Digital Realty Trust, Inc. (NYSE: DLR) — The Right Address for the AI Build-Out, at a Top-of-Cycle Rent

Independent fundamental research. Report date: 2026-06-20. The body (Sections 1–15) carries no recommendation and no price target; the sole exception is the clearly-labeled Claude’s Take block below.

⚡ Claude’s Take

This block is the author’s own independent opinion and general information, not investment advice. The analysis that follows (Sections 1–15) takes no position and names no price target.

Verdict: HOLD / fair-to-full price; accumulate on weakness; not a short. Digital Realty is a genuinely improved business — deleveraged from ~7x to 4.7x net-debt/EBITDA, with cash re-leasing spreads positive for the first time in years, a capital-light private-capital fee platform scaling fast, and Core FFO/share finally inflecting (+10% in 2025, ~+9% guided for 2026) after a lost half-decade of flat per-share earnings. None of that is in dispute. The problem is what you pay for it. At $188.15 DLR trades at roughly 23–24x forward Core FFO, ~25x P/AFFO, ~28x EV/EBITDA, the 96th percentile of its own 10-year price/book and price/sales history, and an implied cap rate (~4.5–5.5%) below where private data-center assets change hands. You are buying the #2 player — on every quality axis that matters — at a multiple only ~6–10% below the best-in-class #1 (Equinix). That discount is too thin for the gap behind it. My fair-value accumulation zone is ~$150–170 (~19–21x forward Core FFO), with a genuine table-pounder only on a rate-or-AI-scare drawdown into the low-$130s (the level DLR actually traded at in April 2025, ~13 months ago).

The framing is “real cyclical inflection, priced as a permanent structural re-rating.” The bull case — power scarcity hands wholesale landlords durable pricing power, DLR’s ~5 GW land-and-power bank is irreplaceable, the fee platform compounds — is all true today. But the load-bearing number is the development yield: DLR builds at ~11.4% stabilized, less than half of Equinix’s ~26–27% retail cash-on-cash, and that ~11.4% is a commodity return the capital cycle is built to mean-revert. Record private capital (~$45.7B into data centers in 2025, plus DLR’s own $3.25B fund) is flooding the exact segment DLR leads, and the only thing holding returns up is a power bottleneck that the grid will, eventually, relieve. Pay 96th-percentile prices for that and your margin of safety is the thickness of the power-interconnection queue. Conviction: medium. What flips me bullish: a ~15–20% de-rate, or a third consecutive year of ~8–10% per-share Core FFO growth with the diluted share count growing under ~3%/yr (proof the inflection is structural, not cyclical). What flips me bearish: development yields on new starts printing below ~9–10%, or re-leasing spreads fading back toward flat as supply catches demand. Tag: you can own the landlord of the AI build-out — just not at the top-of-cycle rent the tape is charging.

📈 Stock Price Action — Five-Year Event Map

Factual price history, not a recommendation. Price moves are FACT; the attributed drivers are INTERPRETATION. No price target, no chart-pattern claims.

DLR has round-tripped and then some over five years. Split/dividend-adjusted, it peaked near ~$150 in late 2021, crashed to a $77.88 trough on 2023-05-24 (a ~48% drawdown), then re-rated all the way to a $202.56 all-time high on 2026-04-20 before easing to $188.15 (2026-06-18) — about 7% off the high, with a 52-week range of $145.96–$202.56. The arc is the story of the whole sector: a long-duration REIT killed by the 2022–23 rate shock and a “wholesale will commoditize” bear thesis, then resurrected by the AI/data-center demand wave and a balance-sheet repair.

The throughline: DLR is a rate-sensitive, long-duration REIT (the single largest factor in its return is a −0.43 loading to interest rates) that was repriced down by rates and up by the AI demand story. The 2025–26 leg is fundamental (real leasing, real spread inflection) layered on a recovered rate backdrop — which is why the stock sits near highs at a multi-year-low dividend yield.

1. Executive Summary

Digital Realty is the world’s second-largest data-center REIT and the scale leader in wholesale/hyperscale capacity — ~310 data centers, ~42.5 million square feet, across 50+ metros and 25+ countries, generating $6.11B of FY2025 revenue (+10%), ~$62B equity value and ~$82B enterprise value. It is a UPREIT (the parent owns ~98.2% of the operating partnership, Digital Realty Trust, L.P.). Structurally it straddles two very different sub-industries: a retail colocation + interconnection book (the “0–1 MW + interconnection” segment) that is genuinely moaty but where DLR is a clear #2 to Equinix, and a much larger wholesale/hyperscale book (the “>1 MW” segment) that is structurally more commoditized but where DLR leads. The revenue tilt toward wholesale is the single fact that explains almost everything about DLR — its lower multiple, its lower margins, its customer concentration, and its lower returns on invested capital than Equinix.

The central tension is a real cyclical inflection priced as a permanent structural re-rating. The good news is real: after a half-decade of flat per-share earnings — Core FFO/share was roughly $6.61 (2021), $6.70 (2022), $6.55 (2023), $6.72 (2024) — DLR finally grew it +10% to $7.39 in 2025 and guides $8.00–$8.10 (~+9%) for 2026. The inflection is driven by genuine forces: cash re-leasing spreads turned +6.7% in 2025 (after years flat-to-negative), primary-market vacancy is at a record-low ~1.4% (CBRE), the new private-capital fee stream doubled to $143.8M, and management deleveraged hard from ~5.8x (2023) to 4.7x net-debt/Adjusted EBITDA (Q1-2026, a multi-year low). The development pipeline scaled +60% to $16.5B / ~6 GW (1.2 GW under construction, 61% pre-leased) and the backlog hit a record $1.8B.

The caution is equally real. First, the per-share inflection is only two years old and coincides with a once-in-a-cycle AI/power-scarcity spike — it is not yet proof that the structural drag (serial ~6.5%/yr dilution + commodity wholesale economics) is permanently fixed. Second, the economics are lower-quality than they look: development yields of ~11.4% are less than half Equinix’s retail cash-on-cash, GAAP returns are a depreciation artifact, AFFO is a flattered distributable-cash proxy, and true post-development free cash flow is deeply negative by design. Third, the capital cycle is flashing late-stage: ~$45.7B of private equity flooded into data centers in 2025, the highest in five years, into the very segment DLR leads. Fourth, customer concentration is real — the largest customer is ~11.7% of annualized recurring revenue and the top two ~21% (versus ~3% for Equinix’s largest). Fifth, the price: DLR trades at the 96th percentile of its own 10-year price/book and price/sales history, at a sub-private-market implied cap rate, and at only a ~6–10% Core-FFO-multiple discount to a materially higher-quality Equinix.

Capital allocation has genuinely improved — the private-capital pivot, the disciplined dividend freeze (held at $4.88 since 2022, letting the payout fall to ~64% of AFFO), and a Core-FFO/share-anchored compensation plan are all positives that mark a turn from the prior era’s value-destructive growth-via-dilution. But the historical record is poor (flat per-share FFO from dilutive M&A), insiders own under 1% with zero open-market purchases, and the entire bull case rests on a cyclical tailwind holding through a capital cycle designed to erode it. The business is better than it was; the price assumes it is better than it is.

2. Business Overview

What Digital Realty sells. DLR is a data-center landlord and connectivity provider operating on two speeds. The first product is turn-key/wholesale data-center capacity — large contiguous blocks of power, space and cooling (the “>a megawatt” category), leased on long-dated contracts to hyperscalers, cloud providers and, increasingly, AI workloads. The second is colocation plus interconnection — smaller deployments (the “0–1 MW + interconnection” category) sold to enterprises and network/IT-services customers, bundled with connectivity products: physical Cross Connects, Metro Connect, the software-orchestrated ServiceFabric global fabric, Internet services and pathway/pre-cabling. These sit inside PlatformDIGITAL, DLR’s global platform branded around the “Pervasive Datacenter Architecture (PDx)” methodology for managing “data gravity.” A third, newer and capital-light revenue line is fee income from managing the private-capital vehicles (the hyperscale JVs and funds) DLR has built since 2023.

Revenue composition (FY2025, total $6,112.7M, +10.0% YoY). “Rental and other services” was $5,968.9M (97.6%) and “Fee income and other” was $143.8M (2.4%) — but the fee line grew +98.3%, doubling year-on-year, and is the strategic growth lever (capital-light, scalable, lumpier). Geographically the book is roughly Americas ~45% / EMEA ~34% / Asia-Pacific ~21% (consistent with the post-Interxion/Teraco footprint), with a substantial euro/GBP/other foreign asset base. Revenue is overwhelmingly recurring and contractual on the rental side; the fee line is the growing non-rental layer.

The two-speed book, in numbers. This is the heart of the business:

• 0–1 MW + interconnection (retail colo): FY2025 bookings of ~$340M (a record, +35% YoY), with a record $96M quarter in Q4-2025 and a new record $98M in Q1-2026; interconnection bookings of ~$18–19M/quarter (+22% YoY in 2025); ~600 new logos added in each of 2024 and 2025; and a record 21% of 0–1 MW bookings AI-oriented in Q1-2026. This is the higher-quality, stickier, denser-connectivity book.
• >1 MW (hyperscale/wholesale): on a 100%-share basis, hyperscale leasing exceeded $800M in 2025, and Q1-2026 included the largest single megawatt lease in company history (a ~200 MW AI-inference deal with an AA-rated hyperscaler in Charlotte, delivering 2027–28). Pricing here runs >$180/kW/month (Q4-2025). This is the larger, more commodity book.

Bookings, backlog and pricing momentum. FY2025 total bookings were $1.2B (a second consecutive >$1B year, ~70% above the prior-five-year pace). Record total backlog hit ~$1.8B at 100% share (~$1.0B at DLR’s share) by Q1-2026, with ~$634M of commencements scheduled for 2026 and $152M for 2027+. Crucially, cash re-leasing spreads turned +6.7% in 2025 — the clearest financial evidence the wholesale supply/demand balance has flipped to landlords after years of flat-to-negative marks during the 2016–2020 glut. Same-capital cash NOI grew +4.5% constant-currency in 2025 (guide +4–5% for 2026).

Customer base. DLR serves >5,000 customers across cloud/IT services, digital content, networks/comms, financial services, manufacturing, energy, gaming, life sciences and consumer verticals. But the concentration is the tell: the largest customer is ~11.7% of annualized recurring revenue and no other single customer exceeds ~9.0% (so the top two are ~21%). These are widely understood to be hyperscalers (not named in the 10-K). This is the structural opposite of Equinix, whose largest customer is ~3% — and it is the signature of a wholesale/hyperscale-tilted portfolio.

Pressure-testing the “data gravity” pitch. DLR markets PlatformDIGITAL around “data gravity” — the idea that as enterprises accumulate data, the cost and latency of moving it pulls compute, storage and networking toward where the data already sits, so customers concentrate workloads in DLR’s connected campuses and become stickier over time. There is a real kernel here: for the retail/interconnection book, gravity is just another name for the network-effect/switching-cost moat, and it is genuine. But applied to the wholesale book it is largely marketing — a hyperscaler leasing a 200 MW block is not captured by data gravity; it is making a build-versus-buy decision on power, speed and price, and it can and does run the same calculation at a competitor or in-house at renewal. The honest reading is that data gravity describes a real force in the part of DLR’s business that is the minority of revenue, and is a narrative gloss over the part that is the majority. Investors should weight the connectivity metrics (interconnection bookings, ServiceFabric reach, new logos) as the true gravity signal, and treat the wholesale book as the commodity it is.

Corporate form. DLR has been a REIT since its 2004 IPO. As with any REIT, GAAP earnings are heavily suppressed by depreciation ($1,894.6M in FY2025) on a historical-cost real-estate base, so Core FFO/share — not GAAP EPS — is the metric management guides and the market underwrites. HQ is Dallas, TX (relocated from San Francisco); CEO Andrew Power has led since end-2022 (he was previously CFO); CFO is Matt Mercier; ~3,936 employees.

Verdict. A recurring-revenue, global data-center landlord with a genuinely improving fee-income leg — but a portfolio tilted toward the more commoditized wholesale segment, carrying real single-name customer concentration that its closest peer does not. The product mix, not the floor space, is the franchise — and DLR’s mix is structurally less premium than Equinix’s.

3. Industry Dynamics

Two sub-industries under one “data center” label. The data-center industry blends two structurally opposite businesses, and DLR straddles both — tilted more toward the weaker one than Equinix is:

• Retail colocation + interconnection — structurally GOOD. Metro-located facilities selling space/power by the cabinet plus dense partner ecosystems and cross-connects. High-margin, sticky, ~95% recurring. The top tier is a genuine oligopoly (Equinix, Digital Realty, NTT, and a long tail) with high, real barriers: ecosystem/network density, irreplaceable metro land and power, an accumulated interconnection graph, and switching costs. DLR plays here through its 0–1 MW + interconnection book, but its connectivity density is an order of magnitude behind Equinix’s (Equinix has ~500,000+ interconnections; DLR does not disclose a comparable figure, and ServiceFabric’s reach of ~300 cloud on-ramps and ~700 interconnected data centers, while material, is far smaller).
• Wholesale / hyperscale — structurally WEAKER. Large contiguous megawatt blocks leased to a single hyperscaler, or self-built by AWS/Azure/Google. Commodity power-and-shell economics, lower margin per MW, lease-like, low barriers (capital + power + land, replicable by any funded developer or by the hyperscaler itself). This is precisely why it commoditizes — and it is where DLR’s revenue is most concentrated.

The capital cycle is flashing late-stage (Marathon lens). Record capital is flooding the segment DLR leads. Private-equity data-center investment hit ~$45.7B in 2025 (a five-year high); PE is expected to fund ~$350B of data-center financing by 2028; and DC M&A topped $115B across ~95 deals (an all-time high in 2024). The mega-players are all in: Blackstone (the ~$10B QTS take-private; a >$25B Pennsylvania energy/DC hub; and now its own data-center REIT, BXDC), KKR (the ~$15B CyrusOne take-private; a $50B AI-DC JV), Vantage, Aligned, the CoreWeave-style “neoclouds,” and hyperscaler self-build. In Marathon’s framework, this is the textbook asset-growth/capital-flooding signal that predicts mean-reverting forward returns on capital. DLR is itself adding to the flood with its $3.25B Hyperscale Fund.

Why does this matter so much for DLR specifically? Because DLR’s forward value creation is development, and development returns are exactly what the capital cycle attacks. When a high return on capital appears (here, ~11–14% unlevered yields on hyperscale builds in 2024–25), capital floods toward it; the flood adds supply; the supply compresses the return back toward the cost of capital. The asset-growth anomaly — that the fastest-growing asset bases tend to deliver the worst forward returns — is the empirical fingerprint of this mechanism, and the data-center industry is growing its asset base as fast as any sector in the market right now. DLR is not a passive observer of this; via its $3.25B Hyperscale Fund and $7B Blackstone JV it is one of the larger pumps adding capacity. The capital-light structure protects DLR’s balance sheet from the overbuild, but it does not protect the return on the marginal build — and that marginal return is what the ~11.4% pipeline yield represents.

But supply is acutely tight right now — the bull’s counter. The capital cycle is currently masked by a power bottleneck. CBRE’s H2-2025 data shows primary-market vacancy at a record-low ~1.4%, record net absorption of ~2,498 MW in 2025 (versus the prior record ~1,810 MW), and — most tellingly — new capacity under construction in primary US markets declined for the first time since 2020 (5.9 GW at YE2025 vs 6.3 GW a year earlier). The binding constraint is not capital; it is power — grid-interconnection queues, generation and transmission upgrades, and permitting/zoning pushback (Northern Virginia/Loudoun, and overseas moratoria precedents in Dublin, Amsterdam and Singapore). That power scarcity is exactly what is producing record-low vacancy and the positive re-leasing-spread inflection despite record capital inflow.

Where DLR sits on AI demand. DLR is a more direct AI-training-capex play than Equinix: its >1 MW tilt plus its ~5 GW power bank and >3,500 MW of developable land (including >1,000 MW in Northern Virginia) let it capture gigawatt-scale training builds directly, where Equinix touches training only via its capital-light xScale JVs. DLR also plays the slower-arriving inference/interconnection layer (a private AI exchange, high-density colo, pre-installed liquid cooling) — but its moat there is weaker than Equinix’s. Net: DLR is higher-beta to the AI build-out, but also more exposed to the commoditized leg and to a future supply glut if power constraints ease.

Regulation. Power availability is the gating operational/regulatory constraint. Data-sovereignty mandates are a modest net positive (they favor distributed in-country footprints — DLR’s 25+ countries qualify). A US Senate (Warren) inquiry into PE data-center investment and the socialization of utility costs onto ratepayers is a watch-item, not yet material. ESG is largely immaterial financially except insofar as power/permitting gates development.

Verdict: mixed / two-tier. The retail-colo+interconnection sub-industry DLR participates in is structurally good, but DLR is the #2 there. The wholesale/hyperscale sub-industry where DLR leads is structurally weaker and sits mid-to-late in a capital cycle with record private-capital inflow — currently masked by an acute power-driven supply shortage. The honest read: DLR is enjoying a genuine cyclical upswing (landlord pricing power has returned) inside a structurally more-commoditized segment whose long-run returns the capital cycle threatens to mean-revert once power catches up.

4. Competitive Position

Naming the moat — a hybrid, asymmetric one. DLR’s competitive advantage is real but narrow and uneven across its two books:

1. On the 0–1 MW + interconnection book: a real but second-tier network-effect/interconnection moat (in Greenwald’s taxonomy, demand-side customer captivity reinforced by economies of scale). The evidence it is real: record 0–1 MW bookings (+35%), interconnection bookings +22%, ServiceFabric reaching 300+ cloud on-ramps and 700+ data centers, ~600 new logos a year, and genuine “connected campus” network effects. The evidence it is second-tier: no disclosed interconnection graph remotely approaching Equinix’s ~500,000 connections, a lower interconnection-revenue mix, and the fact that this is the minority of DLR’s revenue.

2. On the majority >1 MW / hyperscale book: scale without captivity (Greenwald’s term for genuine scale economies but low customer captivity). Hyperscale wholesale has low switching costs, commodity power-shell economics, and is replicable by any funded developer (Blackstone, KKR, Vantage, Aligned) or by the hyperscaler self-building. DLR’s edge here is not captivity — it is a set of supply-side advantages: (i) the largest global land-and-power bank / secured-power runway (~5 GW buildable IT capacity; >3,500 MW developable land), (ii) a 20-year development track record and speed (sourced-to-leased hyperscale in under 18 months in Charlotte), (iii) a global footprint that lets multi-region hyperscalers standardize, and (iv) connected-campus adjacency that situates wholesale next to the interconnection ecosystem. These are durable*-ish*, but they are precisely the kind of advantage that capital can partially buy — which is the whole point of the $45.7B/yr PE inflow.

The decisive quantitative test — development yields. GAAP returns are uninformative for a REIT (DLR’s reported GAAP ROIC of ~2.1% and the garbled ROE are depreciation artifacts; book value per share is near-zero because accumulated dividends in excess of earnings have hollowed out GAAP common equity). The right lens is the return on new capital deployed:

• DLR builds its development pipeline at a ~11.4–11.9% expected stabilized yield (Q4-2025: ~$10B underway at 11.9%; Q1-2026: ~$16.5B / 6 GW at 11.4% average, 61% pre-leased; 2026 yields guided to “stay double-digit”).
• Equinix cites ~26–27% stabilized cash-on-cash on its retail/interconnection book.

DLR’s development yield is less than half of Equinix’s — a direct, quantitative measure of the wholesale-versus-retail economics gap. An ~11.4% unlevered stabilized yield against a ~4–5% incremental cost of debt is a value-creating spread while it holds, but it is a commodity-development return, not a franchise return — and it is exactly the return the capital cycle pressures as more capital chases the same builds.

The private-capital pivot as a tell. DLR is deliberately moving hyperscale off its own balance sheet into JVs and funds (the Blackstone $7B JV; the inaugural $3.25B US Hyperscale Fund; the Mitsubishi JV; ~$15B of total dry powder), keeping a ~20% co-investment plus management/leasing/development fees (fee income doubled to $143.8M). This is the same capital-light model Equinix uses for xScale — strategically smart — and a tacit admission that DLR does not want to hold the lower-return, capital-hungry, more-competed hyperscale assets directly on its own equity. It is good capital allocation that simultaneously diagnoses where the franchise economics are weakest.

The DLR-vs-Equinix scorecard. The cleanest way to see why DLR is the lower-multiple #2 is to lay the two side by side on the axes that drive REIT value. The picture is consistent: DLR wins on absolute scale, growth-headline and price; Equinix wins on every quality metric.

The takeaway: DLR’s ~6–10% multiple discount to Equinix is real but thin against a quality gap that is wide and consistent. You are paying ~94% of the best-in-class multiple for the clearly second-best franchise.

Verdict: a real but narrow, asymmetric moat. DLR is (i) a clear #2 in the genuinely-moated retail-colo/interconnection niche (durable but sub-scale versus Equinix), and (ii) the scale leader in structurally weaker, lower-captivity hyperscale wholesale, where its edge is a land+power+execution supply advantage (partly buyable by capital) rather than customer captivity. The ~11.4% development yield versus Equinix’s ~26–27% cash-on-cash quantifies the quality gap. DLR is the lower-multiple #2 for a reason: more of its book is commodity. The current re-leasing-spread inflection and record-low vacancy are cyclical tailwinds (power scarcity), not proof of a strengthened structural moat.

5. Growth History and Forward Opportunities

The growth was bought, not earned per-share. Revenue compounded ~11.3% from $3,209M (2019) to $6,113M (2025), but the step-ups were dominated by all-stock M&A — Interxion (2020, ~$8.4B, EMEA) and Teraco (2022, ~$3.5B, Africa) — plus the Ascenty LatAm JV. Over the same window, the diluted FFO-basis share/unit count rose from ~262M (2020) to ~359M (Q1-2026), +37% / ~6.5% per year. The result is the single most important fact about DLR’s history:

Core FFO/share was essentially FLAT for four years — ~$6.61 (2021), ~$6.70 (2022), ~$6.55 (2023), ~$6.72 (2024) — and Nareit FFO/share actually declined ($6.20 in 2023 to $6.14 in 2024). Revenue, EBITDA and the asset base all grew; the share count grew as fast; rising rates and development drag absorbed the rest.

This is textbook value-destructive scale — growth that did not accrue to the per-share owner. It is the reason DLR is the lower-multiple #2 to Equinix, and it is the baseline against which the recent inflection must be judged.

The inflection (2025–26). Core FFO/share grew +10% to $7.39 in 2025, is guided to $8.00–$8.10 (~+9%) for 2026 (raised $0.10 in Q1-2026), and printed $2.04 in Q1-2026 (+15% YoY, +11% constant-currency). This is the first real per-share growth in half a decade. The drivers are a mix of cyclical and structural:

• Cyclical pricing: cash re-leasing spreads turned +6.7% in 2025 (versus years flat/negative), with 2026 guidance raised to +6.5–8.5% blended. The vivid Q1-2026 tell was a +74% cash re-leasing spread on >1 MW renewals (on a small $32M volume in Vienna/London/Silicon Valley) — a signal that legacy below-market wholesale leases can re-price sharply up. The durable number, though, is the +6.5–8.5% blended (>80% of renewal volume is 0–1 MW at ~+4.3%).
• Occupancy / power scarcity: record-low ~1.4% market vacancy; power-based occupancy ~91% same-capital; guided +50–100 bps in 2026.
• Fee income (capital-light): doubled to $143.8M, a direct contributor to Q1-2026 growth.
• Deleveraging: lower interest drag as net debt/Adj EBITDA fell to 4.7x.

The forward engine, quantified. The development pipeline scaled +60% sequentially to $16.5B at 100% share / ~6 GW by Q1-2026, with 1.2 GW under construction, 61% pre-leased, at ~11.4% average expected yield. The record $1.8B backlog (~$1.0B at DLR’s share) commences ratably into 2028, giving “strong visibility into 2027 and 2028.” The land-and-power bank (~5 GW buildable; >3,500 MW developable, including a new 873-acre Atlanta parcel supporting a ~1 GW campus) is the scarce input that lets DLR convert AI demand into NOI. Bookings momentum is strong (FY2025 $1.2B; Q1-2026 the second-highest bookings quarter ever).

Verdict: mixed-quality, improving. The history is low-quality growth — volume/M&A/dilution-driven, producing flat per-share FFO for four years (the defining flaw). The forward leg is higher quality on two fronts — capital-light fee income (genuinely accretive, scalable) and cyclical pricing power (positive, accelerating spreads from power scarcity) — but the bulk of the $16.5B pipeline is still commodity hyperscale development at ~11.4% yields (less than half Equinix’s cash-on-cash), and the per-share reacceleration is only two years old and coincides with a once-in-a-cycle AI/power-scarcity spike. The durability of the per-share inflection is the open question on which the whole thesis turns.

6. Financial Quality

FFO/Core FFO/AFFO — the central per-share story. Already covered in Section 5 on the inflection; the quality nuance is that the recent reacceleration is real but young and cycle-aided, and the burden of proof that the flat-per-share era is permanently over has not yet been met (two good years against a five-year record of stagnation). Nareit FFO available to common+units rose $1,915.7M (2023) → $2,027.1M (2024) → $2,399.0M (2025).

AFFO is a flattered distributable-cash proxy. DLR does not publish AFFO in the 10-K (it is in the supplement). The Q1-2026 AFFO payout ratio was ~64% (falling), implying AFFO/share of ~$7.6 run-rate against the $4.88 dividend — comfortable coverage with ~36% retained. But the same conceptual caveat that applied to Equinix applies here: DLR classifies the overwhelming majority of capex as “growth/development” and subtracts only a small “recurring/maintenance capex” line in AFFO (~$169M in Q4-2025; perhaps ~$600–700M annualized including tenant improvements and leasing costs) against a ~$40B+ gross real-estate base and $1.9B of annual depreciation. AFFO is therefore an optimistic distributable-cash figure — a stricter maintenance assumption would lower it and raise the payout. The critique is milder than at Equinix (wholesale shells genuinely refresh less often than dense retail colo), but the direction is the same.

Margins. GAAP operating margin is depressed (~15.1% FY2025) by heavy D&A, transaction/integration expense ($185.1M FY2025, up from $93.9M — integration not fully digested) and recurring impairments ($78.6M FY2025; $191.2M FY2024). Reported EBITDA margin (~46% FY2025) is structurally below Equinix’s (~49% rising to ~51%), diluted by the lower-margin Interxion/Teraco colo and transaction noise; Adjusted EBITDA margin (after add-backs) is higher. Operating leverage is real on the upswing (2026 guide: revenue and Adj EBITDA both >10% constant-currency), but the incremental flow-through is lower than Equinix’s given DLR’s lower recurring/interconnection mix.

Cash flow — deeply FCF-negative by design. Operating cash flow was $2,412M in FY2025 (up from $2,261M in 2024 and $1,635M in 2023). But total development/improvement capex runs ~$3.0–3.5B/yr gross, and with the $16.5B / 6 GW pipeline ramping, DLR is structurally free-cash-flow-negative after development capex — the gap funded by equity (ATM/forward), debt, asset contributions to JVs/funds, and JV/LP capital. (Note: ROIC’s “free cash flow” line equals operating cash flow only and does not subtract development capex — it is not true FCF.) Dividends paid were $1,728.5M in FY2025 (rising in aggregate dollars on more shares despite the flat per-share rate).

How flattered is AFFO? A sensitivity. This matters because AFFO is the number that “covers” the dividend and frames the P/AFFO multiple. DLR subtracts a recurring/maintenance-capex figure of ~$169M in Q4-2025 (call it ~$650M annualized including tenant improvements and leasing commissions) in arriving at AFFO. Against a ~$40B+ gross real-estate base and $1.9B of annual depreciation, that is a maintenance assumption of roughly 1.5–2% of gross assets — defensible for long-life wholesale shells, but at the optimistic end. If a stricter analyst assumed maintenance capex were, say, ~$300M higher per year (closer to ~2.5% of the depreciating asset base), AFFO/share would fall by roughly ~$0.85, the run-rate would drop from ~$7.6 to ~$6.75, the AFFO payout on the $4.88 dividend would rise from ~64% to ~72%, and the P/AFFO multiple at $188 would climb from ~24.8x to ~28x — i.e., above Equinix. The point is not that DLR’s AFFO is wrong, but that the comfortable 64% payout and the apparent multiple discount to Equinix are partly artifacts of a generous maintenance-capex convention; the true distributable-cash cushion is thinner than the headline suggests.

Balance sheet — the deleveraging is the clearest positive. Total debt was ~$19.66B at YE2025 (net debt ~$15.0B). The headline leverage metric — net debt/Adjusted EBITDA — fell to 4.7x at Q1-2026 (a multi-year low), down from ~5.1x at YE2025, ~5.8x in 2023, and ~7x at the 2022 post-Teraco/Interxion peak — well inside the <5.5x target. This was achieved via ~$3.65B of equity raised in 2024 + ~$1.1B in 2025, JV/fund capital recycling, and asset contributions. Financial policy targets are sensible (net debt/Adj EBITDA <5.5x, fixed-charge coverage >3x, floating-rate debt <20%); the weighted-average debt term is ~5.0 years and well-laddered; ratings are investment-grade BBB / Baa2, with global unsecured access in USD, EUR and GBP.

The maturity profile is the other reason the deleveraging matters: a weighted-average term of ~5.0 years means roughly a fifth of the ~$19.7B stack reprices each year, so in a higher-for-longer rate world the blended coupon grinds upward even without new borrowing — which is precisely why management prioritized cutting absolute leverage (via fund/JV capital) over simply terming out. With floating-rate debt held under ~20% of the total and a well-laddered maturity wall, the refinancing risk is manageable but not trivial; it is the single mechanical reason the stock carries a −0.43 interest-rate factor loading despite being a “growth” AI name in the popular narrative. A sustained back-up in long rates would simultaneously raise DLR’s refinancing cost and compress the multiple on its long-duration cash flows — the two effects that produced the 2022–23 drawdown.

FX. The large euro/GBP/other foreign asset base (Interxion, Teraco, Ascenty, APAC) is managed via natural hedging — DLR issues foreign-currency-denominated debt (e.g., €850M 3.875% notes due 2034, issued June 2025) plus cross-currency swaps to match foreign assets, so FX moves hit translated NOI and goodwill (goodwill ~$9.7B swings with FX) but are partly offset on the liability side. FX is a reported-results swing factor (it helped Q4-2025 reported same-capital NOI to +8.6% vs +4.5% constant-currency) but is a translation issue, not a solvency one.

ROIC/ROE. GAAP returns are meaningless here (REIT depreciation). The right operating metrics are the development yield (~11.4%), same-capital cash NOI growth (~4–5%) and the Core FFO/share trajectory.

Verdict. Economics are improving on the cyclical upswing (Core FFO/share +10%/+9%, deleveraging to 4.7x, positive re-leasing spreads) but remain structurally lower-quality than Equinix — roughly half the development yield, a lower recurring/interconnection mix, a deeper FCF deficit, and a half-decade of flat per-share earnings proving prior scale and M&A did not create per-share value. The recent reacceleration is real but young and cycle-aided; AFFO is an optimistic distributable-cash proxy; true post-development FCF is negative.

7. Capital Allocation

The defining 2023–26 story: the private-capital / capital-recycling pivot. Under CEO Andrew Power and CFO Matt Mercier, DLR reoriented from funding development with dilutive common equity + debt toward funding it with third-party private capital, keeping ~20% co-investment stakes plus asset-management/leasing/development fees. The milestones:

• December 2023: the ~$7B hyperscale development JV with Blackstone (Blackstone funds take 80% for ~$700M initial capital; DLR keeps 20%; four campuses in Frankfurt, Paris and Northern Virginia; ~500 MW IT load at full build). (Note: ROIC’s profile dates this “July 2024”; the verified announcement date from the Blackstone/PRNewswire release is 2023-12-07.)
• 2023–2025: an expanded GI Partners JV, a Realty Income data-center JV, TPG Real Estate, and a Mitsubishi JV (January 2025).
• March 30, 2026: the final close of the inaugural US Hyperscale Data Center Fund at $3.25B of LP equity commitments (global institutional LPs — public pensions, SWFs, endowments, insurers), with DLR as manager and ~20% retained. (In May 2025, DLR contributed operating data centers + development projects for ~$937M gross proceeds, recognizing an ~$873M disposition gain and retaining a $661M investment.)
• The monetization: fee income doubled to $143.8M (+98.3%) in 2025.

This is strategically smart — the same capital-light, co-invest-plus-fee model Equinix uses for xScale. It simultaneously (a) funds the 6 GW build off-balance-sheet, (b) drove the deleveraging to 4.7x, and © reduces reliance on dilutive common equity. It is also, as noted, a tacit admission of where DLR’s direct economics are weakest.

Equity issuance / dilution — the historical Achilles heel. DLR has been a serial equity issuer: ~$3.65B raised in 2024 + ~$1.1B in 2025 via ATM/forward, with ~$1.9B still available under the 2024 ATM, and $870.6M of common stock issued in Q1-2026 alone. This ~6.5%/yr dilution is precisely why a half-decade of revenue/EBITDA/M&A growth produced flat Core FFO/share through 2024. The private-capital pivot is partly designed to substitute LP capital for dilutive ATM — and whether per-share growth is now durable hinges on slowing the share-issuance pace. This is the single most important capital-allocation watch-item.

Dividend — held flat since 2022 (a positive here). The common dividend has been frozen at $4.88/yr ($1.22/quarter) since early 2022, after ~17 years of consecutive growth. The freeze is a capital-allocation positive in context: the dividend was stretched on AFFO during the flat-FFO years, and holding it while FFO/share now grows lets the AFFO payout ratio fall (to ~64%) and retains more internal capital for the build, reducing dilution. The dividend is now well-covered and de-risked; a resumption of dividend growth would be a bullish signal that the per-share inflection is durable. The yield is ~2.6% — the low end of DLR’s own history. Preferred dividends are modest (~$40.7M/yr; ~$0.73B outstanding).

M&A history. Interxion (2020, ~$8.4B all-stock) was a transformative EMEA entry; Teraco (2022, ~$3.5B for control, the Africa leader, with a minority put right that GAAP treats as share-settled — a dilution overhang). Integration is not fully digested (transactions+integration expense rose to $185.1M in FY2025), and the all-stock Interxion deal is a prime example of growth-via-dilution that did not lift per-share FFO for years. The pivot away from large M&A toward organic build + funds is an improvement.

Compensation. The 2026 proxy anchors pay on Core FFO Per Share, Net Income from Operations, and Relative TSR — the right per-share REIT metrics, with no raw revenue/MW/empire-growth metric that would reward dilutive scale. The annual incentive is ~75% financial / ~25% individual; the LTI uses RSUs + performance shares measured over multi-year periods on Core FFO/share, constant-currency Core FFO and relative TSR. This is well-aligned and notably better than a pure-growth scorecard.

Insider signal — weak/absent. All directors and executive officers as a group (12 persons) own 516,978 shares — less than 1% of the ~349M shares outstanding. Across ~120 recent Form 4 filings, the transaction codes were grants (90), tax-withholding (31), exercises (8), open-market sales (7), gifts (2) — and zero open-market purchases (code P). No insider bought stock even during the deep 2023 drawdown. This parallels Equinix exactly (zero P buys, ~0.27% ownership): not a governance red flag, but no positive conviction signal either.

Verdict. Capital allocation is improving and now broadly intelligent — the private-capital pivot, the disciplined dividend freeze, the deleveraging, and the Core-FFO/share-anchored comp are genuine positives that mark a turn from the prior era’s value-destructive growth-via-dilution. But the historical record is poor, and the proof is forward-looking: whether share issuance genuinely slows from here. Insiders own <1% with zero open-market buys.

8. Changes and Headwinds — Last Two Years

The strategic pivot (covered in Section 7) is the dominant change — the shift to private capital under Power/Mercier, funding the build with third-party LP equity rather than dilution and debt. Alongside it:

• The re-leasing-spread inflection. After years of flat-to-negative cash renewal spreads in the oversupplied wholesale era, spreads turned solidly positive (+6.7% FY2025) and the 2026 guidance was raised to +6.5–8.5%. This is the clearest evidence the wholesale supply/demand balance flipped to landlords.
• The AI demand wave / record leasing. Q1-2026 saw >$700M of new leases signed (~70% above the prior-highest quarter), a record $1.8B backlog, the largest single lease in company history (the ~200 MW AI-inference deal with an AA-rated hyperscaler in Charlotte), and the development pipeline scaling +60% to $16.5B.
• Guidance raises. 2026 Core FFO/share lifted to $8.00–$8.10 (~+9%).
• Aggressive deleveraging. Net debt/Adj EBITDA from ~7x (2022) to 4.7x (Q1-2026); AFFO payout to ~64%.
• Geographic land-and-power expansion. First Barcelona data center (May 2026), a Malaysia/Cyberjaya 32 MW platform (June 2026), an ORCA quantum lab in London, and an ePlus private-AI managed service on PlatformDIGITAL.
• Leadership. Power as CEO since end-2022; Mercier as CFO since 2023. No disruptive board turnover.

Headwinds. The same forces driving the bull case are the bear’s red flags: record private capital (including DLR’s own funds and Blackstone’s new BXDC REIT) flooding into data centers is the Marathon capital-cycle warning; EU energy/permitting and sustainability mandates raise development cost and constrain supply (a double-edged sword — they also keep supply tight); and AI-financing-fragility headlines (JPMorgan’s “$4.1T AI debt story,” June 2026) feed the overbuild narrative. On the short-seller question: there is no active 2024–26 short campaign that surfaced; Jim Chanos’s famous c.2022–23 short of data-center REITs (the “hyperscalers will self-build and commoditize wholesale” thesis) is historical bear framing that the AI re-leasing inflection has since worked against.

Verdict. On a two-year view these changes net strengthen the thesis — the private-capital pivot and deleveraging materially de-risk the balance sheet and reduce the dilution drag, while the re-leasing inflection and record backlog validate returning pricing power. But the strength arrives precisely at peak capital enthusiasm, and the same demand that lifted the stock to 96th-percentile valuations is the demand the capital cycle is built to over-serve.

9. Risk Analysis (Risk Matrix)

The two risks that matter most are the capital cycle (the structural threat to the ~11.4% development yield that underwrites the whole forward story) and valuation/multiple compression (because the stock is priced for the bull case to come true). They are correlated: a capital-cycle disappointment is the most likely trigger for multiple compression. Catastrophic/total-loss risk is low — DLR is an investment-grade, hard-asset REIT with a global, diversified, mostly investment-grade tenant base and a deleveraged balance sheet; the realistic downside is a de-rating and a slower-growth grind, not insolvency.

10. Valuation Discussion (Embedded Expectations)

Where it trades (2026-06-18, $188.15). Market cap ~$62.2B; enterprise value ~$81.7B (diluted ~$83.2B); TTM revenue $6,340M, EBITDA $2,920M. The headline multiples:

On its own history, DLR is near richest-ever. Own-history valuation percentiles read P/B 96.7th, P/S 96.4th, composite 80th (the P/E percentile is ignored — a GAAP REIT artifact). The dividend yield at ~2.6% sits at the low end of DLR’s own range (it was ~3.5–4.5% in 2022–23), corroborating the rich valuation. EV/EBITDA at ~28x is at the top of the 10-year band but not a clear outlier versus the 2020–21 and 2024 peaks (~29x); the P/S and P/B percentiles are the more extreme reads.

The Equinix cross-read — the key relative call. Equinix (2026-06-12, ~$1,056) trades at ~29.5x EV/EBITDA, ~25x forward AFFO, the 98th percentile of its own P/B and P/S history, and a ~1.8% yield. So DLR is only ~1.5 turns cheaper on EV/EBITDA, ~6–10% cheaper on forward distributable-cash multiple (~23.4x Core FFO / ~24.8x AFFO vs Equinix ~25x AFFO), and offers a higher yield (2.6% vs 1.8%). Is that discount justified? Yes — and arguably it should be wider. DLR’s book is structurally lower-quality on every axis that matters: half the development yield (~11.4% vs ~26–27%); real customer concentration (~11.7% / ~21% vs ~3%); interconnection density an order of magnitude smaller; more wholesale/commodity revenue mix; a half-decade of flat per-share FFO versus Equinix’s historical low-teens compounding; and a lower Adjusted EBITDA margin. A ~6–10% multiple discount looks like thin compensation for that gap — the market is pricing two near-peers when the quality gap warrants a wider spread.

The private-market reality check. TTM EBITDA / EV is a ~3.6% gross yield; on a stabilized NOI basis the implied cap rate sits in the ~4.5–5.5% area. Private-market stabilized data-center assets underwrite to ~5.75–7% gross yields (~6.5% for CMBS-financed stabilized DC). DLR’s implied ~4.5–5.5% cap rate is below where private DC assets trade — i.e., the public market values DLR’s portfolio at a premium to private stabilized assets. That is partly justified (you are buying ~11.4%-yield development optionality and platform/fee value, not just stabilized NOI), but it confirms DLR is not cheap on a NAV/cap-rate basis; the public multiple already capitalizes the development pipeline.

A reverse read on the multiple. Strip the valuation to its components. A buyer at $188 gets a ~2.6% starting cash yield (the frozen dividend) plus whatever per-share Core FFO growth DLR can compound, plus or minus the multiple change. For the stock to deliver a ~9–10% total return with no multiple help, DLR must compound Core FFO/share at ~6.5–7.5% in perpetuity — comfortably above its actual 2021–24 record of ~0% and only sustainable if the cyclical drivers (spreads, occupancy, fee income) become structural. If, instead, the multiple reverts even partway toward DLR’s own pre-AI ~18–20x norm over five years (a ~2–3%/yr drag), the required Core FFO/share growth to still clear ~9–10% rises toward ~9–12%/yr — essentially the bull case running uninterrupted. Put differently: at today’s price, avoiding a poor outcome requires the per-share inflection to be both real and durable; the multiple is doing none of the work for you, and may work against you.

Embedded expectations — what $188 underwrites. At ~23.4x forward Core FFO with a frozen 2.6% yield, the price embeds: (1) the per-share inflection is durable/structural — sustained ~8–10% Core FFO/share compounding for years, not a two-year cyclical pop; (2) ~11.4% development yields hold as record PE capital floods in (the capital cycle does not mean-revert returns soon); (3) re-leasing spreads stay positive/accelerating (power scarcity persists); (4) the fee/private-capital platform scales into a meaningful recurring earnings stream; (5) share issuance slows so growth accrues per-share; and (6) no multiple compression from the 96th-percentile P/S and P/B. That is largely the bull operating stack underwritten as the base case. Correctly priced: the demand/power-scarcity tailwind and the deleveraging/fee-pivot quality improvement are real. Aggressive: extrapolating a two-year, cycle-aided reacceleration into permanent compounding, assuming ~11.4% yields survive the capital cycle, and paying 96th-percentile multiples and a sub-private-market cap rate for a book half the quality of Equinix’s on development economics. The margin of safety is thin.

Scenarios (ranges, NOT price targets; framed as multiple × normalized Core FFO/share).

• Bear (~$120–150): the per-share inflection proves cyclical — power constraints ease, PE supply floods, re-leasing spreads fade toward flat, dev yields compress toward ~8–9%, dilution resumes; Core FFO/share growth slows to low-single-digits (~$8.0–8.5 normalized) and the multiple de-rates from ~23x toward ~16–18x (DLR’s 2022–23 trough zone) as the AI re-rating unwinds.
• Base (~$175–205, ~where it trades): DLR delivers the guide (~$8.05 in 2026, ~8–9% per-share growth into 2027 on backlog), spreads stay positive, fee income scales, leverage holds ~4.7x; the multiple holds ~22–24x forward Core FFO. Total return ≈ ~2.6% yield + high-single-digit FFO growth, less modest multiple give-back = low-double-digit annualized at best. “Improving #2 at a fair-to-full price.”
• Bull (~$220–260): the power-scarcity pricing regime persists, >1 MW mark-to-market re-pricing (the +74% tell) proves large and durable, dev yields hold double-digit through the capital cycle, the fee platform compounds, dilution genuinely slows, and DLR sustains low-double-digit per-share growth; the market keeps (or modestly expands) the ~24–26x multiple as DLR re-rates toward Equinix on a narrowing quality gap (~$9–10 normalized Core FFO × ~25x).

Verdict: full, not cheap. DLR trades near the richest end of its own history, at ~23.4x forward Core FFO / ~24.8x P/AFFO / ~28x EV/EBITDA, and at an implied cap rate below private-market stabilized DC trades. Its ~6–10% Core-FFO-multiple discount to Equinix is real but thin given a materially lower-quality book. The price underwrites the bull operating stack as the base case; embedded expectations are aggressive relative to the structural quality of the franchise.

11. Variant Perception

Consensus. DLR is an AI-infrastructure winner mid-re-rating; the lost half-decade of flat per-share FFO is over; per-share Core FFO is inflecting (+10%/+9%) on AI/power-scarcity demand; the deleveraged balance sheet plus the private-capital fee platform make it a higher-quality, durably-compounding REIT. The stock is up ~27% YTD in 2026, near 52-week highs, with consensus Core FFO of ~$8.06 and a Buy-tilted sell-side (e.g., Truist Buy, PT $208). The market treats DLR as a slightly-cheaper, higher-yield way to own the same AI-data-center theme as Equinix.

Strongest bull case. Power and land scarcity have handed wholesale landlords genuine, returning pricing power (record-low 1.4% vacancy, +6.7% FY2025 spreads accelerating to +6.5–8.5%, the +74% >1 MW mark-to-market tell) — a structural, multi-year regime, not a blip. DLR’s ~5 GW power bank + >3,500 MW of developable land + 20-year execution track record make it one of the few players that can actually deliver gigawatt AI capacity on schedule. The capital-light private-capital platform (fee income doubled to $143.8M) adds a high-margin, scalable, capital-efficient earnings leg while funding the 6 GW / $16.5B pipeline off-balance-sheet. Deleveraging to 4.7x removes the rate drag. The result is durable low-double-digit per-share Core FFO growth that re-rates DLR toward Equinix.

Strongest bear case. This is a late-stage capital cycle — ~$45.7B/yr PE inflow, $115B+ DC M&A, Blackstone/KKR/Vantage/Aligned plus hyperscaler self-build all chasing the same builds — and the ~11.4% development yield (half Equinix’s) is a commodity return that mean-reverts once power constraints ease. DLR’s core wholesale economics are structurally commodity (low captivity, scale-without-captivity, replicable by capital). The per-share inflection is only two years old and coincides with a once-in-a-cycle AI spike, atop a five-year record of flat per-share FFO driven by serial ~6.5%/yr dilution and all-stock M&A. Customer concentration is real (largest ~11.7%, top-2 ~21%, presumed hyperscalers who can self-build). And at 96th-percentile P/S/P/B and a sub-private-market cap rate, the market is capitalizing a cyclical tailwind as a permanent structural re-rating. Zero insider open-market buys; <1% insider ownership.

The 3–5 assumptions that matter most (with falsification tests):

1. Do ~11.4% development yields hold through the capital cycle? (Load-bearing.) Falsify bull: dev yields print <~9–10% on new starts / spreads compress. Falsify bear: yields hold double-digit two-plus years into the PE inflow.
2. Is the per-share Core FFO reacceleration structural or a two-year cyclical pop? Falsify bull: per-share growth decelerates to low-single-digits as spreads/occupancy plateau. Falsify bear: a third-plus consecutive year of ~8–10% per-share growth with slowing issuance.
3. Does share issuance genuinely slow (does the fund/fee model substitute for dilutive ATM)? Falsify bull: ATM issuance reaccelerates and dilutes away FFO growth (the historical pattern repeats). Falsify bear: diluted share/unit count growth falls below ~2–3%/yr while FFO/share grows.
4. Does the power-scarcity pricing regime persist? Falsify bull: vacancy rises / under-construction pipeline re-accelerates / spreads fade to flat. Falsify bear: spreads stay +6%+ and the >1 MW mark-to-market repricing proves large/durable.
5. Can the ~96th-percentile valuation / sub-private-market cap rate persist? Falsify bull: multiple de-rates toward DLR’s own ~18–20x history. Falsify bear: DLR re-rates up toward Equinix on a narrowing quality gap.

The factor-positioning read. A quantitative factor model characterizes DLR as an in-favor, momentum-strong, rate-sensitive REIT recovery — only partially an “AI trade” in the statistical tape. Its dominant loading is Sector Real Estate +1.06 with InterestRate −0.43 the single largest factor; the Momentum loading is only ~+0.095 (so it is not a crowded momentum name in factor terms, despite a huge price run); beta is 0.88 with a positive alpha (+0.08). The leaderboard shows an explosive recent run (m6 +65%, m3 +43% annualized; y3 +25%/yr) but a y5 laggard at +6.6%/yr — the profile of a long-duration REIT that round-tripped a rate-driven drawdown and re-rated on the AI demand story. Its factor-similar peers are all broad-REIT/real-estate ETFs (SRVR, XLRE, ICF, IYR) — no single-stock twin; in factor space DLR trades like the real-estate sector, not like a tech/AI name. The variant view: the quality gap to Equinix, the capital-cycle risk, and the two-year-young per-share inflection are underpriced at a 96th-percentile own-history valuation, and the dominant −0.43 rate sensitivity is the key risk the AI-growth narrative is glossing over.

12. Fact vs. Interpretation Table

13. Open Questions

1. Exact FY2025 AFFO/share and recurring-capex as % of revenue (in the supplement, not the 10-K) — needed to size precisely how flattered AFFO is; this memo used a ~$7.6 run-rate implied by the 64% payout.
2. Do ~11.4% development yields hold as $45.7B/yr of PE capital competes? The single load-bearing capital-cycle question.
3. Is the >1 MW mark-to-market repricing opportunity (the +74% Q1-26 tell) large and durable across the legacy book, or a one-off on three deals?
4. Total interconnection count for DLR (to put numerically against Equinix’s ~500,000) — not disclosed; ServiceFabric reach (300 on-ramps / 700 DCs) is the closest proxy.
5. Will diluted share/unit count growth genuinely slow below ~2–3%/yr now that the fund/fee model can substitute for dilutive ATM?
6. Precise net debt/Adjusted EBITDA bridge (management’s 4.7x vs ROIC’s unadjusted ~5.3x — the gap is EBITDA adjustments + pro-forma JV treatment) and the full year-by-year debt-maturity ladder.
7. Exact terms/dates of the Mitsubishi, GI/TPG and Realty Income JVs (the broader recycling roster beyond Blackstone and the flagship Fund).

14. What Must Be True

Bull case — what must be true. (1) The per-share Core FFO reacceleration is structural, not a two-year cyclical pop — DLR compounds Core FFO/share ~8–10% for several more years. (2) Development yields hold double-digit (~11%+) through the capital cycle, even as record PE capital and DLR’s own funds add supply. (3) Re-leasing spreads stay positive (the power-scarcity pricing regime persists). (4) Share issuance slows materially (the fund/fee model substitutes for dilutive ATM), so growth accrues per-share. (5) The fee platform scales into a durable, high-margin recurring stream. Falsification test: any one of — development yields on new starts printing <~9–10%; per-share Core FFO growth decelerating to low-single-digits; or diluted share count growth staying >~5%/yr while FFO/share stalls — breaks the bull thesis. The cleanest single confirmation would be a third consecutive year of ~8–10% per-share growth with the share count growing under ~3%.

Bear case — what must be true. (1) The current pricing power is cyclical — power constraints ease (the grid catches up, under-construction capacity re-accelerates), and re-leasing spreads fade toward flat. (2) The ~11.4% development yield mean-reverts toward ~8–9% as the capital cycle plays out. (3) DLR resumes dilutive issuance and per-share growth stalls again, as it did 2021–2024. (4) The 96th-percentile valuation de-rates toward DLR’s own ~18–20x history. Falsification test: development yields holding double-digit two-plus years into the PE inflow, spreads staying +6%+, and the share count growth slowing while per-share FFO compounds — that combination would prove the inflection is structural and break the bear thesis.

The two cases share one pivot: the durability of the development yield and the per-share inflection through the capital cycle. Everything else — the multiple, the dividend, the relative call versus Equinix — follows from that.

15. Source Appendix

See the Source Appendix (Appendix B) for the full, dated source list. Principal sources: Digital Realty FY2025 Form 10-K (filed 2026-02-13); Q1-2026 Form 10-Q (filed 2026-05-01); Q4-2025 and Q1-2026 earnings-call transcripts (2026-02-05 and 2026-04-23); the 2026 DEF 14A (filed 2026-04-17); the EDGAR Form 3/4/5 corpus (CIK 0001297996); the Blackstone/Digital Realty JV release (2023-12-07) and the US Hyperscale Fund final-close release (2026-03-30); CBRE North America Data Center Trends H2 2025; aggregated financials/ratios/enterprise value; own-history valuation percentiles and a news feed (2026-06-18/20); a quantitative factor model (2026-06-18); a five-year daily price history; and public Equinix disclosures for the peer cross-read.

APPENDIX A — Standard Diligence Questionnaire

Supplemental to the memo. Fact / Interpretation / Assumption labels applied where it matters. Where a question does not map to a data-center REIT, the correct sector analog is given.

General

What thoughtful questions have other investors asked about this company? The recurring institutional debates are: (1) Is the wholesale data-center business commoditizing? — the Chanos-era short thesis (hyperscalers self-build; oversupply crushes pricing), which the 2024–26 AI re-leasing inflection has worked against, at least cyclically. (2) Will per-share FFO finally grow? — after a half-decade of flat Core FFO/share (~$6.55–6.72, 2021–24) despite revenue/EBITDA growth, the +10%/+9% 2025–26 inflection is the central question of durability. (3) Is the private-capital pivot accretive or an admission of weak direct economics? (4) How much of the AI capex actually lands in DLR’s footprint vs self-build/neoclouds? (5) Is DLR cheap or expensive relative to Equinix given a ~6–10% multiple discount against a materially lower-quality book.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: on a multiple and spread basis, closer to a cyclical high — re-leasing spreads inflected positive (+6.7%), vacancy is at a record low (~1.4%), and the valuation is at the 96th percentile of its own history. On an absolute per-share-earnings basis, DLR is early in a recovery off a flat half-decade (Core FFO/share just $7.39, guided ~$8.05). Both can be true: a recovering earnings base at a top-of-cycle multiple.

Driven by the external environment or internal actions? Both. External: the AI demand wave + a power-scarcity supply bottleneck driving record absorption and positive spreads (the dominant driver). Internal: deleveraging (~7x → 4.7x), the private-capital/fee pivot, and the disciplined dividend freeze.

How stable are revenues? High — overwhelmingly recurring, contractual rental income (~97.6% of revenue) on long-dated leases with built-in escalators, plus a growing (lumpier) fee-income line. Wholesale leases are long-dated; churn is “modest.” The risk is concentration, not stability: the top customer is ~11.7% of ARR.

Outlook for products/services? Strongly positive on demand (AI training + inference + hybrid cloud + interconnection); the question is supply/returns, not demand.

How big will this market be — growing, shrinking, domestic or international? Large and growing globally; management/industry cite a ~$250B+ TAM with AI the fastest-growing sub-segment. DLR is genuinely global (Americas ~45% / EMEA ~34% / APAC ~21%; 25+ countries).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More — record private capital (~$45.7B in 2025) and new entrants (Blackstone’s BXDC, KKR/CyrusOne, Vantage, Aligned, neoclouds, hyperscaler self-build) are flooding in. Today masked by power scarcity.

How profitable is the business (ROIC, ROE)? GAAP ROIC/ROE are meaningless (REIT depreciation artifacts). The right metric is development yield: ~11.4% stabilized — value-creating over a ~4–5% cost of debt, but half of Equinix’s ~26–27% retail cash-on-cash. Same-capital cash NOI growth ~4–5%.

How profitable is the industry — how many competitors, what barriers to entry? Two-tier. Retail colo+interconnection: oligopoly (EQIX/DLR/NTT), high barriers (ecosystem density, metro land/power, switching costs). Wholesale/hyperscale: structurally weaker, low switching costs, replicable by funded developers — barriers are land+power+execution, partly buyable by capital.

Can the business be easily understood? Yes at the model level (a global data-center landlord with a connectivity overlay and a fee-management arm), though the REIT accounting (FFO/AFFO vs GAAP), the JV/fund web, and FX add complexity.

Can it be undermined by foreign low-cost labor? No — it is a hard-asset, location-and-power business; labor is not the cost driver.

Do brands matter? Modestly. “PlatformDIGITAL”/connectivity ecosystem matter for the retail book; for hyperscale wholesale, the “brand” is execution reliability and the ability to deliver power on schedule — not consumer brand.

What is the nature of competition? Securing power and land, delivery speed/reliability, global footprint, price, and (for retail) ecosystem density. For wholesale, increasingly a capital-and-power race.

Customers’ switching costs? High for the dense retail/interconnection book (re-architecting connectivity is painful); low for single-tenant wholesale (the customer can move or self-build at renewal) — which is why the +74% one-off mark-to-market is notable but the durable spread is the colo-weighted +6.5–8.5%.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? Yes, conceptually: the ~5 GW land-and-power bank / >3,500 MW developable land carries embedded value not fully reflected at historical cost; the ~20% co-investment stakes in JVs/funds and the recurring fee stream are platform value. Conversely, GAAP common equity is hollowed out by accumulated dividends in excess of earnings (so book value/share is near-zero — not a sign of distress, a REIT artifact).

Off-balance-sheet liabilities? The JV/fund structures move hyperscale development off DLR’s balance sheet (DLR’s share is equity-method); DLR’s economic exposure is its ~20% co-invest + fees, not full consolidation. Operating commitments and the development pipeline funding obligation are the main forward calls on capital. Teraco’s minority put is a GAAP share-settled dilution overhang.

How conservative is the accounting? Reasonable for a REIT. Positive QoE finding: FY2025 GAAP net income (+117%) was flattered by ~$995.6M of disposition gains (~$873M from the May-2025 Fund contribution), but Nareit FFO explicitly excludes these — Core FFO is clean of them. The aggressiveness sits in AFFO’s low recurring-capex assumption (see below), not in headline FFO.

How CapEx-hungry is the business? Very. ~$3.0–3.5B/yr of development/improvement capex against ~$2.4B operating cash flow makes DLR structurally free-cash-flow-negative after development — by design; the build is the cash use, funded by equity, debt, and JV/LP capital. Analog to “FCF”: AFFO (~$7.6/sh run-rate) is the distributable-cash proxy, but it is flattered by a low maintenance-capex assumption (~$169M Q4-25 recurring capex against a $1.9B annual depreciation base).

Capital Allocation & Management

How much FCF does the business generate, and how is it used? True post-development FCF is negative; the relevant figure is AFFO (~$7.6/sh), ~64% paid as dividend with ~36% retained for the build. Aggregate dividends ~$1.73B (FY2025).

Significant acquisitions recently? The large-M&A era (Interxion ~$8.4B 2020; Teraco ~$3.5B 2022) has given way to organic build + private-capital JVs/funds (Blackstone $7B JV 2023; $3.25B Hyperscale Fund 2026). No new transformative M&A.

Buying back shares? No — DLR is a net issuer (~6.5%/yr historical dilution; $870.6M issued in Q1-2026 alone). The private-capital pivot is partly intended to reduce reliance on dilutive ATM.

Issuing large amounts of new shares to insiders? Routine equity-comp grants (RSUs/PSUs); no unusual insider issuance. Insiders own <1%.

Compensation policy of directors/management? Well-aligned: pay anchored on Core FFO/share, Net Income from Operations, and Relative TSR, with no raw revenue/MW/empire metric. STI ~75% financial / 25% individual; LTI on Core FFO/share + constant-currency Core FFO + relative TSR.

Motivations of management? Interpretation: salaried/granted operators with metrics that reward per-share value and relative performance — better aligned than a growth-at-all-costs scorecard. But <1% ownership and zero open-market purchases mean management is not co-invested at scale; no conviction tell.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? No — a US-domiciled REIT (UPREIT; parent owns ~98.2% of the OP). Common shareholders receive a 1099-DIV (REIT dividends, partly ordinary income / capital-gain / return-of-capital), not a K-1.

Dividend policy? Common dividend frozen at $4.88/yr since 2022 (after ~17 years of growth); ~2.6% yield; ~64% AFFO payout. The freeze de-risks the dividend and funds the build with less dilution.

How profitable is the business? See above — development yield ~11.4% (value-creating but commodity-level); structurally lower-margin and lower-return than Equinix.

Is net income diverging from cash from operations? Yes, structurally — GAAP net income is depreciation-suppressed and disposition-gain-inflated; OCF (~$2.4B) is the better cash read, and even that is dwarfed by development capex. Use Core FFO/AFFO, not net income.

Risks & Downside

What factors would cause the stock to decline? A rate back-up (the −0.43 InterestRate factor loading is the largest single risk); evidence the capital cycle is compressing development yields; re-leasing spreads fading to flat; an AI-capex digestion/overbuild scare; multiple compression from the 96th-percentile valuation; or a hyperscaler concentration event (pause/renegotiation by the ~11.7% customer).

Risk of a catastrophic loss? Low. Investment-grade (BBB/Baa2), hard-asset, globally diversified, deleveraged (4.7x) REIT with a mostly investment-grade tenant base. The realistic downside is a de-rating + slower-growth grind, not insolvency.

Chance of a total loss? Very low — diversified hard-asset base, IG balance sheet, no single point of failure.

Recent News & Events

Has the business environment changed recently? Yes, materially and favorably (cyclically): the AI demand wave + power scarcity flipped wholesale pricing power to landlords (record-low 1.4% vacancy, +6.7% spreads), and DLR posted record bookings/backlog ($1.8B) including the largest lease in its history (a ~200 MW AI-inference deal). The counter-signal is the late-stage capital cycle (record PE inflow, Blackstone’s new BXDC, “$4.1T AI debt” headlines).

Significant acquisitions? None recently (the pivot is to JVs/funds, not M&A). The $3.25B Hyperscale Fund closed March 2026.

Change in accounting policies? None material; DLR is transitioning occupancy/leasing disclosure to power-based (IT-load) metrics starting 2026.

Recent changes — new markets, facilities, management? New markets/facilities: Barcelona (May 2026), Malaysia/Cyberjaya (June 2026), plus land/power additions (873-acre Atlanta parcel). Management is stable (CEO Power since end-2022; CFO Mercier since 2023).

APPENDIX B — Source Appendix

Primary sources first. All accessed 2026-06-20 unless otherwise dated.

Primary — SEC filings (EDGAR, CIK 0001297996)

1. FY2025 Form 10-K (filed 2026-02-13) — business (Item 1), properties (Item 2), MD&A revenue table, FFO/Core FFO reconciliation, balance sheet, cash-flow statement, debt/covenant policy, FX/natural-hedge note, customer-concentration disclosure (largest customer ~11.7% of ARR), Fund-contribution disposition gain (~$995.6M), goodwill, fee income $143.8M (+98%).
2. Q1-2026 Form 10-Q (filed 2026-05-01) — FFO reconciliation ($1.99 diluted/sh), dividend note ($4.88 annual rate; $425.0M common paid), equity issuance ($870.6M), debt.
3. FY2021–FY2024 Forms 10-K (filed 2022–2025) — multi-year FFO/share, revenue, M&A integration, dilution history.
4. 2026 DEF 14A (filed 2026-04-17) — compensation measures (Core FFO/share, Net Income from Operations, Relative TSR), STI/LTI structure, beneficial-ownership table (directors+officers 516,978 shares, <1% of 348,955,463).
5. EDGAR Form 3/4/5 corpus (parsed 2026-06-20) — insider transaction-code read: zero code-P open-market purchases across ~120 recent Form 4s.
6. 8-K material-event corpus (2024–2026) — quarterly results, debt/equity offerings, JV/fund announcements, guidance.

Primary — Company disclosures & transcripts

7. Q4-2025 earnings call transcript (2026-02-05) — FY2025 Core FFO $7.39 (+10%), same-capital cash NOI +4.5% cc, development pipeline ~$10B @ 11.9% yield, $3.225B fund equity, initial 2026 guide $7.90–8.00.
8. Q1-2026 earnings call transcript (2026-04-23) — Core FFO $2.04 (+15%/+11% cc), 2026 guide raised to $8.00–8.10, net debt/Adj EBITDA 4.7x, AFFO payout 64%, pipeline $16.5B/6 GW (1.2 GW UC, 61% pre-leased, 11.4% yield), backlog $1.8B/$1.0B share, renewal spreads +5% blended cash (+74% on small >1 MW volume), 2026 renewal guide +6.5–8.5%, the ~200 MW Charlotte AI-inference lease.
9. Blackstone / Digital Realty hyperscale JV release (2023-12-07, Blackstone / PRNewswire) — ~$7B JV; 80/20; Frankfurt/Paris/Northern Virginia; ~500 MW.
10. US Hyperscale Data Center Fund final-close release (2026-03-30, investor.digitalrealty.com / GlobeNewswire) — $3.25B LP equity commitments; DLR ~20% + manager.
11. DLR 2025 Impact Report (May 2026) — 93% global renewable coverage.

Secondary — industry data

12. CBRE — North America Data Center Trends, H2 2025 — primary-market vacancy ~1.4% (record low), net absorption ~2,498 MW, under-construction pipeline decline (5.9 GW vs 6.3 GW).
13. DatacenterDynamics — under-construction capacity decline; DLR Barcelona/Malaysia openings; Hyperscale Fund close.
14. CommercialSearch — “Who’s Funding the Data Center Boom” — ~$45.7B PE into data centers in 2025.
15. Data Center Frontier — CyrusOne/KKR ~$15B take-private; capital-cycle context.
16. US Senate Banking (Warren) — inquiry into PE data-center investment / utility-cost socialization.
17. CRE Daily / Ropes & Gray / CBRE — private-market data-center cap rates (~5.75–7% gross; ~6.5% CMBS stabilized); NoVa/Phoenix compression.

Quantitative & market data

18. Aggregated fundamentals (accessed 2026-06-20) — income statement, balance sheet, cash flow, profitability/credit ratios, per-share data, enterprise value (Q1-26 EV $81.72B; market cap $62.17B; EV/EBITDA TTM 27.99x; EV/sales 12.89x), valuation multiples (11-yr trend), company profile. Reconciled to filings. Note: GAAP ROE/ROIC and “free cash flow = operating cash flow” are not usable as-is (REIT artifacts).
19. Own-history valuation percentiles (2026-06-18) — P/B 96.7th, P/S 96.4th, composite 80th (P/E percentile ignored — REIT depreciation artifact). Price $188.15.
20. News flow (May–Jun 2026) — strongly positive/momentum skew; macro-caution items (JPMorgan “$4.1T AI debt”; Blackstone BXDC IPO).
21. Five-year daily price history — split/dividend-adjusted; 5-yr low $77.88 (2023-05-24), high $202.56 (2026-04-20), current $188.15.
22. Quantitative factor model (2026-06-18) — loadings (Sector Real Estate +1.06, InterestRate −0.43, Market +0.91, Momentum +0.095; beta 0.88, alpha +0.08), leaderboard (m6 +65%, m3 +43%, y3 +25%/yr, y5 +6.6%/yr ann.; lifetime max DD −56.8%), related names (all REIT/RE ETFs: SRVR, XLRE, ICF, IYR).

Peer cross-read (public)

23. Equinix, Inc. (NASDAQ: EQIX) public disclosures — the direct data-center REIT competitor: ~500,000 interconnections, largest customer ~3%, ~26–27% retail cash-on-cash, xScale capital-light model, an AI-as-indirect-bet profile. The primary relative-valuation and industry-structure benchmark.
24. American Tower (AMT), Crown Castle (CCI), SBA Communications (SBAC), Prologis (PLD) — tower and industrial REIT public disclosures for REIT/capital-cycle framing.