-center REIT — a global landlord of digital infrastructure that leases space, power, cooling, and interconnection to enterprises, cloud providers, and carriers. As of 2025-12-31 the portfolio was 310 data centers (118 US, 113 Europe, 36 LatAm, 18 Asia, 16 Africa, 6 Australia, 3 Canada), totalling ~57.6M rentable sq ft (incl. 9.7M sq ft under active development and 4.7M held for development), with ~2.9 GW of in-place IT capacity and >5 GW of future development capacity ``. Reincorporated/HQ'd in Dallas, TX (the Q1 10-Q lists an Austin address — a relocation in progress); REIT since IPO on 2004-10-29.

How it makes money. It is a hybrid wholesale + colocation operator. Revenue splits two ways by deployment size: the >1 MW (hyperscale wholesale) cohort is 59.7% of annualized rent, and the 0–1 MW (colocation + interconnection) cohort is 34.9% — the latter is the higher-margin, stickier, interconnection-rich business that distinguishes DLR from a pure wholesale developer . Its product spine is **PlatformDIGITAL®** — a "data gravity" thesis: as enterprise data mass grows, it becomes costly to relocate, so colocating compute next to it (in 55+ metros, with **232,000+ cross-connects** and 305+ cloud on-ramps via ServiceFabric®) is the value proposition .

Contract structure. Operating leases, avg remaining lease term ~5 years, generally with contractual escalators plus expense reimbursement (utilities, taxes, CAM) — so power-cost inflation is substantially passed through to tenants ``. This is a recurring, contracted, escalating-rent model — the closest thing to a toll road on AI/cloud compute.

Customers / suppliers / competitors. Customers = the hyperscalers and Global 2000 (Lens 1 detail in Lens 2/9). Suppliers = power utilities, construction/GC, and chip-adjacent cooling vendors. Competitors = Equinix (the #1 by market cap, interconnection-dense retail colo), CoreWeave/CoreSite (American Tower-owned), Vantage, QTS (Blackstone), NTT, CyrusOne (KKR/GIP), plus hyperscaler self-build.

Lens 2 · Supply Chain

Upstream → DLR → end customer, named at each link:

• Power (the true bottleneck). Regional utilities — Dominion (Northern Virginia, DLR's #1 market at 21.4% of rent), plus European/African/APAC grid operators. Power, not land or steel, is the binding constraint; DLR's land bank "could accommodate over 3,500 MW of additional capacity, including >1,000 MW developable in Northern Virginia" — but only as power is delivered ``.
• Construction & capital partners. DLR increasingly does NOT fund builds alone — it syndicates. Named development-capital partners: Blackstone ($7B hyperscale JV, Frankfurt/Paris/NoVA, ~500 MW, 80/20 split ); **Mitsubishi** (added 15% to a JV for $62M in Jan 2025); **TPG Real Estate** ($1.5B NoVA portfolio, 80/20); the **DC Partners NA Fund** (>$3B equity commitments raised in 1H25, DLR retains 20–60% + <2% GP interest) .
• DLR (the platform). Owns/operates the metal; provides Critical Facilities Management®, interconnection fabric, 99.999% uptime (18 consecutive years ``).
• Downstream buyers (the chokepoint risk). Concentrated in a handful of hyperscalers — see Lens 9 concentration. Single-source dependency cuts both ways: 20 of 310 data centers are single-tenant, including assets occupied solely by the top-3 customers ``.

Chokepoint verdict: the supply chain is gated at power availability (favors incumbents with land+power positions) and exposed at customer concentration (a hyperscaler demand-pull, e.g. Microsoft's 2025 lease pullback, ripples straight through).

Lens 3 · Competitive Advantages (moats)

1. Irreplaceable interconnection density. The 232k cross-connects and connected "data communities" in gateway facilities (e.g. the Westin Seattle building, Ashburn) are "difficult for competitors to replicate" ``. This is a genuine network-effect moat on the 0–1 MW colo business — a tenant in a dense exchange will not move for a 10% rent saving.
2. Scale + global footprint as a switching-cost moat. Multinationals get one contract across 55 metros / 30 countries / 6 continents instead of negotiating with dozens of local providers ``. Tenant improvements installed at customer expense raise switching costs further.
3. Power & land bank. 3,500+ MW of developable capacity, >1,000 MW in NoVA, in a world where new power interconnects take years — this is the scarce input, and DLR is long it.
4. Capital-access moat. Investment-grade balance sheet + a stable of institutional capital partners (Blackstone, TPG, Mitsubishi, the Fund) lets DLR fund hyperscale builds without bearing all the equity — see Lens 9.

Bargaining power is asymmetric and worth being honest about. Against a Microsoft or Oracle taking 76 / 42 locations, DLR has limited pricing power (these are sophisticated counterparties with self-build alternatives). Against the long tail of 0–1 MW colo/enterprise tenants locked into dense exchanges, DLR has strong pricing power — evidenced by the re-leasing spreads in Lens 5. The moat is real but concentrated in the colo book, not the hyperscale book.

Lens 4 · Segments

One reporting segment. DLR explicitly manages as "a single global business — with one operating segment and therefore one reporting segment"; the CODM (CEO) uses consolidated net income ``. So there is no product-segment P&L to break out — the meaningful cut is geography and deployment size.

By geography (% of annualized rent, 2025-12-31) ``:

US ≈ ~55–60% of rent; the rest is a genuinely global, FX-exposed (EUR/GBP/SGD/JPY/ZAR/BRL) book — a differentiator vs US-centric peers and a source of reported-FFO volatility.

By deployment size: >1 MW = 59.7% of rent (hyperscale, accelerating on AI), 0–1 MW = 34.9% (colo/interconnection, the margin + moat). The trend is hyperscale-led acceleration — record >1 MW bookings drove the 2026 backlog to an all-time high (Lens 5).

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print: Q1 2026, reported 2026-04-23)

The defining quarter of the current thesis — a beat-and-raise driven by record AI-led bookings.

• Revenue $1,635.2M, +16.2% YoY (from $1,407.6M); rental & other services $1,600.2M, +15.4% ``. Stabilized rental +$113.3M (new leasing, renewals, utilities reimbursement, FX); non-stabilized +$100.1M (development deliveries).
• GAAP net income to common $169.1M; diluted EPS $0.46 (vs $0.27) ``.
• FFO/share (Nareit) $1.99 diluted, +19% (from $1.67); FFO available $700.4M . **Core FFO/share $2.04** vs $1.77 YoY .
• Guidance RAISED: 2026 Core FFO to $8.00–$8.10/share (+9% at midpoint over 2025), constant-currency same-cap cash NOI +4–5% ``.
• Bookings = the headline. Record combined 0–1 MW + interconnection bookings plus stronger >1 MW leasing → $423M of annualized GAAP rent at DLR's share, pushing signed backlog to ~$1.8B at 100% share — an all-time high . (Trajectory: backlog was $826M at 2Q25 → $1.8B by Q1 2026. That is the single most important momentum fact in the file.)
• Balance sheet flags (positive). Net Debt/Adj EBITDA 4.7x — a multiyear low . Total debt $18.14B at **2.85% weighted-avg rate**, ~89% fixed/swapped; only $1.2B floating . Cash $3.45B (YE25). Capex Q1 $910M; FY guide raised +$250M to $3.5–4.0B ``.
• Market reaction / what's priced in. Stock +~27% YTD into mid-2026 ``; trading ~$190 vs 52-wk $146–$208. The raise was expected — DLR has beaten-and-raised every quarter through this cycle, so the bar is high.

Unusual vs its own history: a $134.3M Redeemable-NCI adjustment (Teraco put right, in-the-money from 2026-02-01) ran through Additional Paid-In Capital, not the income statement — a non-cash equity reshuffle worth flagging (Lens 10).

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on disk (transcripts=0); sentiment reconstructed `` from Q4'25 (2026-02-05), Q1'26 (2026-04-23) calls and conference appearances.

• Tone: confidently bullish, increasingly AI-centric. Management has shifted the narrative from "data gravity / hybrid cloud" (the multi-year PlatformDIGITAL story) to "AI infrastructure demand" as the primary booking driver — the 0–1 MW + interconnection record alongside >1 MW strength is the phrase they keep returning to ``.
• What they started saying: "record backlog," "double-digit development yields," "balance sheet at a multiyear low leverage," "private capital partners."
• What they stopped emphasizing: churn/occupancy-drag worries and the 2023-era "funding gap / dilution" defensiveness — the Fund + JV model has answered that critique on the call.
• Oversupply, addressed head-on. CEO Andy Power has publicly engaged the oversupply question (RECM 2026 conference, DCD coverage) — arguing power scarcity and pre-leasing (64% of the 769 MW underway is pre-leased) insulate DLR ``. That he is asked this repeatedly is itself the sentiment tell.

Lens 7 · Comps

Data-center + infrastructure REIT peer set. Multiples are `` with source/date or n/a. Do not treat the n/a cells as zero.

Read: DLR trades at ~108.8% of NAV (a premium) vs EQIX at ~86% of NAV (a discount) ``. On forward P/FFO, DLR (~23–24x) is cheaper than EQIX (~30x), but EQIX is the higher-quality interconnection franchise and is the one screening as relatively cheap to its own history. The honest comp conclusion: DLR is fairly-to-fully valued in absolute terms and only "cheap" relative to a more expensive EQIX. I will not fabricate the missing multiples — they were not sourceable to a dated outlet in this pass.

Lens 8 · Stock-Price Catalysts (what actually moves DLR)

Pattern over the cycle ``:

• AI-demand inflections (up). The 2023→2024 generative-AI capex wave re-rated the whole group; DLR's record-bookings prints are the clearest up-catalysts. 3-yr TSR ~98% ``.
• Hyperscaler capex signals (up AND down). The most dangerous single factor: Microsoft's early-2025 lease cancellations (~200 MW) / 1.5 GW self-build freeze (TD Cowen) hit the entire sector's sentiment ``. DLR moves on what Microsoft/Amazon/Google/Meta/Oracle say about capex more than on its own results.
• Rates (down when rising). As a levered, long-duration REIT, DLR de-rated in the 2022–23 hiking cycle; the de-leveraging to 4.7x is partly a response.
• Capital-recycling / JV announcements (up). Blackstone $7B JV, the Fund close — each validated the "asset-light hyperscale" pivot and lifted the stock. Takeaway: the market prices DLR as a derivative of hyperscaler AI capex, secondarily on rates. Its own execution is necessary but not sufficient.

---

Phase C — Judge people & books

Lens 9 · Management

• Andrew (Andy) Power — CEO since Dec 2022 (~3.5 yrs), prior President (2021) and CFO (2015–2022) . A **finance-pedigree CEO** — career capital-markets operator who "managed >$30B of capital raises and >$19B of M&A" and was on DLR's 2004 IPO underwriting team . Archetype: professional manager / capital allocator, not founder-operator. For a capital-intensive REIT mid-AI-buildout, that is arguably the right archetype — the job is funding and allocation, which is his native skill.
• Greg Wright — CIO, leads investment/M&A/capital allocation ``. The JV-and-Fund flywheel is his fingerprint.
• Track record (quantified). Under Power: FY2025 revenue $6.11B (+10%), Core FFO/sh ~$7.10, leverage cut to 4.7x from the ~6.5x+ of the Bill Stein era, $1.04B total bookings in 2024, 35.8% TSR in 2024 ``. The signature achievement is fixing the balance sheet without halting growth — exactly the 2022 bear thesis (over-levered, serial equity issuer) now substantially neutralized.
• Skin in the game / comp. Insider ownership is modest (typical large-REIT professional-management profile; insider-transactions.csv absent — not quantified here, n/a). Equity comp is sizeable: LTIP/PSU/RSU grant-date fair values $164–177/sh in 2025; unearned comp ~$140M ``.
• Capital-allocation history — the core of the bull case. The "private-capital flywheel": develop at an 11.9% stabilized yield-on-cost, then sell down stabilized/partial interests to Blackstone/TPG/Mitsubishi/the Fund at premium cap rates, recycling proceeds into more development. In 2025 this generated an $873M gain on the Fund contribution alone, plus $271M (Brookfield), $124M (non-core sales), $77M (Dulles to Blackstone), $427M (Q4 Fund top-up) ``. It de-risks the build-out and funds the dividend without max equity issuance.
• Red flags (fair). Serial equity dilution remains the legitimate knock — $1.1B ATM in 2025 at $173.09 avg (down from $2.0B ATM + $1.7B offering in 2024) ``. Share count has crept from ~299M (2023) to ~345M (Q1'26). The Teraco put (Lens 10) is a related-party-adjacent overhang. Dividend held flat at $4.88 since 2024 — not cut, but not growing either, a tell that FFO growth is being funneled to the build, not the payout.

Lens 10 · Forensic Red Flags

Forensic-analyst lens. Every figure labeled.

• Capitalized interest & overhead flatter FFO. DLR capitalized $127.2M of interest and $140.6M of compensation/overhead into construction in 2025 (vs $118.9M / $111.2M in 2024) ``. Legitimate for a developer, but it means reported interest expense and G&A understate the true run-rate; rising capitalized amounts track the build, and would reverse into the P&L if development slowed. Watch this ratio.
• Gains-on-disposition are a large, lumpy earnings component. FY2025 net income to common was $1,267.9M `` — but ~$873M+ of that is the Fund-contribution gain plus other disposition gains. GAAP EPS is gain-inflated; FFO (which excludes RE gains) is the right lens — and FFO confirms the underlying strength, so this is a quality-of-GAAP-earnings flag, not a quality-of-business flag. Watch that the flywheel's gains don't get treated as recurring.
• Redeemable NCI / Teraco put. The Rollover Shareholders can put their Teraco stake to DLR (cash or shares) over a 2-yr window from 2026-02-01. In Q1'26 DLR booked a $134.3M adjustment to Redeemable NCI through APIC as the contractual redemption value exceeded carrying value ``. A real contingent claim on equity/cash; non-cash today but a future dilution/cash drain to size.
• Impairments are recurring (non-core tail). $78.6M (2025) and $191.2M (2024) impairments on "non-core properties in secondary U.S. markets" ``. Modest vs a ~$49B asset base, but a reminder that not all of the portfolio is the AI-prime Ashburn/FLAP-D book.
• FX / equity-method opacity. Ascenty (LatAm) and Teraco debt FX remeasurement swings equity-in-earnings (−$88.2M YoY in 2025) — earnings volatility from unconsolidated JVs that DLR "does not control or manage," so disclosure controls there are "substantially more limited".
• Controls / audit: disclosure controls concluded effective; no changes to ICFR ``.

Regulatory findings (required sub-section).

• SEC Litigation Releases / AAERs: None. Searched EDGAR EFTS (LR + AAER) for "Digital Realty Trust" 2021-06-21→2026-06-21, total_sec_findings: 0 ``.
• 10-K Item 3 (Legal Proceedings): "As of December 31, 2025, we were not a party to any legal proceedings which we believe would have a material adverse effect" . Q1 10-Q reaffirms the same as of 2026-03-31 .
• Non-SEC enforcement (web): No material FTC/DOJ/EU/utility-regulator enforcement actions against DLR surfaced in this pass ``. (Sector-level regulatory pressure — power-grid permitting, data-sovereignty, EU energy rules — is a business risk, not an enforcement finding.)
• Conclusion: No material regulatory or legal findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K/10-Q Item 3 as of 2026-06-21.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

Built bottom-up from FY2025 Core FFO/sh ~$7.10 and the company's own raised 2026 guide. Output ``; inputs labeled. No forecast.ts create (watchlist loop).

Anchors: FY2025 Core FFO/sh ~$7.10 / Nareit FFO **$6.94**. 2026 Core FFO guide $8.00–$8.10 . Drivers: same-cap cash NOI +4–5% CC ; +50–100 bps occupancy off ~89% ; cash renewal spreads +6–8% ; $10B development at 11.9% yield delivering into the backlog ``; offset by ATM dilution (~2–3%/yr) and higher financing as low-rate debt rolls.

The base case is a ~9–10%/yr Core FFO/sh compounder — premium to the ~5–6% the group historically delivered, justified by AI-led demand and the development pipeline, if hyperscaler capex holds. Brier forecast (logged on resolution, not here): "DLR FY2026 Core FFO/sh ≥ $8.00, p≈0.80."

Lens 12 · Bull vs Bear

Bull case. DLR is the AI build-out's picks-and-shovels landlord with (1) a scarce power+land bank (3,500+ MW, >1,000 MW in NoVA) in the one input that is genuinely supply-constrained; (2) record demand — $1.8B backlog, 769 MW underway at 64% pre-leased, 11.9% dev yield; (3) a fixed balance sheet (4.7x, 2.85% blended rate, 89% fixed) that removes the 2022 bear thesis; (4) a private-capital flywheel (Blackstone/TPG/Fund) that funds growth at a premium-to-cost and books real gains. At ~24x forward FFO for a 9–10% grower with a fortress balance sheet, the quality is not expensive.

Bear case (permanent-impairment candidates). (1) AI-capex air pocket — hyperscaler capex ~$600B in 2026 (+36%) is the demand; Microsoft already cancelled ~200 MW and froze 1.5 GW of self-build. If 2026–27 arrives without AI monetization, the demand DLR is building into evaporates and pre-leasing rates fall. (2) Oversupply + power-shifting — peers (EQIX, CoreSite/AMT, Vantage, QTS) are all building; in power-rich secondary metros, supply could outrun demand and compress spreads (note DLR's own guide caveats renewal upside on fixed-rate >1 MW expiries). (3) Negative FCF through FY27 on $3.5–4.0B capex — growth is funded by dispositions + ATM dilution; if capital markets tighten or JV appetite cools, the flywheel stalls and dilution accelerates. Pre-mortem (18 months out, thesis broke): a hyperscaler capex reset (à la a deeper Microsoft pullback) cut new bookings, the backlog stopped growing, dev yields compressed below 10%, the stock de-rated from ~24x to ~18x FFO — a ~25% drawdown despite "fine" reported FFO, because the growth premium came out.

Are multiples too high? Not egregiously — ~24x forward FFO for ~9–10% growth + fortress balance sheet is defensible, but it prices in continued AI-demand strength. There is little margin of safety at ~108% of NAV. Contrarian view: the market treats DLR as a binary AI-capex bet; the underappreciated fact is the balance-sheet transformation and the 0–1 MW interconnection moat, which would let DLR compound respectably even in a "cloud-not-AGI" world — the downside is less binary than the tape implies.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull case.

• Revenue is dangerously concentrated in counterparties with a self-build option. Top-20 = 50.9% of annualized recurring revenue; top-3 ≈ 26%; #1 (Fortune-50 software, ~Microsoft) alone 11.7% ($547M), #2 Oracle 9.0% ($424M) ``. These are the exact firms (a) building their own data centers and (b) already cancelling leases. A renewal-pricing standoff or non-renewal by even one is a multi-point FFO event — and 20 of 310 centers are single-tenant.
• The moat is weaker on the part that's growing. The 0–1 MW interconnection moat is genuine — but the growth is >1 MW hyperscale wholesale, which is commoditized, lowest-margin, and where DLR has the least pricing power. The mix is shifting away from the moat.
• Most dangerous competitor bulls underrate: hyperscaler self-build (Microsoft's 1.5 GW freeze was a self-build freeze — when it un-freezes, that capacity competes with DLR's wholesale), and CoreSite under American Tower plus Blackstone-owned QTS (note Blackstone is simultaneously DLR's JV partner and, via QTS, its competitor — a structural tension).
• Capital allocation: the flywheel is also financial engineering. $873M of FY2025 net income is a disposition gain, not operating earnings; capitalized interest + overhead ($268M combined) flatter FFO; the dividend has been frozen at $4.88 since 2024; share count is up ~15% in two years. A skeptic reads "serial dilution dressed up as a fund strategy."
• What must hold for ~$190: AI-led hyperscaler capex stays at/above current run-rate; dev yields stay ~12%; spreads stay +6–8%; the JV/Fund bid for stabilized assets persists. If FY2027 growth disappoints by 20–30% (Core FFO/sh ~$8.20 not ~$8.85), the multiple compresses to ~18x and the stock is a high-$140s/low-$150s name — back to the bottom of its 52-wk range.
• Single scenario that permanently impairs: a structural finding that AI training/inference does NOT need as much new third-party colo as projected (efficiency gains + hyperscaler self-build absorbing the load) — turning today's land bank into long-dated, lower-yield optionality. Plausibility: moderate — not the base case, but the Microsoft signal proves it is not tail-risk either.

Lens 14 · Management Questions (ordered by information value)

1. Of the $1.8B signed backlog, what share is with your top-5 customers, and what is the weighted-average lease term and pre-leasing % on the 769 MW under construction — i.e. how much is de-risked vs speculative?
2. Microsoft (your #1, ~11.7%) cancelled ~200 MW and froze 1.5 GW of self-build industry-wide in 2025 — what did your renewal economics and expansion bookings with your top-3 hyperscalers actually look like in the last two quarters?
3. The 11.9% stabilized dev yield — what is it on new 2026 starts specifically (post power-cost and construction inflation), and where does it trend as you build into more power-constrained metros?
4. You held the dividend flat at $4.88 since 2024 while AFFO payout fell to 64% — is the policy to grow the payout, or to retain for the build? What would change it?
5. FCF is negative through at least FY27 on $3.5–4.0B capex — under what capital-markets scenario does the JV/Fund + ATM funding model break, and what is plan B?
6. Quantify the Teraco put: expected cash vs share settlement, timing, and the dilution/cash range if Rollover Shareholders fully exercise from 2026.
7. On the >1 MW vs 0–1 MW mix shift toward lower-margin hyperscale — how do you defend blended margins as the growth tilts away from interconnection?
8. Blackstone is both your largest JV partner and, via QTS, a competitor — how do you manage that conflict in deal terms and market overlap?
9. Power is the binding constraint — how many MW of your >5 GW future pipeline have contracted power interconnects vs are awaiting utility delivery, by metro?
10. Same-cap cash NOI +4–5% — what is the organic (ex-FX, ex-development) component, and what is the path to accelerate it?
11. What is the realistic re-leasing spread on the 2026–27 >1 MW expirations given the fixed-rate mix you flagged?
12. How much of reported FFO growth is "real" vs capitalized interest/overhead that would reverse if development pace slowed?
13. Cooling/density for AI racks (liquid cooling, >50 kW/rack) — what retrofit capex is required across the legacy portfolio, and is it in the capex guide?
14. M&A appetite at this point in the cycle — buyer or seller of platforms, and at what cap rates?
15. What is the single demand signal you watch that would make you slow development starts?

---