Digital Realty Trust DLR

Reference table

1. The quarter, on the measures a data-centre REIT is actually judged on

All figures filing-verified from the 8-K filed 2026-07-23, which carries the complete earnings release and financial supplement. The vendor payload contains no FFO field, no backlog field, no bookings field and no renewal-spread field.

Read the FFO ladder carefully, because the three measures tell different stories and only one is the operating number. Core FFO per share excluding net promote — $2.13, up 13.9% — is the measure management guides on and it is the one used throughout this dive. Headline Core FFO per share of $2.65 and FFO per share of $2.73 both include a large net promote contribution, which is the performance fee Digital Realty earns on its private-capital funds and which is lumpy by construction. A reader comparing $2.73 to $1.75 and concluding this business grew 56% would be reading a fee event, not the portfolio.

The leading indicators, and they are the reason the growth score is 7:

Three observations. The renewal spread is the most informative number in this file. A landlord re-signing existing tenants at 25.4% higher cash rents is telling you that replacement supply is scarce and that the in-place rent roll is materially below market. That is a pricing statement, not a demand statement, and it is the hardest thing for a competitor to erode.

The backlog is a revenue floor. $1.4 billion of annualised base rent at Digital Realty's share, signed but not commenced, against FY2025 revenue of $6,113 million, with a nine-month average lag. Most of the 2027 revenue increment is already contracted.

And the mix is improving. The 0-1 megawatt plus interconnection category — smaller deployments plus the cross-connect business — passed $100 million of quarterly bookings for the first time. That is the higher-margin, stickier, network-effect part of a data-centre portfolio, and it is the part that distinguishes a landlord from a wholesale developer. Management's own framing: "We signed more than $100 million of 0-1 MW plus Interconnection bookings for the first time, demonstrating the strength of our connectivity-rich portfolio and boosting near-term growth." Half-weighted as management framing; the $108 million is a disclosed number.

Guidance has been raised twice and the trajectory is worth showing:

2. The cost of the growth, and the number the payload gets wrong

This is the most important section in this dive and it exists because the payload's version of it is wrong by $3.2 billion.

Filed cash flow, from the 10-K ($ thousands):

The vendor's FY2025 capitalExpenditure is ZERO. The 10-K reports $3,181,179 thousand of "Improvements to investments in real estate" on the face of the cash-flow statement, and the FY2024 and FY2023 figures in the same vendor block match the filing to the dollar. This is the "capex is exactly zero" signature — one of the two outright-wrong classes in the documented defect taxonomy — and it is confined to a single year.

The consequence is that the vendor reports free cash flow of +$2,412 million where the filed arithmetic gives MINUS $769 million: a $3.18 billion error that changes the sign. freeCashFlowYieldTTM of 1.90% should be negative. priceToFreeCashFlowRatioTTM of 52.5x and evToFreeCashFlowTTM of 64.5x are both meaningless.

And the corrected picture is the analytical point, not just a data note. Digital Realty has not generated positive free cash flow in any of the four years the payload covers — MINUS $984M in FY2022, MINUS $1,891M in FY2023, MINUS $570M in FY2024 and MINUS $769M in FY2025 — while paying dividends of $1,451M, $1,521M, $1,633M and $1,728M. The gap between what the business generates and what it spends and distributes is funded by issuing equity and debt, and the evidence of that is in the share count:

Total assets rose from $49,410 million at 2025-12-31 to $54,518 million at 2026-06-30 — 10.3% in six months.

The June quarter provides a specific example of the mechanism. The 8-K filed 2026-07-01 discloses that in connection with the closing of the purchase of Blackstone's interests in the Digital Carver Dulles 9 and Digital Carver Brickyard joint ventures on 2026-06-30, Digital Realty filed Articles Supplementary "classifying and designating 12,310,249 shares of its authorized common stock, par value $0.01 per share... as shares of" a new class. Twelve million shares — 3.3% of the count — designated to fund one transaction.

None of this makes the growth illusory. A development REIT that builds assets yielding more than its cost of capital creates value even while free cash flow is negative, and the renewal spread of +25.4% cash is direct evidence that the assets are worth more than their book cost. But it does mean that per-share growth of 13.9% is being achieved against a 5%-a-year rising share count, and that the correct question about this company is not "how fast is it growing" but "what is the return on the capital it keeps raising." The filings in this archive do not answer that question directly, and we say so rather than estimating a development yield.

3. Balance sheet

Filing-verified at 2026-06-30 ($ thousands):

The credit trend is genuinely improving. Fixed charge coverage — adjusted EBITDA including the pro-rata share of unconsolidated entities, divided by fixed charges — went 4.7x, 4.6x, 4.5x, 4.9x, 5.2x across the five quarters to 2026-06-30. Financial debt is essentially flat over six months at roughly $18.6 billion while total assets grew 10.3%, which means the recent growth has been funded predominantly with equity rather than leverage.

The debt is 85% unsecured senior notes ($15.9bn of $18.6bn), with only $710 million drawn on the global revolver — substantial undrawn liquidity, though the facility size is not disclosed in the extracted text.

Retained earnings are MINUS $6,690,722 thousand at 2025-12-31, which for a REIT distributing more than its GAAP earnings is normal arithmetic rather than a warning.

4. Valuation — priced in or room?

The vendor cannot support the standard metric for this security type and we say so before quoting anything. There is no FFO field, no Core FFO field and no AFFO field in the payload. priceToEarningsRatioTTM of 89.3x is computed on GAAP earnings that charge $1,894.6 million of annual depreciation and amortisation against data centres whose renewal rents are rising 25.4%. Every valuation figure below uses Core FFO per share excluding net promote, sourced from the filings.

At $193.885:

On the est block: the epsAvg series is GAAP earnings per share, not FFO, and it is unusable for valuation. FY2026 $2.617, FY2027 $2.520 — a decline — and FY2028 $3.170. A series in which forward earnings per share falls while revenue rises 11% is the depreciation schedule, not the business. None of it is used here. Coverage is 11, 11, 12 and 10 analysts across FY2026 to FY2029.

est.ebitdaAvg and est.ebitAvg are fabricated and rejected — the twelfth consecutive file. In every one of the eight years present, ebitdaAvg is exactly 59.617% of revenueAvg and ebitAvg is exactly 27.406%.

Enterprise value — rebuilt, and the omission is the noncontrolling interests, which are large here. enterpriseValueTTM of $88,121.7M less market capitalisation is $16,388.9M, consistent with the 2026-06-30 financial debt of $18,635.3M less roughly $2,246M of cash — so the net-debt component uses the most recent quarter and correctly excludes operating leases. What it omits is $1,920.3 million of noncontrolling interests plus $1,567.3 million of redeemable noncontrolling interests — $3,487.6 million in total. Rebuilt: $91,609M, 4.0% above the vendor's figure. This is the fifth consecutive batch in which the NCI-omission check has fired, and DLR's is the second-largest instance in this batch after Simon Property Group's.

4a. What today's price assumes (the inversion)

At $193.885 — 23.7x the midpoint of management's Core FFO guidance — the price embeds:

• Core FFO per share excluding net promote of $8.15 to $8.20 in 2026. (Management guidance, raised twice; half-weighted.) The first half delivered $4.17 ($2.04 plus $2.13), so the second half needs about $4.02 — a slight decline, which the guidance range accommodates comfortably. This is the least demanding assumption in the price.
• Growth continuing near 11-14% a year thereafter. (Our derivation from the 13.9% quarterly rate and the $1.4bn backlog.) At $9.10 of 2027 Core FFO the stock is on 21.3x; at $10.10 for 2028, 19.2x.
• The market keeps paying 22-24x Core FFO. (Our number.) This is the most fragile assumption in the price. At 20x an extrapolated 2027 figure of $9.10 the stock is $182 — below today; at 26x it is $237.
• Renewal spreads stay strongly positive. (Our derivation; the filing supplies +25.4% cash.) The current spread reflects leases signed in 2020-2022 at pre-shortage rents rolling into a tight market. As those vintages exhaust, the spread compresses mechanically even if market rents keep rising.
• Equity issuance stays accretive. (Our number.) Shares rose 10.2% in two years while Core FFO per share rose 13.9% in the last one — so issuance has been accretive so far. If development yields compress, it stops being.

4b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: Core FFO per share growth (roughly +11%, from a 2026 midpoint of $8.175 toward an extrapolated $9.10) + multiple drift (modest COMPRESSION, from 23.7x to roughly 23.0x) + shareholder yield (+2.52% — dividend only; there is no buyback) ≈ +10-11%.

Our base assumes the multiple compresses slightly and we should say why. A 23.7x Core FFO multiple on a REIT is a growth multiple, and it is being paid for a business whose free cash flow has been negative for four years and whose share count rises about 5% a year. The offsetting argument — and it is genuine — is that a record backlog and 25.4% renewal spreads justify a growth multiple, and that the street agrees: 31 of 48 analysts are at buy and the lowest of 48 published targets is $197, above the price.

Note the composition: roughly three quarters of the expected return is Core FFO growth and one quarter is the dividend, with none from a re-rating. That is a defensible shape and it produces a modest number at this entry.

4c. Variant perception (where we differ, what would surprise)

• We are 4.2% below the street and the difference is the multiple, not the estimate. Consensus is $218.07, which is 26.7x the guidance midpoint; our base is 23.0x on an extrapolated 2027 figure. We use management's own guidance.
• We think the free-cash-flow position is materially under-discussed and the payload actively conceals it. Four consecutive years of negative free cash flow on the filed numbers, against a rising dividend, funded by a share count up 10.2% in two years. The vendor reports FY2025 free cash flow as POSITIVE $2.4 billion because it records the year's $3.18 billion of construction as zero. Anyone screening data-centre REITs on this field is looking at a different company. Watchable number: "Improvements to investments in real estate" in the FY2026 10-K against $3,181M.
• We think the renewal spread is the most important disclosure and its durability is the real question. +25.4% cash and +32.0% GAAP is exceptional. It is also, mechanically, a function of which lease vintages are rolling. Watchable number: the cash renewal spread in the 2026-10-22 release against +25.4%.
• Positive surprise that would force a re-rate: the backlog exceeding $1.6 billion at share; or a third guidance raise; or 0-1MW plus interconnection bookings above $120 million; or the Columbia Capital acquisition closing with disclosed fee economics that materially lift the $144 million of FY2025 fee income.
• Negative surprise that would break the thesis: cash renewal spreads falling below 15%; or the backlog declining; or a hyperscaler announcing a large self-build programme that displaces third-party leasing; or share issuance accelerating without a corresponding step in Core FFO per share.

Synthos fair values

All three anchors are multiples of Core FFO per share excluding net promote — the measure management guides on — sourced from the filings because the payload contains no FFO field.

• Bear ~$155 — 19.0x the 2026 guidance midpoint of $8.175, or 18.0x a stalled 2027 figure of $8.60. Cross-checks: a 3.15% dividend yield on the current $4.88; 4.8% above the 52-week low of $147.93; 21.3% below the street's low target of $197. The scenario: renewal spreads compress below 15% as the cheap lease vintages exhaust, hyperscaler self-build takes share, development yields fall, equity issuance turns dilutive, and a growth multiple becomes a REIT multiple. MINUS 20.1%.
• Base ~$209 — 23.0x an extrapolated 2027 Core FFO ex-promote of roughly $9.10, or 25.6x the 2026 guidance midpoint. Cross-checks: a 2.34% dividend yield; a rebuilt enterprise value of $99.0bn at that price. Sensitivity, stated openly: 20x the 2027 figure gives $182 — 6.1% BELOW spot — and 26x gives $237. The multiple is doing essentially all the work, because the near-year estimate is management's own guidance with a 0.6% range. The scenario: guidance is met, growth continues near 11%, the backlog converts on the nine-month lag, renewal spreads moderate but stay positive, and the multiple holds near current levels. +7.8%.
• Bull ~$262 — 28.8x an extrapolated 2027 figure of $9.10. Cross-checks: 32.0x the 2026 guidance midpoint; 28.5% above the 52-week high of $203.91; 20.1% above the street's $218.07 and 11.5% above the street high of $235. The scenario: the artificial-intelligence build-out extends the demand curve past 2028, renewal spreads stay above 20%, the 0-1MW plus interconnection business scales into a genuine network-effect franchise, the Columbia Capital transaction adds a fee-earning private-capital platform, and data-centre real estate re-rates as infrastructure rather than property. +35.1%.

Base is 7.8% above spot; asymmetry roughly 1.75:1 (20.1% down, 35.1% up), before a dividend of 2.52%. A base under 10% with a payoff ratio under two, on a business that has not generated positive free cash flow in four years, is a Watch — and the trigger is written as a price and a number in Section 8.

5. Knowledge base — four claims about the word "colocation", none about this company

Raw hits: 4. Case-sensitive entity matches: ZERO. Name-level claims: ZERO.

The sweep ran entity terms DLR, Digital Realty and Digital Realty Trust plus free text on digital realty, data center reit, data centre reit and colocation across all 52,021 distilled claims.

All four hits matched only on the word "colocation" and none is about a data-centre landlord:

> 2026-06-21 · bullish · conviction 55 · real_vision · speaker Mike Alfred · entities: CIFR, WULF — "Colocation-focused infra firms like Cypher and TeraWolf will be quite successful because they'll get really good at building out and leasing sites, even without wielding compute themselves." About two bitcoin miners converting to data-centre hosting.

> 2026-04-30 · bullish · conviction 78 · bankless · speaker Shuyao · entities: MEGA — "MegaETH is a token business, not just block space: USDM stablecoin yield and colocation/proximity-market revenue flow to buying back MEGA." About a blockchain layer-2 token.

> 2026-03-02 · neutral · conviction 60 · odd_lots · categories: market microstructure, regulation, latency — "Colocation is now equalized via coiled equal-length cables, ending the wild-west of drilling walls; speed advantage shifted to inter-datacenter links." About the placement of trading servers inside stock exchanges.

> 2024-08-22 · BEARISH · conviction 80 · no_priors · entities: AWS, Azure, Google Cloud, CoreWeave — "Today's AI cloud isn't real cloud — it's colocation/hosting: forced 3-year fixed reservations, no elasticity, paid up front, no options/futures market to hedge." About GPU cloud contract structure.

All four are discarded. There is no entity match on DLR, Digital Realty or Digital Realty Trust anywhere in the store.

The void is more striking here than on any other name in this batch. This knowledge base is unusually deep on artificial-intelligence infrastructure — it carries claims about compute demand, hyperscaler capital expenditure, GPU cloud economics and bitcoin miners converting to data centres — and it has said nothing whatever about the second-largest listed owner of the buildings.

One adjacent observation, recorded without scoring it. The store's single claim on Simon Property Group, from compound_and_friends on 2026-06-09, notes in passing that "Prologis is turning into a data-center landlord". So the store treats becoming a data-centre landlord as a positive development worth mentioning about an industrial REIT, while holding no view at all on the company that already is one. That is a coverage gap rather than a signal, and it is recorded as such.

This is the fifteenth void this programme has found and reported honestly. The verdict here rests entirely on the filings. Breadth 0, claim count 0, net conviction empty.

6. Data integrity — what we rejected and why

Eight findings, one of which is the largest single-field error in this batch.

1. capitalExpenditure for FY2025 is ZERO against a filed $3,181,179 thousand — and it flips free cash flow from positive to negative. The 10-K reports "Improvements to investments in real estate" of $(3,181,179) thousand for FY2025 on the face of the consolidated statement of cash flows, against $(2,831,740) in FY2024 and $(3,525,598) in FY2023. The vendor's capitalExpenditure for FY2024 and FY2023 matches those exactly; FY2025 reads 0, and investmentsInPropertyPlantAndEquipment reads 0 for all three years. Consequently freeCashFlow for FY2025 reads +$2,412,136 thousand — identical to operating cash flow — where the filed arithmetic gives MINUS $769,043 thousand. A $3.18 billion error that changes the sign of the number most REIT screens rank on. freeCashFlowYieldTTM of 1.90%, priceToFreeCashFlowRatioTTM of 52.5x and evToFreeCashFlowTTM of 64.5x are all rejected. Note the internal contradiction that would have caught it: capexToDepreciationTTM reads 0.802 and capexToRevenueTTM reads 0.234, both of which require a large non-zero capex — so the trailing-twelve-month ratios and the annual cash-flow block disagree with each other inside the same payload.

2. earn_cal switches measurement basis mid-series and produces a spurious 150% beat. The two 2025 entries report epsActual of $1.87 and $1.89, which are Core FFO per share — the $1.87 matches the Q2 2025 figure in the earnings release exactly. The three 2026 entries report $0.24, $0.46 and $1.21, which are GAAP diluted EPS and match inc_q exactly. The 2026-07-23 row therefore shows an actual of $1.21 against an estimate of $0.4829 — a 150% "beat" that is entirely a basis mismatch. The forward estimate of $0.575 for 2026-10-22 appears to be GAAP and is NOT comparable to the $1.89 the same field shows for September 2025. This is the identical defect found on Simon Property Group in this batch — two REITs, same failure — and any screen tracking earnings surprises on either is reading a basis change as a result.

3. There is no FFO, Core FFO or AFFO field, so every earnings-based metric is a depreciation artefact. priceToEarningsRatioTTM of 89.3x, netIncomePerShareTTM of $2.256 and earningsYieldTTM of 1.16% are all computed on GAAP earnings after $1,894.6 million of annual depreciation and amortisation. All valuation in this dive uses Core FFO per share excluding net promote from the filings.

4. bal_a.totalDebt exceeds filed financial debt by $5.78 billion — 31.4% — and the residual is not attributable to any single line. The vendor reports FY2025 shortTermDebt $1,147,428 plus longTermDebt $20,003,886 plus capitalLeaseObligations $3,030,914, totalling $24,182,228 thousand. The filed balance sheet at 2025-12-31 shows global revolving credit facilities $899,090, unsecured term loans $439,536, unsecured senior notes $16,194,441 and secured and other debt $869,068 — financial debt of $18,402,135 thousand. Adding the entire operating lease liability of $1,253,217 gives $19,655,352, still $4,526,876 thousand below the vendor's figure. So the vendor's capitalLeaseObligations of $3,030,914 is 2.4x the disclosed operating lease liability, and roughly $4.5 billion of the total is unattributable to any line in the filed balance sheet. This is the largest and least explicable debt overstatement in this batch, and netDebtToEBITDATTM of 4.81x and debtToEquityRatioTTM of 0.665 are rejected accordingly.

5. enterpriseValueTTM omits $3,487.6 million of noncontrolling interests. The residual of $16,388.9M over market capitalisation is consistent with the 2026-06-30 financial debt of $18,635.3M less roughly $2,246M of cash — so the net-debt side is current and correctly excludes leases, unlike totalDebt. It omits $1,920.3 million of noncontrolling interests and $1,567.3 million of redeemable noncontrolling interests. Rebuilt: $91,609M, 4.0% above the vendor's $88,122M — the second-largest NCI omission in this batch.

6. inc_q.grossProfit is corrupt across multiple quarters, including two negative values. The vendor reports gross profit of $162.1M (Q2 2026), MINUS $62.1M (Q1 2026), MINUS $24.2M (Q4 2025), $867.0M (Q3 2025) and $831.6M (Q2 2025). A negative gross profit for a REIT reporting a 49.8% EBITDA margin is not possible, and the series swings by a factor of fourteen across adjacent quarters with revenue moving less than 20%. Every other field in those rows — revenue, operating income, net income, diluted EPS, interest expense — is consistent with the filings. The partial-row class: one field wrong across several quarters while the rest of each row is sound.

7. seg_geo contains single-letter labels in two years. The FY2019 entry reads "G": $288,200,000, "Non Us": $627,400,000 and "U": $2,600,000,000; FY2020 reads "Non Us": $1,300,000,000 and "U": $2,600,000,000. "U" and "G" are truncated XBRL element names, not geographies. The recent years are fine — FY2025 United States $3,167.7M plus Non-US $2,945.0M equals $6,112.7M, which is consolidated revenue to the dollar — so the block is usable from FY2021 forward and unusable before it. This is the ninth consecutive batch in which seg_geo has failed on at least one name, and the fifth failure in this batch.

8. est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature, and est.epsAvg is GAAP rather than FFO. Exactly 59.617% and 27.406% of revenueAvg in all eight years. And the epsAvg series has FY2027 ($2.520) BELOW FY2026 ($2.617) on revenue 11.1% higher — which is the depreciation schedule, not a forecast, and confirms the series is GAAP. None of the est block is used in this valuation.

VERIFIED CLEAN — recorded because a clean check is a finding.

FY2024 and FY2023 capital expenditure are exactly correct at $2,831,740 and $3,525,598 thousand, matching the 10-K to the dollar — which is what makes the FY2025 zero a single-year corruption rather than a systematic one, and what allowed it to be caught.

seg_geo reconciles exactly for FY2025 and FY2024, and seg_prod reconciles exactly: FY2025 rental and other services $5,968,918 plus fee income and other $143,774 equals $6,112,692 thousand, which is total revenue to the dollar.

The enterpriseValueTTM net-debt component is current and lease-free, using the 2026-06-30 balance sheet — a better construction than the totalDebt field in the same payload, which is worth noting because the two fields in the same file disagree with each other by roughly $5.5 billion.

Non-equity tripwire — checked and passed. DLR is common stock, $0.01 par value, NYSE-listed. Price of $193.885 is not par-like; beta is 1.053; volume was 1.65M shares (~$321M of turnover); the 52-week band of $147.93 to $203.91 is a 38% range. This is common equity. Note the 12,310,249 shares of a newly designated class created by Articles Supplementary on 2026-06-30 in connection with the Blackstone acquisition, and the $1,567.3 million of redeemable noncontrolling interests in the operating partnership; neither is this security.

Vendor composite rating — noted, and it is the lowest in this batch. C / 2 overall, with 1 of 5 on discounted cash flow, price-to-earnings and price-to-book. All three sub-scores are computed on GAAP earnings and on the free-cash-flow figure rejected in finding 1, so the rating is scoring a REIT as though it were an operating company. It carries no weight — but it is worth observing that the rating and the analyst consensus (31 of 48 at buy) could not disagree more completely, and both are looking at the same GAAP numbers.

7. Technicals and insiders

• Price $193.885. MINUS 4.9% from the 52-week high of $203.91; +31.1% above the low of $147.93. Position within the annual range: 82nd percentile.
• Above both moving averages: +4.7% above a 50-day average of $185.20; +10.0% above a 200-day average of $176.32, with the 50-day above the 200-day. A clean uptrend.
• RSI 66.0 — firm, not overbought. MACD +2.58, positive.
• Current distance from the six-year peak: MINUS 4.9%, identical to the 252-day figure — the six-year peak falls inside the last twelve months and the coincidence is correct.
• Relative performance is mixed and the shape matters: 3-month MINUS 2.3% against SPY +7.6%; 6-month +17.4% against SPY +11.1%; 12-month +12.6% against SPY +24.3% — an 12-point deficit over a year but a 6-point outperformance over six months. The stock rallied hard from February through June as the guidance was raised twice, and has gone sideways since.
• Sentiment: 1 strong buy, 30 buy, 16 hold and 1 sell across 48 analysts — the only name in this batch with a sell rating other than Carvana and Colgate. Consensus target $218.07 (+12.5%), median $220.50, high $235 (+21.2%), low $197 — 1.6% ABOVE spot.

Insiders — nothing usable. Eight rows: seven A-Award director grants (153 to 1,815 shares, at $0, on 2026-05-29 and 2026-06-30 — the standard annual retainer) and one F-InKind tax withholding of 53 shares at $176.32 by the chief accounting officer. There is not one open-market purchase or sale, and no transaction by the chief executive or chief financial officer. The file is uninformative rather than corrupt, and it is reported as such.

8. Verdict, kill-criteria and flip conditions

Watch.

The business deserves better than a Watch and the price does not.

What is genuinely excellent, and all of it is filing-verified. Core FFO per share excluding net promote of $2.13, up 13.9%. A record backlog of $1.9 billion of annualised base rent at 100% share, $1.4 billion at Digital Realty's share, against $6.1 billion of annual revenue and a nine-month signing-to-commencement lag — so most of the 2027 revenue increment is already contracted. Renewal leases signed at rental rates 25.4% higher on a cash basis and 32.0% higher on a GAAP basis, which is the clearest evidence of landlord pricing power anywhere in this batch. Quarterly bookings of $208 million at share, including more than $100 million from the 0-1 megawatt plus interconnection category for the first time. Full-year guidance raised twice. Fixed charge coverage improved from 4.7x to 5.2x.

What holds it at Watch, in three specifics.

First, the price. 23.7 times the midpoint of management's own Core FFO guidance is a growth multiple, and our base of $209 — 23.0x an extrapolated 2027 figure — is only 7.8% above spot with 1.75:1 asymmetry. At 20x the stock is $182, 6.1% below today.

Second, and this is the substantive objection: the growth consumes more cash than it produces. On the filed numbers, free cash flow was MINUS $769 million in FY2025, MINUS $570 million in FY2024, MINUS $1,891 million in FY2023 and MINUS $984 million in FY2022 — four consecutive negative years — while dividends rose from $1,451 million to $1,728 million. The difference is funded with equity: weighted diluted shares rose 10.2% in two years, and the June quarter designated 12,310,249 new shares for a single acquisition. That is a legitimate way to build a development business and it is not a legitimate reason to pay a premium multiple.

Third, the renewal spread that makes the current numbers look so good is partly a vintage effect. +25.4% cash reflects leases signed in a looser market rolling into a tight one. As those vintages exhaust, the spread compresses even if market rents keep rising, and nothing in this archive says how much of the rent roll is still below market.

Why not Hold. Because there is a specific, dated, watchable event — the 2026-10-22 print — with three numbers that would change the answer, and because a business compounding per-share cash earnings at 13.9% with a record contracted backlog belongs on a watchlist rather than in the dead file.

Why not Buy. Because 7.8% of upside on 1.75:1 asymmetry, at 23.7x, on a business with four consecutive years of negative free cash flow, is not a position.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• A price below roughly $170, which is 20.8x the guidance midpoint and restores a payoff ratio above 2.5:1 — and is 3.6% above the 200-day average.
• A third guidance raise, taking Core FFO per share excluding net promote above $8.25.
• The backlog exceeding $1.6 billion at Digital Realty's share, from a record $1.4 billion.
• Cash renewal spreads holding above 20% for two more quarters, which would demonstrate the pricing power is market-driven rather than vintage-driven.
• Free cash flow turning positive, or capital expenditure guidance falling below $3 billion with growth maintained.
• 0-1MW plus interconnection bookings above $120 million in a quarter.

Pre-registered KILL criteria — what would take this to Hold or Avoid:

• Cash renewal spreads below 15%.
• The backlog declining for two consecutive quarters from $1.4 billion at share.
• Weighted diluted shares rising more than 6% in a year without a corresponding step in Core FFO per share.
• Fixed charge coverage falling below 4.5x, from 5.2x.
• Any impairment of the $9,592 million of goodwill, which is 17.6% of total assets.
• A hyperscaler announcing a self-build programme that materially displaces third-party leasing.

Where DLR fits in the Synthos Framework Portfolio. The real-estate sleeve, on the watchlist at zero weight, with a 2% starter contingent on a price below $170 or a third guidance raise. Sizing note: this is the AI-infrastructure exposure in this batch that is expressed through a balance sheet rather than a semiconductor, which means it carries interest-rate and equity-issuance risk that the underlying demand curve does not. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $193.885, with the fair-value anchors, kill criteria and upgrade conditions all gradeable.

Single biggest risk: that growth this good requires more capital than the business generates. Digital Realty spent $3,181 million on improvements to investments in real estate in FY2025 against $2,412 million of operating cash flow, and paid $1,728 million of dividends on top. It has not generated positive free cash flow in any of the four years the payload covers. The gap is funded by equity and debt, and the equity side shows: weighted diluted shares up 10.2% in two years, with 12,310,249 shares designated in June alone to buy Blackstone's interests in two Northern Virginia joint ventures. This is how development REITs grow, and while the assets being built yield more than the cost of capital it creates value — the 25.4% cash renewal spread is direct evidence that the existing portfolio is worth more than its book cost. The risk is not that the model is wrong; it is that the model requires continuous access to capital markets at attractive prices, and that a 23.7x multiple already assumes it. A data-centre REIT is a levered bet on the artificial-intelligence build-out with a rate-sensitive funding structure attached, and the honest place to own it is at a lower multiple than this one.

Provenance & disclosures

• Traceability: ZERO knowledge-base claims name Digital Realty out of 52,021 distilled claims (raw hits 4, case-sensitive entity matches 0, name-level claims 0, discarded 4; breadth 0, claim count 0, net conviction empty). All four hits matched only on the free-text term "colocation" and none concerns a data-centre landlord: real_vision 2026-06-21 (bullish 55, speaker Mike Alfred, entities CIFR and WULF — about two bitcoin miners converting to hosting), bankless 2026-04-30 (bullish 78, entity MEGA — about a blockchain layer-2 token whose revenue includes "colocation/proximity-market revenue"), odd_lots 2026-03-02 (neutral 60 — about the placement of trading servers inside stock exchanges and latency arbitrage), and no_priors 2024-08-22 (bearish 80, entities AWS, Azure, Google Cloud, CoreWeave — about GPU cloud contract structure). All four discarded. The void is the most striking in this batch: this store is unusually deep on artificial-intelligence infrastructure and carries no view whatever on the second-largest listed owner of the buildings. One adjacent observation is recorded without scoring: the store's single claim on Simon Property Group notes in passing that "Prologis is turning into a data-center landlord" — so becoming a data-centre landlord registers as a positive development about an industrial REIT while the company that already is one is absent. This is the fifteenth void this programme has found and reported honestly, and the verdict rests entirely on the filings. All quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — balance sheet, debt composition and the June 2026 quarter through 2026-06-30, filing-verified from the 10-Q filed 2026-07-31 and the 8-K filed 2026-07-23 carrying the complete earnings release and financial supplement; annual figures for FY2025, FY2024 and FY2023 from the 10-K filed 2026-02-13, including the cash-flow statement's "Improvements to investments in real estate" line · corporate actions from the 8-Ks of 2026-06-29 and 2026-07-01 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873599 = 2026-08-04T19:59:59Z ($193.885, +1.39%; 50-DMA $185.20; 200-DMA $176.32; RSI 66.0; MACD +2.58; beta 1.053) · knowledge-base claims 2026-08-04. Digital Realty's fiscal year is the calendar year. All figures come from the Synthos vendor data file for DLR or from the SEC filings in the DLR archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2026-02-13 (fiscal 2025 — the source of the $3,181,179 thousand of improvements to investments in real estate that the payload records as zero); 10-Q filed 2026-05-01 (March 2026 quarter); 10-Q filed 2026-07-31 (June 2026 quarter — the source of the balance sheet and the debt composition); 8-K filed 2026-06-29, disclosing the purchase of Blackstone's interests in the Digital Carver Dulles 9 and Digital Carver Brickyard joint ventures; 8-K filed 2026-07-01, disclosing that on 2026-06-30 Digital Realty filed Articles Supplementary classifying and designating 12,310,249 shares of common stock in connection with the closing of that acquisition; 8-K filed 2026-07-23 (177,699 characters), carrying the complete second-quarter earnings release and financial supplement — the source of the Core FFO ladder, the record $1.9bn / $1.4bn backlog, the $208M of at-share bookings including $108M from 0-1MW plus interconnection, the +25.4% cash and +32.0% GAAP renewal spreads, the raised guidance and the 5.2x fixed charge coverage ratio. All carry preserved [TABLE] statement data.
• Where the filings contradicted or corrected the vendor (detailed in Section 6): capitalExpenditure for FY2025 of ZERO against the 10-K's $3,181,179 thousand — the single largest field error in this batch — flipping freeCashFlow from the filed MINUS $769,043 thousand to a reported +$2,412,136 thousand, a $3.18 billion error that changes the sign, and invalidating freeCashFlowYieldTTM (1.90%), priceToFreeCashFlowRatioTTM (52.5x) and evToFreeCashFlowTTM (64.5x) — with the internal contradiction that capexToDepreciationTTM of 0.802 and capexToRevenueTTM of 0.234 both require a large non-zero capex, so two parts of the same payload disagree; earn_cal, which reports Core FFO per share for the 2025 quarters and GAAP diluted EPS for the 2026 quarters, producing a spurious 150% "beat" on 2026-07-23 ($1.21 actual against a $0.4829 estimate) — the identical defect found on Simon Property Group in this same batch; the complete absence of any FFO, Core FFO or AFFO field, making priceToEarningsRatioTTM of 89.3x a depreciation artefact; bal_a.totalDebt of $24,182,228 thousand against filed financial debt of $18,402,135 thousand — a 31.4% overstatement of which roughly $4.5 billion is unattributable even after adding the entire $1,253,217 thousand operating lease liability, with capitalLeaseObligations of $3,030,914 thousand being 2.4x the disclosed operating lease balance; enterpriseValueTTM, which omits $1,920.3 million of noncontrolling interests and $1,567.3 million of redeemable noncontrolling interests, leaving it 4.0% below a rebuilt $91,609M; inc_q.grossProfit, corrupt across multiple quarters including MINUS $62.1M and MINUS $24.2M readings for a REIT reporting a 49.8% EBITDA margin, while every other field in those rows is sound; seg_geo, whose FY2019 and FY2020 entries contain the truncated XBRL labels "U" and "G" in place of geographies; and est, whose ebitdaAvg and ebitAvg are fabricated at exactly 59.617% and 27.406% of revenue in all eight years and whose epsAvg series is GAAP rather than FFO — shown by FY2027 ($2.520) falling below FY2026 ($2.617) on 11.1% higher revenue. Where vendor and filing AGREED — recorded, because a clean check is a finding: FY2024 and FY2023 capital expenditure are exactly correct at $2,831,740 and $3,525,598 thousand, which is what makes the FY2025 zero a single-year corruption and what allowed it to be caught; seg_prod reconciles exactly, with FY2025 rental and other services of $5,968,918 plus fee income and other of $143,774 equalling total revenue of $6,112,692 thousand to the dollar; seg_geo reconciles exactly for FY2025, with United States $3,167.7M plus Non-US $2,945.0M equalling $6,112.7M; and the enterpriseValueTTM net-debt component uses the 2026-06-30 balance sheet and correctly excludes operating leases — a better construction than the totalDebt field in the same file, with which it disagrees by roughly $5.5 billion.
• REIT-metric note: all valuation in this dive uses Core FFO per share excluding net promote, sourced from the filings, because the payload contains no FFO field. Second-quarter 2026: $2.13, against $1.87 a year earlier. Full-year 2026 guidance: $8.15 to $8.20, raised twice from an initial $7.90-$8.00. Three FFO measures are reported in the release and only one is the operating number: headline Core FFO per share of $2.65 and NAREIT-defined FFO per share of $2.73 both include a large net promote fee contribution that is lumpy by construction, and a reader comparing $2.73 to the prior year's $1.75 would be reading a fee event rather than the portfolio. AFFO is NOT computed anywhere in this dive: the filings give no split between maintenance and development capital within the $3,181,179 thousand of improvements to investments in real estate, and that split is required for AFFO. We state the absence rather than estimating around it.
• Estimate coverage and basis: the est block's epsAvg series is GAAP earnings per share and is used nowhere in this valuation, on 11, 11, 12 and 10 analysts across FY2026 to FY2029. The fair values are struck on management's own Core FFO guidance, whose $0.05 range is 0.6% — so the multiple assumption is doing essentially all the work in the fair-value spread, and that is stated rather than concealed.
• Peer note: the vendor peer set — American Tower, CBRE, Crown Castle, Equinix, Realty Income, Public Storage, Simon Property Group — contains exactly one direct comparable (Equinix), plus two tower REITs, a property services firm, a net-lease REIT, a storage REIT and a mall REIT. It is the shortest peer set in this batch at seven names. No peer-multiple comparison is drawn, because a like-for-like Core FFO multiple would require the same FFO data for each peer and this archive contains it only for Digital Realty.
• Fair-value caveat: the $155 / $209 / $262 anchors are multiples of Core FFO per share excluding net promote — 19.0x the 2026 guidance midpoint of $8.175 for the bear, 23.0x an extrapolated 2027 figure of roughly $9.10 for the base, and 28.8x the same for the bull — each cross-checked against the resulting dividend yield and against a rebuilt enterprise value. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 20x the 2027 figure gives $182, which is 6.1% BELOW spot, and 26x gives $237. The base assumes modest multiple compression from 23.7x toward 23.0x, on the grounds that a growth multiple is being paid for a business with four consecutive years of negative free cash flow and a share count rising about 5% a year.
• Guidance and management-framing note: the Core FFO per share range of $8.15-$8.20, the constant-currency range of $8.10-$8.15, the NAREIT FFO range of $8.40-$8.45, and the characterisation of the 0-1MW plus interconnection bookings as "demonstrating the strength of our connectivity-rich portfolio" are management statements from the July 2026 earnings release, half-weighted under the standing house rule. The reviewed and disclosed figures — the FFO reconciliation, the bookings and backlog amounts, the renewal rate changes, the balance sheet, the fixed charge coverage ratio and the cash-flow statement — are not.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.