American Tower AMT

> ## REIT BANNER — THE VENDOR CANNOT VALUE THIS COMPANY

> The payload contains NO funds-from-operations field of any kind. est.epsAvg and the earnings calendar both carry GAAP earnings per share, and the reported 24.1x price-to-earnings is a depreciation artefact on a company that charged $1,032.6 million of depreciation, amortisation and accretion in six months against $2,887.4 million of operating cash flow.

> The 10-Q defines and discusses both Nareit FFO and AFFO. The per-share figures did not survive the extraction window searched, and this dive does not invent one.

> Valuation here is therefore built on Adjusted EBITDA — $1,808.2 million in the quarter, filing-verified — on the recomputed free-cash-flow yield, and on the dividend. Every price-to-earnings figure in this document is labelled as a GAAP artefact and is used for nothing.

Reference table

1. What the business is, and the segment realignment

From the 10-Q filed 2026-07-28:

> "Our primary business is the leasing of space on communications sites... We also hold a portfolio of highly interconnected data center facilities and related assets in the United States. We refer to the business encompassing the above as our property operations, which accounted for 98% and 98% of our total revenues for the three and six months ended June 30, 2026, respectively, and includes our U.S. & Canada property, Africa & Asia-Pacific ("APAC") property, Europe property and Latin America property segments and Data Centers segment."

Note the segment structure: Africa and Asia-Pacific have been COMBINED into a single "APAC" segment. The vendor's seg_geo block reflects the consequence without labelling it — FY2025 carries a line reading "Property, Africa" at $1,422.9 million with NO Asia-Pacific line at all, where FY2024 carried Africa at $193.7 million and Asia-Pacific at $1,208.0 million and FY2023 carried Africa at $1,225.6 million and Asia-Pacific at $1,150.8 million. A reader looking only at the payload would conclude that Asia-Pacific revenue vanished and African revenue grew sevenfold. Neither happened; the segments were merged. Chief executive Steven O. Vondran; approximately 4,866 employees.

FY2025 revenue by geography, from the vendor block with the aggregate corrected in Section 4:

Two readings. The core is slow: the United States is 52.5% of revenue and grew 2.7%. The fast line is Data Centers at 13.9% and Europe at 12.3%, and Data Centers is now 9.9% of the company — a tenth of American Tower is an interconnection business rather than a tower business, and it is growing five times as fast as the towers.

Two disclosures worth recording from the 10-Q, neither in any vendor field:

The DISH churn, quantified. "DISH represented approximately 2% and 4% of our total annual property revenue and total annual U.S. & Canada property revenue, respectively, for 2025. DISH is reflected in churn for the six months ended June 30, 2026. During the six months ended June 30, 2026, we recorded impairment charges related to DISH of $17.5 million." A named tenant representing 4% of the largest segment's revenue moving into churn is a material, dated headwind on a business whose United States revenue is growing 2.7%.

And a disposal in the comparative period: the tax discussion references "taxes related to the sale of South Africa Fiber during the six months ended June 30, 2025" — one of a series of international simplifications alongside the earlier India exit.

2. The quarter

From the 10-Q ($M):

Three observations.

The 133% net income increase is almost entirely a swing in "other (income) expense" and it is not operating. That line went from a $373.9 million expense in the June 2025 quarter to a $53.8 million income in June 2026 — a $427.7 million swing — and across the half from $712.1 million of expense to $144.0 million of income, an $856.1 million swing. For American Tower this line is dominated by foreign currency translation on international debt and by unrealised marks; it is volatile and it is not earnings. The vendor's inc_q reports GAAP diluted EPS of $1.86 against $0.78 with no indication of that.

The operating business grew 3.0%. Adjusted EBITDA of $1,808.2 million on revenue up 4.7% means margin contracted 93 basis points to 65.8% — costs grew faster than the contractual escalators, which for a business of this type is the number that matters.

And the "beat" is on the wrong metric. The earnings calendar records $1.86 actual against a $1.57 estimate — an 18.5% beat. That is GAAP diluted earnings per share, on a REIT, in a quarter whose net income was inflated by a $427.7 million non-operating swing. The industry values this company on AFFO. The payload has no AFFO field, so the "beat" measures something nobody underwrites. The four-quarter record is uniformly positive on the same wrong metric: +18.5%, +15.0% ($1.84 against $1.60), +18.2% ($1.75 against $1.48) and +6.1% ($2.78 against $2.62).

3. Cash flow — the capex check, verified clean

The 10-Q's cash-flow statement gives the filed half-year figures:

The trailing recomputation:

Our TTM capex ÷ TTM operating cash flow is 1,815.1 ÷ 5,774.9 = 0.3143. The vendor's capexToOperatingCashFlowTTM is 0.31431. Our yield of 4.852% reproduces freeCashFlowYieldTTM of 4.85207% to three decimal places. capexToDepreciationTTM of 0.876 is plausible for a REIT whose depreciation is dominated by long-lived tower and intangible assets. Both the diagnostic and the datum were tested separately against the filing, per the MPC lesson; both pass. No repair applied. This is the fifth clean capex verification in this batch.

One REIT-specific caveat, stated because it matters for the dividend. American Tower's $1,815 million of trailing capital expenditure is dominated by DISCRETIONARY construction and augmentation, not maintenance. The filing's own line is "Payments for purchase of property and equipment and construction activities". True maintenance capital intensity for a tower portfolio is a small fraction of that, which is why AFFO — had the payload carried it — would be materially higher than the $3,960 million of free cash flow computed here. The $3,960 million figure is therefore a CONSERVATIVE measure of distributable cash, and the dividend's coverage is better than it appears: $3,157.2 million paid in FY2025 against $3,960 million of trailing free cash flow after ALL capital spending.

4. The two fields that are wrong, and both were predicted

totalDebt overstates borrowings by 20.8%, and the reconciliation is exact

From the 10-Q balance sheet at 2025-12-31: current portion of long-term obligations $3,387.8M plus long-term obligations $33,832.5M = $37,220.3M of debt. Separately: current portion of operating lease liability $584.9M plus operating lease liability $7,158.7M = $7,743.6M.

The vendor's totalDebt for the same date reads $44,963.9M.

$37,220.3M + $7,743.6M = $44,963.9M — to the dollar.

The vendor has added the entire operating lease liability to borrowings. The filing wins. This is the lease-inclusion defect class at its largest in this batch — 20.8%, against Sherwin-Williams' 19.1%, HCA's 8.0% (double-counted), Cummins' 7.4%, Phillips 66's 16.1% and Royal Caribbean's 3.3%. For a tower REIT the error is particularly misleading, because ground leases under towers are a genuine and permanent operating cost, not a financing decision.

Consequential correction, and it runs in the company's favour: netDebtToEBITDATTM reads 6.15x on the inflated base. On the filing's debt at 30 June 2026 — current $5,226.5M plus long-term $31,963.1M = $37,189.6M, less cash of $1,762.5M — net debt is approximately $35,427M, and against trailing Adjusted EBITDA of roughly $7,150M that is approximately 5.0x. Every leverage figure in this dive uses the corrected basis.

Enterprise value omits $6,702.8 million of noncontrolling interests — the largest here

enterpriseValueTTM reads $124,828M against a market capitalisation of $81,611M, implying $43,217M of net claims against FY2025 net debt of $43,489M — a $272 million residual. The $6,702.8 million of noncontrolling interests is entirely excluded.

Adding it gives approximately $131,802M, 5.6% higher. AMT was named in advance as a likely instance of the KKR / MPC / FCX class and it is confirmed — and at $6.70 billion this is the largest ABSOLUTE noncontrolling-interest omission in this batch, ahead of Phillips 66's $1,148M, HCA's $3,433M and Cummins' $1,059M. The June 2026 balance sheet shows the same order of magnitude: total equity of $10,252.5M against American Tower common equity of roughly $3,720M, so noncontrolling interests are approximately $6,532M — 63.7% of total equity, reflecting minority partners in the international property businesses.

Rebuilding enterprise value on the June balance sheet with corrected debt: market capitalisation $81,611M + debt $37,190M − cash $1,763M + noncontrolling interests $6,532M = approximately $123,570M. Against trailing Adjusted EBITDA of roughly $7,150M that is approximately 17.3x, and that is the multiple used throughout this dive.

And the geography block over-counts by exactly its services line

seg_geo FY2025 lists five "Property," geographies plus a "Services Segment" line of $339.6M, summing to $10,984.2M against inc_a revenue of $10,645.1M — 103.2%. The five geographic lines alone sum to $10,644.6M, which is total revenue to within 0.005% — so the geographies already INCLUDE services revenue, and the separate "Services Segment" line double-counts it exactly. The CDNS / CSX over-count class, with the excess equal to the added line to the dollar. The five-line version is used throughout this dive.

A second, uglier instance in FY2024: the block reports "Property, Africa: $193,700,000" — against $1,225.6M in FY2023 and $1,422.9M in FY2025 — and $193.7M is exactly the FY2024 "Services Revenue" figure from seg_prod. The FY2024 African revenue line has been overwritten with the services figure. FY2024 is rejected.

seg_prod itself is clean. FY2025: Property $10,305.0M + Services $339.6M = $10,644.6M against inc_a revenue of $10,645.1M — a 0.005% difference. No over-count and no omission.

5. Valuation — priced in or room?

At $175.145 (market cap $81.611B, 465.96M shares):

The GAAP earnings row is shown only to be labelled. A REIT that charged $1,032.6 million of depreciation, amortisation and accretion in six months against $2,887.4 million of operating cash flow does not earn its economics in GAAP net income, and a 24.1x price-to-earnings ratio on that basis carries no information. The est.epsAvg and earn_cal fields both track GAAP EPS — the FY2025 consensus of $5.122 against a reported GAAP diluted $5.39, and the June-quarter calendar actual of $1.86 matching the 10-Q's GAAP figure exactly — so the payload contains no FFO or AFFO anywhere. This dive values on Adjusted EBITDA, the recomputed free-cash-flow yield and the dividend, and states that the standard REIT metric could not be sourced.

est.ebitdaAvg and est.ebitAvg are REJECTED for a fixed-ratio signature, with an unusual note. In every year from FY2026 to FY2030, ebitdaAvg is exactly 64.16% of revenueAvg and ebitAvg is a correspondingly fixed proportion. The unusual part is that 64.16% is very close to the actual trailing EBITDA margin of 64.21% and the Adjusted EBITDA margin of 65.8% — so unlike DASH's 170%-overstated instance, this fabrication happens to land near the truth. It is still a fixed ratio applied mechanically across five forward years and is rejected on principle: a row that would be right by accident is not a row to value on. The FY2027 Adjusted EBITDA figure used in this dive is built from the filed first-half run rate instead, and is labelled as our estimate.

Estimate coverage on the GAAP EPS row is thin and irrelevant — 7 analysts on FY2026 and 10 on FY2027 — while revenue coverage is good at 16 and 14. FY2029 and FY2030 EPS rest on 10 and 5 analysts and are excluded from every conclusion.

5a. What today's price assumes (the inversion)

At $175.145 — approximately 17.3x corrected enterprise value to Adjusted EBITDA, a 3.99% dividend yield and a 4.85% free-cash-flow yield — the price embeds:

• Revenue compounds 3.3-3.4% through FY2027. (Consensus; 16 and 14 analysts.) Against a June quarter that grew 4.7% and a United States property business that grew 2.7% in FY2025. Reasonable and unaggressive.
• Adjusted EBITDA margin holds near 65-66%. (Our derivation; the June quarter was 65.8%, down 93 basis points.) This is the most fragile assumption in the price, because it is the one already moving in the wrong direction: revenue grew 4.7% and Adjusted EBITDA grew 3.0%.
• DISH churn is contained at the disclosed level. (Company disclosure.) DISH was "approximately 2% and 4% of our total annual property revenue and total annual U.S. & Canada property revenue" for 2025 and is now in churn, with $17.5 million of impairment already taken. On a United States business growing 2.7%, a 4% churn event is more than a year of growth.
• The dividend keeps growing. (Company practice; $6.98 trailing, $3,157.2 million paid in FY2025.) It is covered by $3,960 million of trailing free cash flow — after ALL capital spending including discretionary construction — with roughly $800 million of headroom. On an AFFO basis the coverage would be materially better. A REIT yielding 3.99% whose distribution stops growing re-rates immediately.
• Long-term interest rates do not rise materially. (Our derivation.) $37.2 billion of debt, a 3.99% yield and a 42.3% peak-to-trough drawdown in twelve months are the same fact stated three ways. At 15x corrected EV/EBITDA the equity is approximately $154; at 21x it is approximately $252.

5b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: Adjusted EBITDA growth (+4%) + multiple drift (EXPANSION, from approximately 17.3x toward 18.0x) + dividend yield (+3.99%) ≈ +16% — but note the leverage amplification: with net debt and noncontrolling interests of roughly $42 billion against $81.6 billion of equity, each turn of EV/EBITDA is approximately $15.30 per share, 8.7% of the price.

Our base assumes the multiple expands modestly and the reason is where it already sits. Tower REITs traded well above 20x EV/EBITDA through the low-rate decade; approximately 17.3x reflects a rate-driven de-rating rather than a deterioration in the leases. What makes this a Watch rather than a Buy is that the same leverage which amplifies a re-rating amplifies its absence, and this dive has no view on long-term interest rates and says so.

Note also the structure: roughly a quarter of the expected return is the dividend, which is covered and growing, and the rest requires either EBITDA growth above 4% or a multiple recovery. That is a more defensive composition than most of this batch and it is the reason the risk score is 6 rather than 7.

5c. Variant perception (where we differ, what would surprise)

• We think the leverage is materially better than the standard data shows, and this is the clearest quantified disagreement in the file. netDebtToEBITDATTM reads 6.15x. The correct figure is approximately 5.0x. The difference is $7,743.6 million of operating lease liabilities added to borrowings — a reconciliation exact to the dollar against the filing. Anyone screening REITs on a leverage cut-off is seeing American Tower a full turn worse than it is.
• We think the 18.5% "beat" is meaningless and the 3.0% Adjusted EBITDA growth is the number. The reported beat is on GAAP EPS, in a quarter whose net income rose 133% largely because "other (income) expense" swung $427.7 million on currency translation. Watchable number: Adjusted EBITDA margin in the 2026-10-27 release, 65.8% in June against 66.7% a year earlier.
• We differ from the street on the level by only 3.2% — our $203 against a consensus target of $209.75. The notable feature is the distribution: 40 of 50 analysts positive, ZERO sells, and a LOW target of $188 that is 7.3% ABOVE the market price. As on Capital One, Williams, Robinhood, Moody's, ICE and Elevance in this programme, a covering community with no downside case is treated as a caution rather than a comfort — particularly on a security that fell 42.3% peak to trough within the last twelve months.
• The knowledge base has NOTHING, and on a name flagged in advance as collision-prone the absence of even homograph noise is worth recording. Zero entity matches in 52,021 for both AMT and American Tower. Communications-site real estate is a genuine sector void. On a levered, rate-sensitive security with a 42-point twelve-month relative deficit, having no independent expert view at all is a real limitation and is stated as one.
• Positive surprise that would force a re-rating: Data Centers revenue growth above 15% with separately disclosed segment economics — at $1,053.1 million and 13.9% growth it is the only part of this company on a different curve; a dividend increase above 5%; corrected net leverage below 4.75x; or United States property organic growth above 4% once DISH churn laps.
• Negative surprise that would break the thesis: Adjusted EBITDA margin below 64%; the dividend held flat; DISH churn exceeding the disclosed 4% of United States and Canada property revenue, or a second tenant entering churn; a currency move that materially raises reported leverage on international debt; or corrected net leverage rising above 5.5x.

Synthos fair values

All three anchors are multiples of estimated FY2027 Adjusted EBITDA of approximately $7,580 million — built from the filed first-half 2026 run rate of $3,643.4 million grown at approximately 4%, and labelled as our estimate rather than consensus, because est.ebitdaAvg carries a fixed-ratio fabrication signature and is rejected. Each is converted to an equity value by deducting net debt of approximately $35,427 million and noncontrolling interests of approximately $6,532 million and dividing by 465.96 million shares. Cross-checked against the dividend yield.

• Bear ~$138 — 14.0x FY2027E Adjusted EBITDA. Cross-check: a 5.06% dividend yield at that price; 14.9% below the 52-week low of $162.11; consistent with the 42.31% peak-to-trough drawdown this security has already experienced within twelve months. The scenario: long-term rates rise, the Adjusted EBITDA margin slips below 64%, DISH churn proves larger than disclosed, and a 5.0x-levered REIT de-rates further. −21.2%.
• Base ~$203 — 18.0x FY2027E Adjusted EBITDA. Cross-check: a 3.44% dividend yield at that price; 4.7% below the 52-week high of $212.98; a 4.20% free-cash-flow yield. Sensitivity, stated openly, and the leverage amplification is the point: each turn of EV/EBITDA is approximately $15.30 per share, 8.7% of the price, so 17x gives $188 and 19x gives $218. The whole answer lives in a 17-19x band and a levered REIT converts small multiple changes into large equity moves. The scenario: revenue compounds 3.3%, margin holds near 65%, the dividend keeps growing, and the multiple recovers a modest part of the rate-driven de-rating. +15.9%.
• Bull ~$248 — 20.7x FY2027E Adjusted EBITDA. Cross-check: a 2.82% dividend yield; 16.4% above the 52-week high. The scenario: long-term rates fall, Data Centers accelerates above 15% and is separately valued, DISH churn laps, and tower REITs re-rate toward the multiples they held before the rate cycle. +41.6%.

Base is 15.9% above spot; asymmetry roughly 1.96:1 (21.2% down, 41.6% up), with a 3.99% dividend yield underneath it. That is four hundredths short of the bar, and this dive is not going to move a multiple to close the gap. The verdict is Watch with a written trigger, and the reasons for not rounding up are in Section 9.

6. Knowledge base — nothing at all

Raw hits: 0. Name-level claims on American Tower: 0.

The case-sensitive entity sweep ran AMT and American Tower across all 52,021 distilled claims. It returned NOTHING.

Two things follow and both are worth stating.

There is no homograph contamination either, and this name was flagged in advance as collision-prone. The ticker AMT collides in ordinary usage with "alternative minimum tax" and "automated manual transmission", and the store returned neither — a clean zero rather than a contaminated positive. That is a materially better outcome than CEG (26 of 29 matches were Constellation Software), PWR (Quanta Computer at conviction 85), EMR (whose only entity match was Emerson Collective, an unrelated LLC), USB (Universal Serial Bus) or MMM ("$3M" strings) produced, and it is recorded as a favourable data result.

And there is no adjacent thematic lane to quote at zero weight. On several other names in this batch — Royal Caribbean, Williams, HCA — a free-text sweep returned sector-level material that could at least be reported as discarded. Here there is nothing a reader would recognise as relevant to tower-leasing economics: no claim about site leasing, tenant escalators, carrier capital expenditure or REIT structure. The lane is empty and is not padded.

Communications-site real estate joins US natural gas midstream, cruise lines, acute-care hospitals, US refining and marketing, managed care, insurance brokerage, environmental services, diversified industrials, architectural coatings and regulated exchanges on the list of genuine sector voids in this store. That is now eleven sectors identified across this batch alone, which is itself a finding about the coverage of the ingest pipeline rather than about any of these companies.

Conclusion. Breadth 0, claim count 0, net conviction none. The Synthos knowledge base has nothing whatever to say about American Tower Corporation. On a security carrying approximately 5.0x net leverage, a 42-point twelve-month relative deficit and a 42.3% peak-to-trough drawdown, the complete absence of independent expert support is a real input to the verdict and is why the trigger in Section 9 is a price rather than a judgment.

7. Data integrity — two predicted defects confirmed, and a clean capex check

Six findings.

0. THE REIT PROBLEM — the payload contains no funds-from-operations field of any kind. est.epsAvg and earn_cal both carry GAAP earnings per share: the FY2025 consensus of $5.122 against a reported GAAP diluted $5.39, and the June-quarter calendar actual of $1.86 matching the 10-Q's GAAP figure exactly. priceToEarningsDilutedRatioTTM of 24.09x and priceToBookRatioTTM of 21.95x are therefore artefacts of depreciation and of a REIT's distribution-driven equity base, and neither is used anywhere in this dive. The 10-Q defines Nareit FFO and AFFO and reconciles both; the per-share figures did not survive the extraction window searched, and this dive does NOT invent one. Valuation is built on Adjusted EBITDA ($1,808.2M in the quarter, filing-verified), the recomputed free-cash-flow yield (4.85%) and the dividend (3.99%), and the substitution is disclosed rather than silent. This is a structural limitation of the vendor for every REIT in the universe, not a defect specific to American Tower.

1. totalDebt overstates borrowings by 20.8% and the reconciliation is EXACT — REJECTED, filing substituted. Vendor $44,963.9M at 2025-12-31 against the balance sheet's current portion of long-term obligations $3,387.8M plus long-term obligations $33,832.5M = $37,220.3M. The $7,743.6M gap equals the current portion of operating lease liability ($584.9M) plus the operating lease liability ($7,158.7M) to the dollar. The largest instance of the lease-inclusion class in this batch — ahead of Sherwin-Williams (19.1%), Phillips 66 (16.1%), HCA (8.0%, double-counted), Cummins (7.4%) and Royal Caribbean (3.3%). Consequential correction: netDebtToEBITDATTM of 6.15x becomes approximately 5.0x on the filing's June-2026 debt of $37,189.6M less cash of $1,762.5M against trailing Adjusted EBITDA of roughly $7,150M. The correction runs in the company's favour and is the clearest quantified variant perception in this dive.

2. Enterprise value omits $6,702.8 million of noncontrolling interests — CONFIRMED, predicted in advance, and the largest absolute instance in this batch. enterpriseValueTTM of $124,828M implies $43,217M of net claims against FY2025 net debt of $43,489M — a $272M residual — with noncontrolling interests entirely excluded. Adding them gives approximately $131,802M, 5.6% higher. AMT was named in advance as a likely instance of the KKR (35.6%) / MPC ($6,772M) / FCX (11.5%) class and it is confirmed at $6.70 billion — ahead of HCA's $3,433M, Phillips 66's $1,148M and Cummins' $1,059M in this batch. The June 2026 balance sheet shows noncontrolling interests of approximately $6,532M against total equity of $10,252.5M — 63.7%. The corrected enterprise value of approximately $123,570M, built on the June balance sheet with corrected debt, is what this dive uses.

3. seg_geo FY2025 over-counts by exactly its services line, and FY2024 has a line overwritten — REJECTED as published, rebuilt. FY2025 lists five "Property," geographies plus a "Services Segment" line of $339.6M, summing to $10,984.2M against revenue of $10,645.1M — 103.2%. The five geographic lines alone sum to $10,644.6M, matching revenue to within 0.005%, so they already include services and the separate line double-counts it exactly — the CDNS / CSX over-count class with the excess equal to the added line. Worse in FY2024: the block reports "Property, Africa: $193,700,000" against $1,225.6M in FY2023 and $1,422.9M in FY2025, and $193.7M is EXACTLY the FY2024 "Services Revenue" figure from seg_prod — the African revenue line has been overwritten with the services figure. FY2024 is rejected; the FY2025 five-line version is used.

4. The 2026 SEGMENT REALIGNMENT is not labelled and produces a misleading series. The 10-Q states that property operations "includes our U.S. & Canada property, Africa & Asia-Pacific ("APAC") property, Europe property and Latin America property segments and Data Centers segment" — Africa and Asia-Pacific have been COMBINED. The vendor reflects the consequence without the label: FY2025 shows "Property, Africa" at $1,422.9M and NO Asia-Pacific line, where FY2023 showed Africa at $1,225.6M and Asia-Pacific at $1,150.8M. A reader would conclude that Asia-Pacific revenue vanished and African revenue grew sevenfold from a corrupted FY2024 base. Neither happened.

5. est.ebitdaAvg and est.ebitAvg carry a fixed-ratio fabrication signature that happens to land near the truth — REJECTED on principle. In every year from FY2026 to FY2030, ebitdaAvg is exactly 64.16% of revenueAvg. American Tower's actual trailing EBITDA margin is 64.21% and its Adjusted EBITDA margin is 65.8%, so unlike DASH's 170%-overstated instance this fabrication is close to correct in level. It is still a fixed ratio applied mechanically across five forward years, and a row that is right by accident is not a row to value on. The FY2027 Adjusted EBITDA used in this dive is built from the filed first-half run rate and is labelled as our estimate.

6. inc_q for the September 2025 quarter reports interestExpense of ZERO — corrupt, and noted. The surrounding quarters read $342.6M, $350.0M, $344.4M and $354.5M. A zero interest expense for a company with $37.2 billion of debt is not a number that exists. It affects no conclusion here because the filing's figures are used throughout, and it is flagged.

Clean verifications worth recording, and there are three. The capital-expenditure check PASSED — the recomputed trailing free cash flow of $3,959.8M, built from the 10-Q's H1 2026 figure of $770.4M and H1 2025's $635.7M, reproduces freeCashFlowYieldTTM (4.85207% against our 4.852%) and capexToOperatingCashFlowTTM (0.31431 against our 0.3143) exactly, with capexToDepreciationTTM of 0.876 plausible and corroborated. The share count is EXACT — the 10-Q reports 465,959 thousand shares outstanding at 2026-06-30 against 465,957 thousand implied by market capitalisation ÷ price. And seg_prod ties — FY2025 Property $10,305.0M plus Services $339.6M = $10,644.6M against inc_a revenue of $10,645.1M, a 0.005% difference, with no over-count and no omission.

Non-equity tripwire — checked and passed. AMT is common stock, $0.01 par value, NYSE-listed, per the 10-Q balance sheet (1,000,000 thousand shares authorised, 480,163 issued, 465,959 outstanding). Price of $175.145 is not par-like; beta is 0.909; the dividend is a declared and historically rising quarterly distribution ($6.98 trailing, 3.99% yield); volume was 1.97M shares (~$344M of turnover); the 52-week band of $162.11 to $212.98 is a 31% range. No preferred stock is outstanding; noncontrolling interests of approximately $6,532 million are NOT this security and are addressed in finding 2. The registrant also has multiple listed senior note series (0.450% due 2027 through 4.625% due 2031 and beyond, per the 8-K cover pages) which are separate instruments. This is common equity.

Vendor composite rating — noted, and substantially void. B+ / 3 overall with 5 on return on equity, 5 on return on assets, 4 on discounted cash flow, and 1 on debt-to-equity and 1 on price-to-book. The debt-to-equity score of 1 is computed on the 20.8%-overstated debt of finding 1 and on a REIT equity base depleted by distributions; the price-to-book score of 1 on the same base. Not used.

8. Technicals and insiders

• Price $175.145. −17.76% from the 52-week high of $212.98; +8.04% above the low of $162.11. Maximum drawdown from peak over the trailing year: −42.31% — more than twice the current distance from the high, so this security fell over 40% and has recovered part of it within twelve months.
• Almost exactly on both moving averages: 0.5% BELOW a 50-day of $175.98; 2.1% BELOW a 200-day of $178.82. Both averages sit within 2.1% of the price — the flattest configuration in this batch.
• RSI 59.3; MACD +0.20 — as close to neutral as this batch produces.
• Beta 0.909, which for a security that fell 42.3% peak to trough is a reminder that beta measures correlation, not risk.
• Relative performance: 3-month −3.17% against SPY +7.59%; 6-month +0.83% against SPY +11.09%; 12-month −17.42% against SPY +24.26% and QQQ +30.80% — a FORTY-TWO-POINT and forty-eight-point deficit, the second-worst in this batch after Intercontinental Exchange.

AMT closed 2026-08-04 at $175.145, up 1.24% or $2.15 from $172.995, having opened at $170.905 and traded $169.03 to $175.55 — closing at the day high on 1.97 million shares. No company-specific filing is dated 2026-08-04; the last event was the 2026-07-28 earnings release and 10-Q, seven days earlier.

Insiders — every substantive transaction is a sale

Five of the eight rows are open-market SALES; one is a gift; two are empty. There is not one purchase. Robert Joseph Meyer sold 5,000 shares at $178.89 on 29 July 2026 — the day after the earnings release — and Ruth T. Dowling sold in four tranches across April and July at $169.54 to $178.48, all above today's close of $175.145.

The amounts are small — roughly $1.4 million in total — and the pattern is consistent with programmed selling. But every substantive transaction in the file is a disposal, executed at prices at or above the current one, on a security that has underperformed the index by 42 points over twelve months. On a name where the knowledge base is completely empty, the only outside signal available is insider behaviour, and it is uniformly negative. It is reported as such and it is one of the reasons a 1.96:1 payoff is not rounded up.

9. Verdict, kill-criteria and flip conditions

Watch.

The arithmetic first, because it is close. Base fair value $203 (+15.9%), bear $138 (−21.2%), bull $248 (+41.6%) — 1.96:1, with a 3.99% dividend yield underneath. Four hundredths short of the bar, and this dive is not going to move a multiple to close the gap.

What is genuinely good, and two of these are corrections rather than opinions. The capital-expenditure check came back verified clean, with a recomputed trailing free cash flow of $3,959.8 million — a 4.85% yield — reproducing the vendor's own ratios exactly from filed half-year figures, and covering the $3,157.2 million dividend with roughly $800 million to spare AFTER all discretionary construction spending. The leverage is a full turn better than the standard data shows: totalDebt of $44,963.9 million against the filing's $37,220.3 million, with the $7,743.6 million difference reconciling to the operating lease liability to the dollar, taking net leverage from a reported 6.15x to approximately 5.0x. Adjusted EBITDA margin is 65.8%. The de-rating has already happened — approximately 17.3x corrected enterprise value to Adjusted EBITDA against a tower-REIT history well above 20x — and the shares sit within 2.1% of both moving averages after a 42.3% peak-to-trough drawdown.

What holds it at Watch, in four items.

First, the metric this industry is valued on is not available. The payload contains no funds-from-operations field of any kind, and the AFFO per-share figure did not survive the filing extraction window searched. We are valuing a REIT on EBITDA and free cash flow because the standard measure is missing, and that is a real limitation on precision, not a stylistic choice.

Second, the operating trend is the wrong way at the margin. Revenue grew 4.7% and Adjusted EBITDA grew 3.0% — margin down 93 basis points to 65.8%. The United States, 52.5% of revenue, grew 2.7% in FY2025. And DISH, disclosed at approximately 4% of United States and Canada property revenue, is now in churn with $17.5 million of impairment already taken — more than a year of growth in the largest segment.

Third, the knowledge base is completely empty and the only other outside signal is negative. Zero claims in 52,021 — with no homograph noise either, so the zero is real. And every substantive insider transaction on file is an open-market sale, all executed at or above today's price.

Fourth, the driver is a variable nobody in this document can forecast. $37.2 billion of debt, a 3.99% yield and a 42.3% twelve-month peak-to-trough drawdown are the same fact three ways: this is a rate-sensitive security, and the base case requires a modest multiple recovery that depends on long-term rates rather than on anything management does.

And the honest counterweight. Forty of fifty analysts are positive, there is not one sell rating, and the LOWEST target on the street is $188 — 7.3% above the price and 7.4% below our base. Our $203 is only 3.2% below their consensus. This is not a case where we see something the market does not; it is a case where we want a slightly better price for a leveraged, un-covered, un-verifiable-on-the-right-metric security.

Pre-registered BUY trigger — what would make this Buy — Tactical:

• A price at or below $160 with the FY2027 Adjusted EBITDA estimate intact — approximately 16.4x on our construction, 1.3% below the 52-week low, and a 4.36% dividend yield, restoring approximately 4.0:1 asymmetry against the $138 bear and the $248 bull.
• Or, at any reasonable price, disclosure of AFFO per share growing above 5% together with a dividend increase above 5% and corrected net leverage below 4.75x — which would resolve simultaneously the metric problem, the distribution question and the leverage question that are the three reasons this is not already a Buy.

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

• Adjusted EBITDA margin below 64%, from 65.8% and falling.
• The dividend held flat, which for a REIT yielding 3.99% is an immediate re-rating event.
• DISH churn exceeding the disclosed 4% of United States and Canada property revenue, or a second named tenant entering churn.
• Corrected net leverage above 5.5x on the filing's debt basis.
• United States property revenue declining year on year, from +2.7%.
• Data Centers growth falling below 8%, which would remove the only accelerating line in the company.

Where AMT fits in the Synthos Framework Portfolio. No position today. Tracked on the real-estate / infrastructure watch list at the $160 buy trigger above, with a 2% initial size and a 4% target if triggered. Sizing note: this is a security whose equity value moves approximately 8.7% for every turn of enterprise-value-to-EBITDA because of the leverage, which means the position size should be set against the leverage rather than against the beta of 0.909 — a low correlation to the index is not the same as low volatility, and a 42.3% peak-to-trough drawdown inside twelve months is the evidence. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $175.145, with the fair-value anchors, both triggers, the kill criteria and the 5a assumptions all gradeable. Also logged as a STRUCTURAL DATA flag for the programme rather than for this name alone: the vendor carries no FFO or AFFO field for REITs, so every real-estate name in this universe is unvaluable on its industry-standard metric from the payload, and the two corrections found here — a 20.8% debt overstatement from operating leases and a $6.70 billion noncontrolling-interest omission — are both likely to recur across the REIT cohort.

Single biggest risk: long-term interest rates against $37.2 billion of debt. American Tower's leases are contracted, its escalators are written down, and its Adjusted EBITDA margin is 65.8% — none of which prevented the shares falling 42.31% from peak to trough within the last twelve months or underperforming the index by 42 percentage points. A real estate investment trust that distributes $3.16 billion a year, yields 3.99% and carries approximately 5.0x net leverage is priced against the long bond whatever the quality of its tenants. The corrections in this dive make the balance sheet look better than the standard data shows — a full turn of leverage better — and they do not change the direction of that sensitivity. What would change it is the metric this file cannot source: if AFFO per share is growing at 5% and the dividend with it, then approximately 17.3x corrected enterprise value to Adjusted EBITDA is cheap and the street's $209.75 is the right number. This dive could not establish that from the data available, and it says so rather than assuming it.

Provenance & disclosures

• Traceability: ZERO name-level knowledge-base claims name American Tower Corporation out of 52,021 distilled claims (raw hits 0, name-level 0; breadth 0, claim count 0, net conviction none). Recorded separately as a favourable data result: this ticker was flagged IN ADVANCE as collision-prone — AMT collides in ordinary usage with "alternative minimum tax" and "automated manual transmission" — and the store returned NO homograph contamination whatever. A clean zero rather than a contaminated positive, materially better than CEG (26 of 29 matches were Constellation Software), PWR (Quanta Computer at conviction 85), EMR (Emerson Collective, an unrelated LLC), USB (Universal Serial Bus) or MMM ("$3M" strings). There is also no adjacent thematic lane to quote at zero weight: no claim in the store addresses site leasing, tenant escalators, carrier capital expenditure or REIT structure in a form a reader would recognise as relevant. The lane is empty, is reported as empty, and is not padded. Communications-site real estate joins US natural gas midstream, cruise lines, acute-care hospitals, US refining and marketing, managed care, insurance brokerage, environmental services, diversified industrials, architectural coatings and regulated exchanges — ELEVEN sectors identified as genuine voids across this batch alone, which is a finding about the ingest pipeline's coverage rather than about any of these companies. On a security carrying approximately 5.0x net leverage, a 42-point twelve-month relative deficit and a 42.3% peak-to-trough drawdown, the complete absence of independent expert support is a real input to the verdict.
• Data as-of: fundamentals — income statement, balance sheet, cash flow, Adjusted EBITDA reconciliation, segment structure, share counts and the DISH disclosure through 2026-06-30, all filing-verified from the 10-Q filed 2026-07-28; annual figures for FY2025 and earlier from the 10-K filed 2026-02-24 · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873597 = 2026-08-04T19:59:57Z ($175.145, +1.24%; 50-DMA $175.98; 200-DMA $178.82; RSI 59.3; MACD +0.20; maximum twelve-month drawdown −42.31%; beta 0.909) · knowledge-base claims 2026-08-04. American Tower's fiscal year is the calendar year and the company is a real estate investment trust. All figures come from the Synthos vendor data file for AMT or from the SEC filings in the AMT archive; no figure comes from memory, recall or external retrieval.
• THE REIT LIMITATION — stated prominently because it governs the whole document: the vendor payload contains NO funds-from-operations field of any kind. est.epsAvg and earn_cal both carry GAAP earnings per share — the FY2025 consensus of $5.122 against a reported GAAP diluted $5.39, and the June-quarter calendar actual of $1.86 matching the 10-Q's GAAP figure exactly. priceToEarningsDilutedRatioTTM of 24.09x is therefore a DEPRECIATION ARTEFACT on a company that charged $1,032.6 million of depreciation, amortisation and accretion in six months against $2,887.4 million of operating cash flow, and priceToBookRatioTTM of 21.95x is an artefact of a distribution-depleted equity base. Neither is used anywhere in this dive. The 10-Q defines and reconciles both Nareit FFO and AFFO; the per-share figures did not survive the extraction window searched, and this dive does NOT invent one. Valuation is built instead on Adjusted EBITDA ($1,808.2M in the quarter and $3,643.4M in the half, filing-verified), the recomputed free-cash-flow yield (4.85%) and the dividend (3.99%), with the substitution disclosed at every point. This is a structural limitation of the vendor for every REIT in this universe, not a defect specific to American Tower, and it is logged as such.
• Filing archive contents: 10-K filed 2026-02-24 (fiscal 2025); 10-Q filed 2026-04-28 (March 2026 quarter); 10-Q filed 2026-07-28 (June 2026 quarter — the source of the balance sheet, the cash-flow statement, the Adjusted EBITDA and FFO/AFFO reconciliation framework, the segment description including the combined Africa & APAC segment, the DISH churn and impairment disclosure, the South Africa Fiber sale reference, the equity statement and the share counts); 8-K filed 2026-06-02 and 8-K filed 2026-07-28 (cover-page-only filings listing the registrant's common stock and its many listed senior note series — 0.450% due 2027 through 4.625% due 2031 and beyond — with no substantive item content in the extraction); 8-K filed 2026-06-04 (signed by Rodney M. Smith, EVP, Chief Financial Officer and Treasurer; forward-looking-statements language only in the extraction). All carry preserved [TABLE] statement data (tables: true); the 10-Q's 86 preserved tables made every check in Section 7 possible. No spin-off, split, notes offering or acquisition appears with substantive detail in the 8-K record.
• Where the filings contradicted or corrected the vendor (detailed in Section 7): totalDebt, at $44,963.9M against the balance sheet's $37,220.3M (current portion of long-term obligations $3,387.8M plus long-term obligations $33,832.5M) — a 20.8% overstatement in which the $7,743.6M gap equals the current portion of operating lease liability ($584.9M) plus the operating lease liability ($7,158.7M) TO THE DOLLAR — the largest instance of the lease-inclusion class in this batch, ahead of SHW (19.1%), PSX (16.1%), HCA (8.0%, double-counted), CMI (7.4%) and RCL (3.3%) — with the consequential correction taking netDebtToEBITDATTM from 6.15x to approximately 5.0x, a full turn IN THE COMPANY'S FAVOUR; enterpriseValueTTM, which omits $6,702.8M of noncontrolling interests (approximately $6,532M and 63.7% of total equity on the June 2026 balance sheet), confirming the KKR / MPC / FCX class flag raised on this name IN ADVANCE and constituting the LARGEST ABSOLUTE NCI omission in this batch, ahead of HCA's $3,433M, PSX's $1,148M and CMI's $1,059M; seg_geo FY2025, which over-counts to 103.2% of revenue because a separate "Services Segment" line of $339.6M duplicates services revenue already inside the five geographic lines (which alone sum to $10,644.6M, matching revenue to within 0.005%), and seg_geo FY2024, in which the "Property, Africa" line reads $193.7M — EXACTLY the FY2024 "Services Revenue" figure — having overwritten the true African figure (which was $1,225.6M in FY2023 and $1,422.9M in FY2025); the unlabelled 2026 SEGMENT REALIGNMENT, with the 10-Q disclosing a combined "Africa & Asia-Pacific ('APAC') property" segment that the payload reflects as a vanished Asia-Pacific line and a sevenfold increase in African revenue from a corrupted base; est.ebitdaAvg/ebitAvg, rejected for a fixed-ratio fabrication signature (exactly 64.16% of revenueAvg in every year FY2026-FY2030) which — unusually, and unlike DASH's 170%-overstated instance — happens to land near the actual 64.21% trailing margin, and is rejected on principle because a row that is right by accident is not a row to value on; and inc_q interestExpense of ZERO for the September 2025 quarter, against $342.6M-$354.5M in the surrounding quarters, on a company with $37.2 billion of debt. Where vendor and filing AGREED — recorded, because clean verifications are real findings: the capital-expenditure check PASSED — the recomputed trailing free cash flow of $3,959.8M, built from the 10-Q's H1 2026 "Payments for purchase of property and equipment and construction activities" of $770.4M and H1 2025's $635.7M, reproduces freeCashFlowYieldTTM (4.85207% against our 4.852%) and capexToOperatingCashFlowTTM (0.31431 against our 0.3143) exactly, with capexToDepreciationTTM of 0.876 plausible and corroborated, datum and diagnostic tested separately per the MPC lesson; the share count is EXACT (10-Q 465,959 thousand outstanding at 2026-06-30 against 465,957 thousand implied by market capitalisation ÷ price); and seg_prod ties (FY2025 Property $10,305.0M plus Services $339.6M = $10,644.6M against inc_a revenue of $10,645.1M, a 0.005% difference — no over-count, no omission).
• Basis note: this dive uses Adjusted EBITDA as the primary earnings measure, as the company does. The 10-Q defines it as "Net income before Income (loss) from equity method investments; Income (loss) from discontinued operations, net of taxes; Income tax benefit (provision); Other income (expense); Gain (loss) on retirement of long-term obligations; Interest expense; Interest income; Other operating income (expense), including Goodwill impairment; Depreciation, amortization and accretion; and stock-based compensation expense", and the June-quarter figure of $1,808.2M is filing-verified. GAAP net income is NOT used as a valuation input, and the reason is visible in the quarter: net income rose 133% to $887.5M largely because "Other (income) expense" swung $427.7 million — from a $373.9M expense to a $53.8M income — on currency translation and unrealised marks, and across the half by $856.1 million. The reported 18.5% "beat" ($1.86 against $1.57) is on GAAP diluted EPS and measures something this industry does not underwrite.
• Estimate coverage: revenue coverage is good — 16 analysts on FY2026, 14 on FY2027 and 14 on FY2028 — and EPS coverage is thin and irrelevant, being 7 and 10 analysts on a GAAP measure this dive does not use. FY2029 and FY2030 EPS rest on 10 and 5 analysts and are excluded from every conclusion. The FY2027 Adjusted EBITDA figure anchoring all three fair values is OUR estimate — approximately $7,580M, built from the filed first-half 2026 run rate of $3,643.4M grown at approximately 4% — and is labelled as such because est.ebitdaAvg is rejected in Section 7.
• Peer note: the peers block for American Tower was not usable for comparison in this dive and no peer-multiple comparison is drawn. The relevant comparison set — Crown Castle, SBA Communications, Equinix and Digital Realty — would in any case require the AFFO metric this payload does not carry, so a peer multiple could not be constructed on the right basis even with a correct peer list.
• Price-target note: pt reports targetHigh $240, targetLow $188, targetConsensus $209.75 and targetMedian $209 — a genuine four-point distribution. The LOW target of $188 is 7.3% ABOVE the spot price of $175.145, and grades shows 1 strong buy, 39 buy, 10 hold and ZERO sell across 50 analysts — 40 of 50 positive. As on Capital One, Williams, Robinhood, Moody's, ICE and Elevance in this programme, this dive treats a covering community with no downside case as a caution rather than a comfort — particularly on a security that fell 42.31% peak to trough within the last twelve months. Our base of $203 is 3.2% below their consensus and 7.4% above their most bearish analyst, so this is not a case of seeing something the market does not. Separately, quote.yearHigh/yearLow ($214.79/$160.06) disagree with tech.hi52/lo52 ($212.98/$162.11) by 0.8% and 1.3%; tech is used throughout, and note that tech.max_dd_from_peak of −42.31% is more than twice pct_from_hi of −17.76%, which is material context a reader would otherwise miss.
• Fair-value caveat: the $138 / $203 / $248 anchors are 14.0x, 18.0x and 20.7x estimated FY2027 Adjusted EBITDA of approximately $7,580 million — OUR estimate, built from the filed H1 2026 run rate, because est.ebitdaAvg carries a fixed-ratio fabrication signature — each converted to an equity value by deducting net debt of approximately $35,427M (on the FILING's debt figure, not the vendor's) and noncontrolling interests of approximately $6,532M, and dividing by 465.96 million shares. Cross-checked against the dividend yield (5.06%, 3.44% and 2.82%). No AFFO multiple is used, because the metric could not be sourced. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed and the LEVERAGE AMPLIFICATION is the key point: with roughly $42 billion of net debt and noncontrolling interests against $81.6 billion of equity, each turn of EV/EBITDA is approximately $15.30 per share — 8.7% of the price — so 17x gives $188 and 19x gives $218. A levered REIT converts small multiple changes into large equity moves in both directions, and no greater precision is claimed than the $138-$248 band.
• Timing: second-quarter 2026 results were released 2026-07-28, seven days before this dive, with the 10-Q filed the same day. GAAP diluted EPS of $1.86 beat the $1.57 estimate by 18.5% — on a metric this industry does not use, in a quarter whose net income rose 133% largely on a $427.7 million non-operating swing — while Adjusted EBITDA grew 3.0% and its margin fell 93 basis points to 65.8%. The next print is 2026-10-27, 84 days away (consensus GAAP EPS $1.63, revenue $2,756M). The most recent insider filings are dated 2026-07-29 and are open-market SALES; five of the eight rows in the file are sales, one is a gift, two are empty, and there is not one purchase. 2026-08-04 carried no company-specific filing.
• 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.