American Tower Corporation (NYSE: AMT) — The Blue-Chip Tower, Derisked and Repaired, Now Priced to Match

Independent fundamental research. Report date: 2026-06-13. The analysis body carries no recommendation and no price target; the sole exception is the clearly-labeled Author’s Take block below.

⚡ Author’s Take

This block is the author’s own subjective opinion. It is general information, not investment advice. The analysis that follows takes no position and names no price target.

Verdict: HOLD / accumulate-on-weakness — the derisked, best-balance-sheet, most-optionality tower at a fair (not cheap) price. The quality middle of the trio. American Tower is the blue-chip of the three public tower REITs: the largest portfolio (~150,000 sites across the US, Europe, Africa, and Latin America), the only one with a real data-center/interconnection business (CoreSite — the differentiated AI/edge optionality the other two lack), the lowest leverage (~5.0x net, versus SBAC’s ~6.6x and CCI’s ~8.7x-pre-paydown), the best interest coverage (~5.1x), the most conservative AFFO payout (~63%, versus ~90% at both peers), and an investment-grade, ~97%-fixed, fully-unsecured balance sheet. It has spent 2023–2025 cleaning itself up — exiting chronically-troubled India (~$2.5B, removing the Vodafone Idea credit overhang), resetting a stretched dividend to delever, putting ROIC into the comp plan, and initiating buybacks — a genuine, disciplined repair after the 2021 binge (Telxius + CoreSite, ~$19.5B in one year at the cycle top) that loaded ~$27B of goodwill and pushed leverage to 9.2x.

The catch is the same one that holds it back from being a table-pounding buy: AFFO/share has been essentially flat for three years (~$10.5–10.9, FY2023→FY2026E) on India-exit dilution, FX translation (Brazil/Africa), the DISH default (~400bps of 2026 churn), and refinancing step-ups — and the stock, while de-rated hard from ~$292 (2021) to ~$187, trades at ~17x AFFO / ~18.7x EV/EBITDA and the 44th percentile of its own valuation history — fairly valued, neither the bargain SBAC is (7th percentile) nor the premium CCI commands. The framing is “highest-quality, most-derisked, most-optionality tower — bought at a fair price for exactly that.” This is the own-it-and-sleep name: lowest risk, best balance sheet, real CoreSite/AI upside, a ~3.8% growing dividend — but the near-term per-share growth is flat and you are paying fair value, so the return is moderate rather than asymmetric. My fair-value accumulation zone is ~$165–185 (~15–17x AFFO) — near the 52-week low (~$165) is where the risk/reward genuinely tilts in your favor and the ~4% yield plus a 2027 AFFO re-acceleration would compound nicely. Conviction: medium. Within the trio I’d rank SBAC (value) ≥ AMT (quality-at-fair-price) > CCI (avoid). What flips me more bullish: a confirmed 2027 AFFO/share re-acceleration to mid-single-digits, CoreSite’s AI-interconnection inflection proving durable, or a de-rate into the low-$170s/$160s. What flips me bearish: an EM/FX shock (Brazil, Africa) or rates staying high enough to keep AFFO/share flat into 2027. Tag: the blue-chip tower, repaired and de-risked — and the market already knows it.

1. Executive Summary

American Tower is the largest of the three public US-listed tower REITs and the most diversified business in the group — a global portfolio of roughly 149,700 communications sites (US & Canada ~42,200; Latin America ~47,100; Europe ~32,500; Africa & APAC ~27,900) plus CoreSite, a carrier-neutral data-center and interconnection platform (~28 facilities, ~10% of revenue) that gives AMT a genuine AI/edge/interconnection growth engine the other two towers lack. FY2025 revenue was ~$10.6B and adjusted EBITDA ~$7.0B at a ~65.5% margin — roughly 2.5x the size of CCI and nearly 4x SBAC.

The investment character of AMT is “blue-chip, derisked, fairly priced.” On the measures of balance-sheet quality it leads the group decisively: net leverage ~5.0x (versus SBAC ~6.6x and CCI ~8.7x pre-paydown), interest coverage ~5.1x (the best of the three), a ~63% AFFO payout (the most conservative — leaving room for both deleveraging and the buybacks it has now initiated), an investment-grade, ~97%-fixed-rate, fully-unsecured debt stack, and — uniquely among the three — positive book equity. It is also the most diversified: developed-market stability (US, Europe) plus emerging-market growth (Africa, LatAm) plus the CoreSite data-center optionality.

The last two years have been a story of disciplined repair under CEO Steven Vondran (appointed February 2024). AMT had overreached at the 2021 cycle top — acquiring Telxius (~$9.4B) and CoreSite (~$10.1B) in a single year, driving goodwill and intangibles to ~$27B and net leverage to ~9.2x. Since then it has exited India (September 2024, ~$2.5B to a Brookfield-led consortium, removing ~76,000 low-margin, churn- and credit-impaired sites and the Vodafone Idea overhang), exited Australia/New Zealand and South Africa fiber, reset and paused its dividend in 2024 to accelerate deleveraging, embedded ROIC into the compensation plan, and begun opportunistic buybacks. This is a textbook pivot from empire-building to balance-sheet and return discipline.

The honest complication, shared with SBAC, is near-term per-share growth. AFFO/share has been roughly flat at ~$10.5–10.9 across FY2023–FY2026E, weighed down by India-exit dilution, FX translation (a weak Brazilian real, Nigerian naira, Mexican peso), the January-2026 DISH default (~400bps of 2026 AFFO churn), and a refinancing step-up as low-coupon notes roll. Underlying organic growth is healthy and bifurcated — US ~5% ex-DISH, Europe ~4–5%, Africa ~11% (FX-eroded), LatAm negative on Brazil churn, and CoreSite a standout ~13–17% — and management guides AFFO/share re-acceleration to mid-to-upper-single digits in 2027 as India/DISH lap, Brazil churn normalizes, and CoreSite compounds.

Valuation reflects all of this as roughly fair. At ~$187 — down sharply from a ~$292 peak in 2021 — AMT trades at ~17x forward AFFO, ~18.7x EV/EBITDA, and a ~3.8% dividend yield, sitting at the 44th percentile of its own ten-year valuation history (P/S 45th, P/E noise, P/B elevated at the 82nd on a small positive book). That is materially cheaper than CCI (~21x AFFO, a premium) and modestly richer than SBAC (~17x but at the 7th percentile of its own history). GAAP EPS (~$5.40) is distorted by REIT depreciation and India discontinued-operations effects; AFFO and EV/EBITDA are the right lenses.

The investment question is therefore one of price for quality, not business quality: AMT is unambiguously the highest-quality, most-diversified, most-derisked, best-financed of the three tower REITs, with a unique data-center/AI option — but it is priced as such, with flat near-term per-share growth, so the return is contingent on the 2027 re-acceleration and modest rather than asymmetric.

2. Business Overview

What AMT is. American Tower is a global REIT that owns and operates communications real estate — principally macro towers, plus distributed antenna systems and, through its CoreSite subsidiary, data centers. It leases vertical space and colocation capacity to wireless carriers, broadcasters, governments, and enterprises under long-term contracts. Founded in 1995 and based in Boston, it is the largest independent tower operator in the world by site count and revenue.

Segment and geographic mix (FY2025). The business is reported across four geographies plus data centers and services:

The crucial structural point: US & Canada is ~51% of property revenue and the hard-currency profit core, with Europe adding developed-market stability, Africa/LatAm adding (volatile, FX-exposed) growth, and CoreSite adding a structurally different, double-digit-growth data-center leg. This is the widest diversification in the tower group — CCI is 100% US; SBAC is US-plus-Brazil-heavy; AMT spans four continents plus data centers.

The tenant base. Concentration is the lowest of the three: T-Mobile ~18%, AT&T ~17%, Verizon ~14%, Telefónica ~10% — top four ~59% of total revenue, with the balance spread across international carriers (América Móvil/Claro, MTN, Airtel, Orange, TIM/Vivo) and CoreSite’s cloud/enterprise customers (AWS, Azure, etc.). Critically, the India exit removed the Vodafone Idea credit overhang that had plagued AMT’s largest-by-count market for years.

The lease model. Towers are leased on 5–10-year initial terms with multiple renewals, ~3% fixed escalators in the US and CPI-linked escalators internationally, and ~98% retention — the same site-monopoly annuity that defines the sector. CoreSite operates on the data-center model: multi-year colocation leases plus high-margin, sticky interconnection revenue (cross-connects between tenants), which carries genuine network-effect characteristics — the more networks and clouds present in a facility, the more valuable it is to the next tenant.

The annuity’s durability. AMT’s leases carry the bond-like characteristics that define the sector: 5–10-year initial terms, multiple tenant renewal options, contractual escalators (~3% fixed in the US, CPI-linked internationally), narrow termination rights, and ~98% retention. The contracted, non-cancellable future tenant-billings backlog runs to tens of billions of dollars with a multi-year weighted-average life, underwriting the bulk of forward revenue. As with peers, the practical churn AMT experiences is rarely competitive defection (almost never economic for a carrier) but episodic decommissioning after carrier mergers — Sprint in the US, Oi in Brazil, the DISH default — which is identifiable and lumpy rather than chronic. The diversification across ~150,000 sites and dozens of carriers means no single decommissioning event, however painful in a given year, threatens the annuity as a whole — a structural resilience CCI’s ~90%-concentrated US-only book lacks.

CoreSite — the differentiator. Acquired in 2021 for ~$10.1B, CoreSite is a carrier-neutral, interconnection-rich data-center platform concentrated in major US metros. It is small relative to pure-plays like Equinix or Digital Realty, but it is the highest-growth and arguably highest-return leg of AMT, growing ~13–17% organically, adding ~200MW of development capacity, and positioned as an AI-inference and hybrid-multi-cloud on-ramp — the “edge” of the AI buildout where low-latency inference and enterprise cloud connectivity live, distinct from the hyperscale training campuses. It is the single feature that makes AMT more than a tower company.

The co-location flywheel. AMT’s economics, like the sector’s, are driven by adding tenants to existing towers. The first (anchor) tenant covers the largely-fixed ground rent, maintenance, and tax cost and earns a modest return; the second and third tenants’ rent is almost pure incremental margin, lifting site-level returns from high single digits toward the 20s%. AMT’s ~65.5% EBITDA margin is the financial signature of this leverage across a ~150,000-site base. The same mechanism runs in reverse on churn — which is why the DISH default and Brazil/Oi decommissioning bite EBITDA disproportionately — but AMT’s lower customer concentration (top-4 ~59%) and geographic spread blunt the impact of any single tenant or market relative to CCI’s ~90%-concentrated US-only book.

Why scale and diversification matter here. A ~150,000-site global footprint spanning the US, Europe, Africa, and Latin America gives AMT three structural advantages a single-market operator lacks: procurement and operating scale (the largest land, power, and maintenance buyer in the industry), relationship depth with multinational carriers (Telefónica, Orange, MTN, América Móvil span many of AMT’s markets), and diversification of demand and currency exposure so that a downturn in one market (Brazil) is partly offset by growth in another (Africa, Europe, CoreSite). The cost of that breadth is operational complexity and FX translation noise — the tension that runs through the whole AMT story.

Corporate form. AMT is a REIT; GAAP EPS (~$5.40 in 2025) is heavily distorted by depreciation and by India discontinued-operations/CTA effects, so the sector metrics — AFFO/share and EV/EBITDA — are the right lenses. Unlike CCI and SBAC, AMT carries positive book equity (~$3.65B attributable, ~$10.4B including ~$6.7B of non-controlling interests), so P/B is at least defined (if elevated and not very meaningful given the M&A-built intangibles).

Verdict. The most diversified, largest-scale communications-real-estate business in the world, with a hard-currency developed-market core, emerging-market growth optionality, and a genuinely differentiated data-center/interconnection engine in CoreSite. The breadth is both the strength (diversification, optionality) and the source of complexity (FX, EM operations, M&A-built intangibles). The asset quality is top-tier.

3. Industry Dynamics

The US core — the shared oligopoly. AMT’s US & Canada segment sits in the same structurally attractive three-player macro-tower oligopoly as CCI and SBAC: three dominant tower owners, investment-grade carrier customers, high barriers to entry (zoning, site scarcity, anchor-tenant economics), near-zero new-supply risk, ~98% retention, and ~3% escalators. As across the sector, the US is now a stable three-network market (Verizon, AT&T, T-Mobile) after DISH/EchoStar’s greenfield build collapsed into a January 2026 default, and carrier capex has normalized off the 5G mid-band peak, decelerating organic leasing toward the low-to-mid single digits pending the next spectrum/6G cycle (~2027+).

The international dimension — broader and more developed-market-weighted than SBAC. Where AMT’s industry exposure differs from both peers is the breadth and quality-mix of its international footprint:

• Europe (~32,500 sites) is the standout differentiator versus SBAC’s Brazil-heavy book — a developed-market, harder-currency (EUR), structurally stable geography with a long Telefónica build-to-suit pipeline running ahead of underwriting and ~4–5% organic growth. This is higher-quality international exposure than emerging-market towers.
• Africa & APAC (~27,900 sites) is the highest-growth (~11% organic) but highest-risk leg — fuel/power cost inflation, currency volatility (Nigerian naira, Ghanaian cedi), and carrier-credit considerations.
• Latin America (~47,100 sites) is the current pain point — Brazil’s Oi unwind driving elevated churn and a weak real/peso eroding reported dollar growth (LatAm organic was negative in 2026 guidance).

Importantly, AMT has made a deliberate strategic decision to shrink emerging markets as a share of the portfolio — directing ~85% of 2026 discretionary capital to developed markets and CoreSite — explicitly to reduce earnings volatility. The India exit was the largest expression of that strategy.

The data-center/interconnection industry. Through CoreSite, AMT also participates in the retail, interconnection-rich data-center market, competing with Equinix and Digital Realty. That industry is structurally attractive — secular cloud/AI demand, network-effect economics in interconnection, high switching costs — and gives AMT a growth vector and an AI-adjacency the pure towers lack. (Equinix is the premium pure-play; CoreSite is the smaller, higher-return, interconnection-curated cousin.) The key industry distinction worth drawing for an AMT investor is which layer of the AI buildout CoreSite occupies: it is not a hyperscale-training-campus play (those multi-hundred-megawatt, single-tenant builds in cheap-power exurbs largely bypass interconnection-rich metro facilities), but rather the inference, edge, and hybrid-multi-cloud on-ramp layer — the metro interconnection points where enterprise cloud traffic, low-latency inference, and network peering converge. That layer arrives later and grows more steadily than the training-capex spike, and it is defended by the same interconnection network effect that anchors Equinix. It is a smaller, higher-quality, more durable AI exposure than a hyperscale landlord — and it is the single feature that lets AMT credibly claim an AI growth vector that CCI and SBAC simply do not have.

The demand drivers. AMT’s long-term demand rides the same secular forces as the sector — compounding mobile data traffic, network densification (more antennas, bands, and amendments per tower), fixed-wireless access, and the next spectrum/6G deployment cycle — but across a far wider geographic base. In the US, the next leasing leg depends on C-band continuation and the planned upper-C-band/AWS auctions (~mid-2027, deploying ~2029–30) plus eventual 6G. Internationally, the runway is longer and earlier-stage: Europe is mid-5G, Africa is still densifying 4G/5G off a low tower-density base, and Latin America is rebuilding post-Oi. Layered on top, CoreSite adds an entirely different demand vector — cloud, hybrid-multi-cloud, and AI inference — uncorrelated with the carrier capex cycle. This breadth of demand drivers is a genuine diversification of growth sources that neither single-market peer enjoys: AMT is never wholly dependent on the US carrier cycle in any given year.

Capital cycle (Marathon lens). Towers themselves show no new-supply response to high returns (the sector’s saving grace), so the down-leg shows up as slower growth rather than impaired economics. Where capital did flood was tower M&A multiples at the 2020–21 low-rate peak — and AMT was the most aggressive buyer of all three (Telxius + CoreSite), which is precisely why its subsequent repair (deleveraging, India exit) has been the most pronounced. The industry is now in its quiet, higher-cost phase.

Verdict: a structurally good industry, and AMT has the widest and most developed-market-weighted exposure within it. The diversification (developed + emerging + data centers) is a genuine quality advantage over SBAC’s Brazil concentration and CCI’s US-only profile, at the cost of operational complexity and FX translation noise. The deliberate tilt back toward developed markets and CoreSite is a sensible response to the EM volatility that has dogged reported results.

4. Competitive Position

The moat, named. AMT’s moat is the sector-standard local economies of scale plus customer captivity (switching costs) — each tower a local quasi-monopoly, with relocation slow, costly, and coverage-risking, reinforced by long contracts and escalators. It passes Greenwald’s diagnostics (stable market share, persistently high site-level returns) and is visible in the ~65.5% EBITDA margin and ~98% retention.

AMT’s specific edges. Within the shared moat, AMT’s distinctive advantages are:

1. Scale — the largest global tower portfolio, with attendant procurement, land, and operating scale, and the deepest relationships with multinational carriers.
2. Diversification — the only one of the three with meaningful developed-market international exposure (Europe) plus a data-center business, lowering single-market/single-currency risk.
3. CoreSite’s interconnection network effect — a genuine, if smaller, second moat: the vast majority of CoreSite revenue is with customers interconnected to five or more others, the textbook stickiness that makes an interconnection ecosystem hard to replicate (the same mechanism that anchors Equinix).
4. The India-exit derisking — removing ~76,000 low-margin, churn- and credit-impaired sites and the Vodafone Idea overhang structurally improved the quality of the remaining portfolio.
5. The best balance sheet — investment-grade, lowest-levered, best-covered, which lowers cost of capital and funding risk versus peers.

Why ROIC trails SBAC — a denominator story, not an operating one. AMT’s ROIC of ~7.85% (2025, recovered from ~5.7% in 2023) sits between CCI (~7%) and SBAC (~12.4%) — and the gap to SBAC is almost entirely a function of the price AMT paid for growth, not weaker per-tower economics. AMT carries ~$12.3B of goodwill plus ~$14.5B of other intangibles (~$27B, ~42% of assets) from Telxius, CoreSite, and EM acquisitions; that inflated invested-capital base mechanically depresses ROIC. EBITDA margins are essentially identical across the three (~65%); SBAC simply grew more organically and cheaply and runs leaner SG&A. AMT’s returns reflect what it paid, not how the assets perform.

The goodwill/ROIC mechanics, worked through. It is worth being precise about why the best-quality franchise posts a middling ROIC, because it is the crux of the AMT-versus-SBAC quality debate. ROIC is roughly NOPAT divided by invested capital. AMT’s NOPAT margin and per-tower cash economics are essentially identical to SBAC’s (~65% EBITDA margins at both). The difference is the denominator: AMT’s invested capital includes ~$12.3B of goodwill and ~$14.5B of other intangibles — ~$27B, roughly 42% of total assets — booked when it paid premium prices for Telxius (~$9.4B) and CoreSite (~$10.1B) at the 2021 cycle top. That ~$27B sits in the ROIC denominator earning the same operating cash flow it would have without the premium, so the return on the price paid is lower even though the return on the underlying assets is not. SBAC, which grew more organically and bought at lower multiples, carries far less goodwill and so converts the same per-tower economics into a ~12% ROIC. The practical implication: AMT’s ~7.85% ROIC is a verdict on its 2021 prices, not its 2025 operations — and it is rising (from ~4.6% in 2022) as the acquisitions season and EM stabilizes. Putting ROIC into the comp plan is management’s explicit acknowledgment of exactly this.

Pressure-test. The moat is real and the diversification genuine, but two honest caveats: (1) the M&A-built balance sheet means AMT’s returns on invested capital will structurally lag a more organically-grown operator like SBAC until the goodwill is earned through; and (2) the emerging-market towers (Africa, LatAm) are genuinely lower-quality than the developed-market core — more FX, more churn, more operational risk — which is why management is deliberately shrinking their share.

The Greenwald and Marathon tests. On Greenwald’s diagnostics, towers pass emphatically on market-share stability — sites rarely change operator hands and tenants almost never relocate — and on persistently high site-level returns. AMT passes the share-stability test as clearly as its peers; where it scores lower is the corporate-return test, for the goodwill reason above. On the Marathon capital-cycle lens, AMT is the most instructive of the three: it was the most aggressive deployer of capital at the cycle top (Telxius + CoreSite, ~$19.5B in 2021 when money was free and tower multiples peaked), and it is now living through the predictable mean-reversion — impairments, FX pain, a de-rate from ~$292 to ~$187, and a multi-year repair. The Marathon discipline AMT lacked in 2021 it is now exercising in 2024–2026 (pruning, deleveraging, ROIC-linked pay). For a forward-looking investor, the relevant point is that the cost of the cycle-top binge is now largely in the price and the balance sheet, and the company is being run with far more capital discipline than it was four years ago — the cycle has turned and management has turned with it.

Verdict: a real, durable, top-tier moat — the most diversified and most derisked of the three — with the best balance sheet and a unique data-center option. The asset quality is first-rate; the ROIC simply reflects the premium prices paid to assemble the empire, and the recent strategy is precisely about improving returns (ROIC now in the comp plan, EM being shrunk, leverage coming down). It is the highest-quality franchise of the three even if SBAC is the highest-return operator.

5. Growth History and Forward Opportunities

The flat-AFFO reality. The central near-term fact about AMT, as with SBAC, is that AFFO/share has been essentially flat for three years: ~$10.06 (FY2023) → ~$10.54 (FY2024) → ~$10.76 (FY2025) → ~$10.78–10.95 guided (~$10.87 midpoint, FY2026E). That is roughly 1–3% annual growth, well below the high-single-to-low-double-digit compounding AMT delivered in the 2010s. This must be confronted directly: AMT is not currently compounding AFFO/share at an attractive rate.

Why — identifiable, largely transitory headwinds. The flatness is the sum of several one-off and cyclical drags: (1) India-exit dilution (lost EBITDA from the divested sites); (2) FX translation (weak Brazilian real, Nigerian naira, Mexican peso eroding reported dollar growth); (3) the DISH/EchoStar default (January 2026 — 100% of DISH revenue moved to churn, a ~400bps 2026 AFFO headwind); (4) a refinancing step-up as low-coupon notes roll into higher rates (~100bps headwind); and (5) services-mix noise. Management’s framing: ex-FX and ex-DISH, AFFO/share grew ~4% in Q1-2026, and FY2025’s reported ~8% as-adjusted growth was the “clean” underlying rate.

The organic picture — healthy and bifurcated. Underneath the flat headline, organic tower billings growth is solid and varied by region: US & Canada ~5% ex-DISH (~1% reported, dragged by DISH churn), Europe ~4%, Africa & APAC ~11% (FX-eroded), LatAm negative (~−2%, Brazil Oi churn), and CoreSite ~13–17% — the clear bright spot, driven by hybrid/multi-cloud, AI inference, and a Q1-2026 “inflection in interconnection activity.” Consolidated organic was ~1% reported (~4% ex-DISH) in 2026 guidance.

The organic detail, region by region. The bifurcation is worth laying out because it is the whole near-term story. US & Canada organic was ~1% reported in early 2026 but ~5% excluding the DISH churn — i.e. the underlying domestic engine is healthy and the headline is a one-off. Europe ran ~4% with the Telefónica build-to-suit pipeline adding sites ahead of underwriting — developed-market, EUR-denominated, the highest-quality international growth in the group. Africa & APAC grew ~11% organically — the fastest, but partly eroded by naira/cedi weakness and carrying the highest operational risk (power costs, currency controls). Latin America was negative (~−2%) as Brazil’s Oi-driven churn (~8% LatAm churn in 2026) and a weak real overwhelmed escalators — the single biggest drag, and the one management says peaks in 2026. Netting these, consolidated organic was ~1% reported / ~4% ex-DISH. The pattern is identical in shape to SBAC’s: a healthy underlying business whose reported number is depressed by a stack of identifiable, largely-transitory items (DISH, Brazil churn, FX).

CoreSite and the AI optionality. CoreSite is the growth and optionality story that differentiates AMT. Growing ~13–17%, adding ~200MW of capacity, with ~$700M+ of success-based development capex planned for 2026, it is positioned at the AI-inference/edge and hybrid-cloud on-ramp layer. This is not a hyperscale-training play (those campuses bypass interconnection-rich metro facilities), but it is a genuine, compounding, AI-adjacent growth engine that neither CCI nor SBAC possesses — and it is the most credible source of above-tower growth for AMT over the next decade.

New builds and capital deployment. AMT builds selectively — ~700+ European sites planned for 2026 (Telefónica build-to-suit plus opportunistic), at WACC+200bps initial returns scaling to teens as tenancy seasons, with ~85% of discretionary capital directed to developed markets and CoreSite. It builds far less in the US than SBAC does internationally, reflecting its developed-market, lower-risk tilt.

The re-acceleration thesis. Management guides AFFO/share growth to resume at mid-to-upper-single digits in 2027 as: India and DISH lap out of the comparisons, Brazil churn (pulled forward into 2026) normalizes with positive LatAm growth returning in 2027–2028, Europe builds season, CoreSite compounds, and the refinancing headwind stabilizes. The thesis is plausible and partly mechanical (the lapping of one-offs), but, as with SBAC, it is guided rather than delivered.

Verdict: bifurcated growth quality, flat near-term per-share, with a genuine 2027 re-acceleration case and a differentiated CoreSite engine. The developed-market organic (US ex-DISH ~5%, Europe ~4–5%) plus double-digit CoreSite is high-quality; the EM leg is higher-growth but volatile and FX-eroded. Reported AFFO/share is flat on identifiable one-offs. Unlike SBAC, AMT has the lowest leverage and the CoreSite/AI optionality to fund and supplement the recovery — but the per-share growth case still rests on the 2027 inflection materializing.

6. Financial Quality

Read AFFO and EV/EBITDA. GAAP EPS (~$5.40 in 2025) is distorted by REIT depreciation and India discontinued-operations/CTA effects and should be set aside. The metrics that matter: AFFO/share ~$10.76 (FY2025), ~$10.87 guided (FY2026E); adjusted EBITDA ~$7.0B at a ~65.5% margin. At ~$187 that is ~17x AFFO and ~18.7x EV/EBITDA (on a correctly-computed EV of ~$131B — note that ROIC’s headline EV of ~$50B is erroneous, undercounting debt; the real EV is market cap ~$87B + net debt ~$43.5B).

Margins and returns. EBITDA margin ~65.5% — in line with the sector’s best, confirming the per-tower economics are top-tier. ROIC ~7.85% (recovered from ~4.6% in 2022) trails SBAC’s ~12.4% purely on the ~$27B goodwill/intangible base, not on operating performance. FCF is strong: OCF ~$5.46B less ~$1.68B capex = ~$3.78B, an ~8.7% FCF yield.

The AFFO bridge and what it includes. The bridge from GAAP runs: net income + real-estate depreciation and amortization (~$2.0B) + stock-based comp (~$174M) + non-cash and one-time items (India CTA, impairments, deferred tax) − sustaining capex ± straight-line/FX adjustments → AFFO of ~$5.0B, or ~$10.76/share attributable in FY2025. Two honest quality notes. First, AFFO adds back SBC (~$174M) — a real economic cost — as the whole sector does; netting it would trim AFFO/share by ~$0.35. Second, AMT’s AFFO is attributable (it strips the ~$6.7B of non-controlling interests’ share of certain consolidated subsidiaries, principally in Africa and the data-center/JV structures), which is the correct, conservative treatment — the headline AFFO already excludes minority partners’ claims. Net of these, AMT’s ~$10.76 AFFO/share is a clean, conservatively-defined distributable-cash figure, and the ~63% payout against it leaves genuine cushion — the opposite of CCI’s ~90% strained payout.

Why the de-rate happened — and what it implies. AMT’s multiple compressed from ~21x P/FCF (2021) to ~11.5x (2025), and the stock from ~$292 to ~$187, for identifiable reasons: rising rates (REITs are rate-sensitive bond-proxies), the leverage overhang from the 2021 binge, the India/EM drag, and the flat AFFO/share. Most of those are now either resolved (leverage down to ~5.0x, India exited) or in the price (the flat AFFO is guided to inflect in 2027). The de-rate is therefore as much an opportunity as a warning — a higher-quality, lower-leverage AMT than 2021’s now trades at ~60% of its prior multiple — provided the AFFO re-acceleration materializes. The valuation section weighs whether the fair-not-cheap multiple adequately discounts that.

The balance sheet — the best in the group. This is where AMT’s quality is clearest:

• Net leverage ~5.0x (AMT’s reported metric; ~6.45x gross debt/EBITDA including ~$7.8B of finance leases), down from ~9.2x at the 2021 peak — the lowest of the three (SBAC ~6.6x, CCI ~8.7x pre-paydown), and the highest credit rating.
• Interest coverage ~5.1x EBITDA/interest — the best of the three (SBAC ~3.87x, CCI ~2.95x).
• ~97% fixed-rate debt — a +10% move in rates changes interest expense by only ~$6.7M — so AMT is far less exposed to floating rates than the headline leverage suggests.
• Fully unsecured, investment-grade senior notes (15+ series, coupons ~0.40–4.625%, maturities 2026–34) plus ~$10B of revolvers and a term loan — AMT is repaying its legacy secured Tower Revenue Notes and moving entirely to an unsecured IG structure, a contrast with SBAC’s securitized ABS model and a sign of balance-sheet maturity. Total liquidity ~$11B.
• Positive book equity (~$3.65B attributable) — unlike the negative equity at both peers (a buyback/impairment artifact there) — though P/B (~24x) is not very meaningful given the M&A-built intangibles.

The refinancing exposure — real but well-managed. Like the rest of the sector, AMT must refinance low-coupon legacy notes into a higher-rate market, a ~100bps AFFO headwind. But its ~97%-fixed, laddered, investment-grade, fully-unsecured structure and ~5.1x coverage make it the best-positioned of the three to absorb that step-up — the opposite end of the spectrum from CCI’s thin ~2.95x coverage.

The unsecured-IG model versus SBAC’s securitization. AMT’s debt architecture is worth contrasting with SBAC’s because it reflects a different philosophy. SBAC finances primarily through securitized Tower Revenue Notes (ABS) backed by tower cash flows — historically the cheapest funding, but secured and structurally senior. AMT funds through a deep stack of senior unsecured investment-grade notes (15+ series, coupons ~0.40–4.625%, maturities laddered 2026–2034) plus large revolvers and a term loan, and is actively repaying its legacy secured notes to move fully unsecured. The unsecured-IG route costs marginally more in normal times but buys flexibility, simplicity, and resilience — no collateral encumbrance, broad market access, and a single consolidated credit. With ~97% of debt fixed-rate, a +10% move in rates changes interest expense by only ~$6.7M, so AMT is far less exposed to floating rates than its ~$45B gross debt suggests. The refinancing step-up (rolling sub-2% pandemic-era notes into higher coupons, ~100bps of AFFO headwind) is real but the most manageable in the group given the ~5.1x coverage and laddered maturities. This is the balance sheet of a company that has chosen durability over the last basis point of funding cost — appropriate for the blue-chip of the sector.

The payout — conservative, and better than the peer tables suggest. A note of correction: AMT’s AFFO payout is ~63% ($6.77 dividend / ~$10.76 AFFO), not the ~93% figure that circulates in some peer comparisons (which conflates a GAAP-EPS-based payout). That ~63% is the most conservative of the three (SBAC ~41%, CCI ~90%) and is what funds AMT’s simultaneous deleveraging, dividend, and newly-initiated buybacks. The GAAP payout exceeds 100%, but that is the usual REIT depreciation artifact, not a coverage problem.

Verdict: the highest financial quality of the three. Top-tier margins, the lowest leverage, the best coverage, ~97% fixed-rate unsecured IG debt, positive book equity, a conservative ~63% payout, and strong FCF. The only blemish is the ROIC drag from the M&A-built intangible base — a price-paid issue, not an operating one. Economics improve with scale here; AMT simply paid up to build the scale, and is now earning it through.

7. Capital Allocation

The two-act story: the binge, then the repair. AMT’s capital-allocation record is genuinely two distinct chapters, and the verdict depends on weighing them.

Act one — the 2021 binge. At the cycle top, with rates near zero, AMT deployed ~$19.5B in a single year: Telxius (~$9.4B, ~31,000 European/LatAm sites) and CoreSite (~$10.1B, ~$170/share). These drove goodwill and intangibles to ~$27B, pushed net leverage to ~9.2x, and were followed by ~$684M (2022) and ~$740M+ (2023) of impairments. In hindsight this was aggressive empire-building at peak multiples with peak-cheap debt — the Marathon cautionary pattern. CoreSite has proven the better of the two (double-digit growth, AI/cloud optionality); the EM-heavy Telxius and the broader EM portfolio have been the source of much of the subsequent FX/churn drag.

Act two — the repair (2023–2025, under CEO Vondran). The pivot has been disciplined and shareholder-friendly:

• India exit (September 2024): ~$2.5B (including ~$320M of Vodafone Idea debentures) to a Brookfield-led consortium, proceeds to debt repayment — removing ~76,000 low-margin sites and the VIL credit overhang, and structurally derisking the portfolio. AMT also exited Australia/New Zealand and South Africa fiber. The strategic logic was sound even at a seemingly low headline price: India was AMT’s largest market by site count but a chronic drag on quality — years of carrier consolidation (the collapse from a dozen operators to a troubled three), brutal price competition, the Vodafone Idea solvency overhang that forced AMT to take reserves and accept payment in VIL securities, and structurally low margins and high churn. Exiting converted a low-return, high-volatility, capital-absorbing market into ~$2.5B of debt paydown and a measurably higher-quality residual portfolio. The book loss on exit (~$978M in 2024 discontinued operations) was the accounting recognition of value already impaired; the cash and the derisking were the real economics. This is the kind of disciplined pruning — selling a sub-scale, low-return position rather than defending it — that distinguishes the current regime from the 2021 acquire-everything posture.
• Dividend reset and pause: after a decade of ~20%/yr dividend growth, AMT held the dividend flat through 2024 (and actually reset the quarterly rate down from ~$1.70 to ~$1.62) to accelerate deleveraging — a notable, disciplined break from a long streak, prioritizing the balance sheet over the dividend narrative. It resumed ~5% growth in 2025 (now ~$1.79/quarter, ~$7.16 annualized, ~3.8% yield).
• Deleveraging: from ~9.2x to ~5.0x net, with disciplined ~$1.68B/yr growth capex tilted to developed markets and CoreSite.
• Buybacks initiated: ~$560M+ across Q4-2025/Q1-2026 (the largest repurchase activity since 2017), a genuine new lever in the capital-allocation framework that the prior, leverage-constrained AMT could not pull.

The capital-return mix versus peers. AMT’s capital-allocation posture now sits between its two peers in an instructive way. SBAC runs the lowest payout (~41%) and the most aggressive buyback, shrinking its share count — the per-share-compounding model. CCI runs the highest payout (~90%), having just cut and rebased, with little buyback capacity. AMT, at a ~63% payout, occupies the middle: enough retained cash to both delever and pay a growing dividend and now initiate buybacks (~$560M across Q4-2025/Q1-2026, the most since 2017), without straining the balance sheet. The newly-initiated repurchase is the tell that management views the balance-sheet repair as substantially complete — the prior, 9.2x-levered AMT could not have bought back stock. It is a more balanced (if less per-share-aggressive) capital-return model than SBAC’s, appropriate for the larger, more diversified, lower-risk franchise, and a world away from CCI’s post-crisis constraint.

Compensation — return-aware, and improving. The 2026 proxy shows an annual incentive on Total Property Revenue (30%) and Adjusted EBITDA (50%), and — the important part — long-term PSUs weighted 40% to cumulative attributable AFFO/share, 30% to average ROIC, and 30% to relative TSR. Putting ROIC explicitly in the long-term plan directly addresses the return-on-capital concern that the M&A binge created — management is now paid to earn the goodwill through, not to build more scale. Insider ownership is <1%.

Verdict: a nuanced, improving record — aggressive scale-building that overreached, now followed by genuine, disciplined repair. Be balanced: the 2021 binge loaded the balance sheet with goodwill and leverage at the worst possible time, and the EM acquisitions have been a persistent drag. But the Vondran-era pivot — India exit, dividend discipline, deleveraging to the best balance sheet in the group, ROIC in the comp plan, and buybacks initiated — is exactly the rational capital allocation an owner would want, and it is showing in the recovering ROIC and the falling leverage. This is not the unbroken excellence of SBAC, nor the value-destruction of CCI’s fiber saga; it is a high-quality franchise that overpaid, recognized it, and is fixing it.

8. Changes and Headwinds — Last Two Years

• CEO transition (Feb 2024). Steven Vondran (previously head of CoreSite and the US tower business) succeeded long-time CEO Tom Bartlett; Rod Smith continues as CFO. Internal, orderly — and notably installs a CoreSite/data-center-experienced leader, signaling the strategic emphasis.
• The India exit (Sept 2024). ~$2.5B sale of ~76,000 sites to a Brookfield-led consortium — the single biggest derisking event, removing a chronically troubled, low-margin, Vodafone-Idea-credit-impaired market. Reported in discontinued operations; drove a ~−$978M disc-ops loss/impairment in 2024 but cleaned the portfolio.
• Dividend reset/pause (2024). Held flat (reset down from ~$1.70 to ~$1.62/quarter) to accelerate deleveraging — a disciplined break from a decade of ~20%/yr growth; resumed ~5% growth in 2025–2026.
• Deleveraging to ~5.0x from the ~9.2x 2021 peak, and the move to a fully-unsecured investment-grade balance sheet.
• Buybacks initiated (Q4-2025/Q1-2026, ~$560M+). A new capital-return lever, signaling balance-sheet confidence.
• The DISH/EchoStar default (Jan 2026). ~400bps of 2026 AFFO churn as DISH revenue moves to churn; litigation overhang. A sector-wide event (it hit all three).
• Portfolio pruning. Exited Australia/New Zealand (2024) and South Africa fiber (2025) — continued focus on developed markets + CoreSite.
• The flat-AFFO air-pocket (2023–2026). The cumulative effect of India, FX, DISH, and refinancing — the single biggest near-term headwind, with a guided 2027 re-acceleration.

Verdict: net strongly stabilizing. The changes — India exit, dividend discipline, deleveraging, ROIC-linked comp, buybacks — collectively represent the most constructive capital-allocation and balance-sheet repair in the group. The headwinds (flat AFFO, FX, DISH) are real but identifiable and largely transitory. AMT enters 2026 as a cleaner, safer, better-financed business than it was in 2021, managing through a cyclical/FX/rate air-pocket from a position of balance-sheet strength.

9. Risk Analysis

The dominant risks are EM/FX translation and the failure of AFFO/share to re-accelerate past the 2026 trough — the two together determine whether the flat-AFFO period is an air-pocket or a plateau. Unlike CCI and SBAC, balance-sheet/rate risk is comparatively low for AMT (best coverage, ~97% fixed, lowest leverage), which is the core of its blue-chip character. The catastrophic-loss probability is very low (diversified, irreplaceable assets, IG balance sheet); the disappointing-total-return probability — if FX and EM stay weak and the re-acceleration slips — is the real risk, mitigated by the fair (not stretched) valuation and the CoreSite optionality.

How the risks interact. AMT’s risk chain is shorter and better-buffered than its peers’. The primary channel runs EM/FX → reported AFFO/share → multiple. A weak real/naira/peso lowers reported dollar AFFO and stalls the per-share line; a stalled per-share line caps the return at the dividend yield and risks the multiple staying mid-range rather than re-rating. But — and this is the blue-chip distinction — the balance-sheet link that makes CCI and SBAC fragile is largely absent at AMT: with ~97% fixed-rate debt, ~5.1x coverage, the lowest leverage in the group, and a conservative ~63% payout, a cash-flow or rate disappointment does not threaten the dividend or force deleveraging at a bad time. So AMT’s realistic bear case is “fairly-valued, low-growth, FX-buffeted dead money,” not “distress.” The diversification further mutes any single shock — a Brazil downturn is offset by Africa/Europe/CoreSite growth in a way CCI (US-only, ~90% three-customer) and SBAC (Brazil-heavy) cannot match. The flip side is that diversification also dilutes the upside: AMT will rarely be the cheapest or fastest-growing, by construction. The risk profile is the lowest of the three, and the return profile is correspondingly the most moderate.

The EM/FX point deserves emphasis because it is the recurring, mechanical drag on AMT’s reported results: roughly a third of revenue is non-US, much of it in volatile-currency emerging markets, so a strong dollar persistently erodes reported growth even when local-currency organic growth is healthy. Management’s deliberate tilt back toward developed markets and CoreSite (~85% of 2026 discretionary capital) is precisely an effort to shrink this drag over time — but it is a multi-year reshaping, and in the interim FX remains the single most common reason AMT’s reported AFFO/share disappoints relative to its underlying operating performance.

10. Valuation Discussion (Embedded Expectations)

Where it trades (2026-06-13, ~$187.18). Market cap ~$87B (~466M shares); enterprise value ~$131B on ~$43.5B net debt (note: ROIC’s ~$50B EV is erroneous — it undercounts debt). On REIT metrics: ~17x forward AFFO (~$10.87 2026E), ~18.7x EV/EBITDA, ~3.8% dividend yield. The stock has de-rated sharply from a ~$292 peak (2021) and a ~21x P/FCF then to ~11.5x P/FCF now. On its own ten-year history, AMT sits at the 44th percentile of its composite valuation (P/S 45th, P/E noise, P/B elevated at the 82nd on a small positive book) — fairly valued, mid-range, neither the bargain SBAC is nor the premium CCI carries. GAAP P/E (~35x) is meaningless.

The peer comparison — AMT is the quality middle.

The table places AMT cleanly as the quality middle: the best balance sheet (lowest leverage, best coverage), the most conservative payout, the most diversification, and the only data-center optionality — trading at the same ~17x AFFO as SBAC but at a higher percentile of its own history (44th vs 7th), i.e. less of a relative-value discount. Versus CCI, AMT is both higher-quality and cheaper. The honest read: you pay a fair price for the best balance sheet and the most optionality, whereas SBAC offers the best returns and the cheaper-vs-history entry, and CCI offers the worst of both.

Embedded expectations / reverse read. At ~$131B EV and ~$7.0B EBITDA growing slowly near-term, with AFFO/share ~flat into 2026, a ~17x AFFO multiple embeds an expectation that 2026 is the trough and AFFO/share re-accelerates to mid-single-digits in 2027 as India/DISH lap, Brazil normalizes, and CoreSite compounds. If that holds, the return math is reasonable: a ~3.8% growing dividend + mid-single-digit AFFO growth + modest buyback accretion ≈ ~8–10% total return, with CoreSite/AI as unpriced optionality and a possible modest re-rating from the 44th percentile. If the air-pocket extends (FX, EM, rates), AFFO/share stays flat and the return collapses to roughly the dividend yield — but the downside is cushioned by the best balance sheet in the group and a valuation already at fair (not stretched) levels.

The total-return decomposition. Strip the narrative and the arithmetic is straightforward. A buyer at ~$187 collects a ~3.8% dividend (growing ~5%/yr). On top of that, AFFO/share is flat in 2026 and guided to mid-single-digit growth from 2027 as one-offs lap and CoreSite compounds — call it ~5% over a multi-year horizon if management delivers, plus a small contribution from the newly-initiated buybacks. Holding the multiple constant, that is a ~3.8% + ~5% ≈ ~9% expected annualized total return, with two free options attached: a possible re-rating from the 44th percentile of its own history, and CoreSite’s AI/interconnection upside that the sum-of-the-parts arguably under-credits. That is a respectable, lower-risk return — but note it is roughly the same base return SBAC offers at a cheaper own-history multiple and a higher ROIC, which is why, on pure expected return, SBAC edges AMT; AMT wins on risk-adjusted return through its superior balance sheet, diversification, and optionality.

The trio choice, made explicit. Having now examined all three, the relative-value picture is clean. SBAC is the highest-return operator at the cheapest own-history multiple, in a cyclical air-pocket — the value pick. AMT is the highest-quality, most-diversified, best-financed franchise with the only AI/data-center option, at a fair price — the quality/safety pick. CCI is the weakest franchise at a premium — the avoid. For an investor who can hold only one and prioritizes sleeping at night — lowest leverage, best coverage, broadest diversification, real optionality — AMT is the defensible default, accepting that you pay fair value for that quality and that SBAC offers more upside for more (FX/leverage) risk. For an investor optimizing expected return and willing to underwrite Brazil and the air-pocket, SBAC. Neither argues for CCI here.

Sum-of-the-parts sanity check. AMT is genuinely two businesses: a ~$9.6B-revenue global tower franchise and a ~$1.05B-revenue, double-digit-growth CoreSite. Pure-play data-center REITs (EQIX, DLR) trade at ~25–29x EV/EBITDA; even a conservative ~20–22x on CoreSite’s EBITDA implies the data-center leg is worth a meaningful and growing slice of AMT’s EV that the blended ~18.7x multiple does not separately credit. As CoreSite compounds at ~13–17% and the AI-interconnection narrative builds, the SOTP gap is a quiet source of upside optionality — not a thesis on its own, but a reason the fair-value-on-blended-multiple read may understate the parts.

Scenarios (return drivers; no price target).

• Bear: the air-pocket extends — EM/FX stays weak, Brazil churn lingers, the 2027 re-acceleration slips — and AFFO/share stays flat. The multiple holds (~17x, already mid-range), so the return is roughly the ~3.8% yield with little growth — dead-money, not a collapse, and the strong balance sheet limits the damage.
• Base: 2026 is the trough; AFFO/share re-accelerates to mid-single-digits in 2027 as one-offs lap and CoreSite compounds; the dividend grows ~5%, buybacks add modestly. Total return ~= ~3.8% yield + ~5–7% AFFO growth ≈ ~9–11%, with optionality on a re-rating.
• Bull: CoreSite’s AI/interconnection inflection proves durable and accelerates, EM/FX turns, AFFO/share re-accelerates faster, and the market re-rates the derisked, lowest-leverage, optionality-rich name back toward the upper half of its history. Low-to-mid-teens total return.

Verdict: fairly valued for genuine quality. AMT is the highest-quality, most-diversified, best-financed, most-optionality tower at ~17x AFFO and the 44th percentile of its own history — a fair price, not a bargain. The distribution is roughly balanced: the best balance sheet and the fair valuation cushion the downside, while the flat near-term AFFO and the mid-range multiple cap the upside short of a true value setup. You are paid fairly to own the best-derisked franchise with the only AI/data-center option; you are not paid a discount.

11. Variant Perception

Consensus. AMT is viewed as the blue-chip, flight-to-quality tower REIT — largest, most diversified, best balance sheet, with the CoreSite data-center/AI angle — that has successfully derisked (India exit, deleveraging) but whose reported AFFO/share growth is temporarily depressed by FX, DISH, and refinancing. The debate is about when per-share growth re-accelerates and whether the CoreSite/AI optionality deserves more credit.

The strongest bull case. The highest-quality, most-diversified tower franchise in the world, freshly derisked (India gone, Vodafone Idea overhang removed), with the best balance sheet (lowest leverage, best coverage, ~97% fixed IG debt), the most conservative payout (room for dividend growth + buybacks), and a unique, double-digit-growth, AI-adjacent data-center engine in CoreSite — trading at ~17x AFFO and the 44th percentile of its own history after a ~36% de-rate from its peak. Own the safest, most-optionality name in a great industry at a fair price and collect a growing ~3.8% yield while the 2027 re-acceleration and CoreSite/AI compound.

The strongest bear case. AFFO/share has been flat for three years and may stay flat if EM/FX and rates don’t cooperate — and you are paying a fair (not cheap) multiple for that, at a higher own-history percentile than SBAC, which offers better returns and a cheaper entry. AMT overpaid massively in 2021 (Telxius/CoreSite), still carries ~$27B of goodwill that depresses ROIC to ~7.85% (well below SBAC’s 12.4%), and remains meaningfully exposed to volatile emerging markets. If the 2027 re-acceleration slips, this is a fairly-valued, low-growth, FX-buffeted compounder — fine, but not compelling versus the cheaper, higher-return SBAC.

The 3–5 assumptions that matter most. (1) Does AFFO/share re-accelerate to mid-single-digits in 2027, or stay flat? (2) Does CoreSite’s AI/interconnection growth prove durable and earn an optionality premium? (3) Do EM currencies (BRL, naira, peso) stabilize? (4) Does the multiple re-rate from the 44th percentile, or is fair value the ceiling? (5) Within the trio, does AMT’s quality-and-safety premium justify owning it over the cheaper SBAC?

What would falsify each side. Bull falsified by: a 2027 AFFO/share guide still flat-to-low-single-digit; CoreSite growth decelerating; EM/FX deteriorating further. Bear falsified by: a confirmed mid-single-digit 2027 AFFO re-acceleration; CoreSite AI-interconnection inflection sustaining; EM/FX stabilizing; or a re-rating toward the upper half of AMT’s history on the flight-to-quality + AI-optionality narrative.

12. Fact vs. Interpretation

13. Open Questions

1. Does AFFO/share re-accelerate in 2027, or does EM/FX keep it flat into a multi-year plateau?
2. How much is CoreSite/AI worth? Does the interconnection/AI-inference inflection sustain double-digit growth and earn an optionality premium the sum-of-parts doesn’t yet reflect?
3. Where do EM currencies go? BRL, naira, and peso are the swing variables on reported growth.
4. Does AMT keep shrinking EM, and would a further developed-market tilt (or more buybacks) improve the quality/return profile?
5. Does the goodwill get earned through — i.e., does ROIC keep climbing toward double digits as the M&A seasons and EM stabilizes?
6. Within the trio, is the quality/safety premium worth paying over the cheaper, higher-return SBAC?

14. What Must Be True

Bull case — what must be true:

• 2026 is the trough; AFFO/share re-accelerates to mid-to-upper-single digits in 2027 as India/DISH lap, Brazil churn normalizes, and CoreSite compounds.
• CoreSite’s AI/interconnection growth proves durable and earns an optionality premium.
• EM currencies stabilize; the developed-market tilt + deleveraging continue to improve quality and ROIC.
• The market re-rates the derisked, lowest-leverage, optionality-rich name from the 44th percentile.
• Falsification test: a 2027 AFFO/share guide still flat-to-low-single-digit, or CoreSite growth decelerating, would break the bull case.

Bear case — what must be true:

• AFFO/share stays flat as EM/FX and rates persist; the 2027 re-acceleration slips.
• The fair (not cheap) multiple offers no re-rating cushion; SBAC’s cheaper, higher-return profile is the better relative bet.
• The ~$27B goodwill keeps ROIC structurally below the best operator, and EM volatility recurs.
• Falsification test: a confirmed mid-single-digit 2027 AFFO re-acceleration, a sustained CoreSite AI inflection, or EM/FX stabilization would break the bear case.

15. Source Appendix

See AMT_source_appendix.md for the full source list. Primary sources: American Tower FY2025 Form 10-K and Q1-2026 Form 10-Q (SEC EDGAR, CIK 0001053507); 2026 DEF 14A proxy; Q4-2025 and Q1-2026 earnings releases and 2026 guidance (8-K exhibits); Q1-2026, Q4-2025, and Q2-2025 earnings-call transcripts; the India divestiture disclosures (Sept 2024); the Telxius and CoreSite acquisition history; Form 4 insider-transaction corpus (CIK 0001053507); SBA Communications’ and Crown Castle’s FY2025/Q1-2026 results for the three-way peer comparison; Equinix’s results for data-center context; and quantitative cross-checks via third-party financial-data aggregators (noting an erroneous aggregator AMT EV, recomputed by hand). All figures reconciled to filings where possible; management commentary is treated as hypothesis and validated against filings and external evidence.

APPENDIX A — Standard Diligence Questionnaire

American Tower Corporation (NYSE: AMT) — supplemental to the research memo. Report date: 2026-06-13.

Answers are grounded in the underlying analysis; Fact/Interpretation/Assumption labels applied where it matters.

General

What thoughtful questions have other investors asked about this company? The dominant questions: (1) When does AFFO/share re-accelerate after three flat years? (2) How much is CoreSite/AI worth, and does it deserve an optionality premium? (3) Where do emerging-market currencies (BRL, naira, peso) go? (4) Is AMT cheap after the de-rate from ~$292 to ~$187, or just fairly valued? (5) Has management’s capital-allocation pivot (India exit, deleveraging, ROIC-linked pay) durably fixed the returns problem the 2021 binge created? (6) Within the tower trio, why own AMT over the cheaper, higher-return SBAC?

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Interpretation: AFFO/share is at a cyclical low/plateau (~$10.5–10.9 for three years), depressed by India exit, FX, DISH churn, and refinancing. Management frames 2026 as the trough with re-acceleration in 2027. The underlying assets are not impaired; the per-share metric is cyclically/transitorily depressed.

Driven by the external environment or internal actions? Both: external (EM currencies, rates, DISH default) and internal (the deliberate India/Australia/South-Africa exits and dividend reset). The operating engine (organic billings) is healthy ex-one-offs.

How stable are revenues? Very stable at the contract level — ~98% retention, long leases, escalators, a large multi-year contracted backlog. Reported dollar revenue is destabilized mainly by FX translation and episodic merger-driven churn, not demand volatility.

Outlook for products/services? Towers: a durable annuity with secular data tailwinds, US ~5% ex-DISH / Europe ~4–5% / Africa ~11% (FX-eroded) / LatAm recovering. CoreSite: double-digit growth with AI/edge upside.

How big is this market — growing, shrinking, domestic/international? Global: US (mature), Europe (developed, mid-5G), Africa (early, fast, risky), LatAm (recovering), plus the secularly-growing data-center/interconnection market via CoreSite. The most diversified demand base of the three.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Structurally stable (3-player US oligopoly; concentrated tower ownership abroad); no new-supply threat. Competition is for incremental carrier dollars, intensifying as the US cycle matures, offset by EM and CoreSite growth.

How profitable is the business (ROIC, ROE)? Fact: ROIC ~7.85% (2025, recovering from ~4.6% in 2022) — between CCI (~7%) and SBAC (~12.4%); the gap to SBAC is the ~$27B M&A goodwill/intangible base, not weaker per-tower economics. EBITDA margin ~65.5%.

How profitable is the industry — competitors, barriers? Highly profitable (65%+ EBITDA margins across all three); high barriers (zoning, site scarcity, anchor tenant). AMT earns the same per-tower economics; its corporate ROIC reflects the prices paid for M&A.

Can the business be easily understood? Mostly — a global tower-leasing annuity plus a data-center business; the geographic breadth and FX add complexity.

Can it be undermined by foreign low-cost labor? No — physical real-estate assets; labor is minor.

Do brands matter? No — location, reliability, and price drive tower selection; CoreSite competes on interconnection density and location.

Nature of competition? Location, footprint, cost-to-lease, execution (towers); interconnection density and metro location (CoreSite). AMT’s edges are scale, diversification, and the CoreSite network effect.

Customers’ switching costs? High — relocating integrated antennas is costly and risky; CoreSite’s interconnection ecosystem is sticky (customers interconnected to 5+ others). The core of the moat.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The franchise value of ~150,000 site-monopoly towers and the contracted backlog exceed book value. Conversely, ~$27B of goodwill/intangibles from M&A sits on the books (and depresses ROIC).

Off-balance-sheet liabilities? Ground-lease commitments are substantially capitalized (~$7.8B finance leases); standard long-dated operating-lease and purchase commitments.

How conservative is the accounting? Reasonable; AFFO is conservatively defined (attributable, net of NCI). Watch the SBC add-back (~$174M, a real cost) and the recurring impairments tied to EM/M&A (~$684M 2022, ~$740M+ 2023, India in 2024).

How CapEx-hungry is the business? Capital-light to maintain; discretionary growth capex ~$1.68B/yr (builds + CoreSite development), tilted ~85% to developed markets + CoreSite.

Capital Allocation & Management

How much FCF, and how is it used? Strong (~$3.78B FCF 2025). Uses: a conservative ~63%-payout growing dividend, deleveraging, newly-initiated buybacks (~$560M Q4-25/Q1-26), and disciplined growth capex.

Philosophy? Post-2021-binge repair under CEO Vondran: delever, prune (India/Australia/SA), tilt to developed markets + CoreSite, ROIC in the comp plan, initiate buybacks. A pivot from empire-building to return discipline.

Significant acquisitions recently? No — the recent story is divestiture (India 2024). The 2021 Telxius (~$9.4B) + CoreSite (~$10.1B) binge is the relevant M&A history.

Buying back shares? Yes, newly — ~$560M+ (Q4-25/Q1-26), the most since 2017; a genuine new lever signaling balance-sheet confidence. Share count ~flat ~466M (SBC ~$174M).

Issuing stock to insiders? No material dilution; SBC ~$174M.

Compensation policy? Annual incentive on Total Property Revenue (30%) + Adjusted EBITDA (50%) + operational; LTI PSUs 40% AFFO/share + 30% average ROIC + 30% relative TSR — ROIC explicitly in the plan, directly addressing the return concern. Insiders <1%.

Motivations of management? New CEO Steven Vondran (Feb 2024, ex-CoreSite/US-tower head) + CFO Rod Smith; mandate is repair and return discipline. Two small director open-market buys (~$994K, ~$495K) near current price — neutral-to-slightly-positive; no conviction wave.

Valuation & Market Data

ADR, MLP, or K-1 issuer? None — US C-corp taxed as a REIT; issues a 1099-DIV.

Dividend policy? ~$7.16/yr annualized (~$1.79/quarter, ~3.8% yield), ~63% AFFO payout (most conservative of the three). Notably paused/reset in 2024 (held flat, reset down from ~$1.70 to ~$1.62) to delever; resumed ~5% growth in 2025–26.

How profitable is the business? Top-tier margins (~65.5% EBITDA); ROIC ~7.85% (M&A-goodwill-depressed); strong FCF (~8.7% yield).

Is net income diverging from cash from operations? Yes — GAAP net income is distorted by REIT depreciation and India discontinued-ops/CTA; OCF and AFFO are the honest measures. Use AFFO, not P/E.

Risks & Downside

What would cause the stock to decline? Persistent EM/FX weakness keeping AFFO/share flat past 2026; the 2027 re-acceleration slipping; a broad tower/REIT de-rate on rates; CoreSite/AI disappointing. Less likely: a balance-sheet/rate shock (AMT is the best-insulated of the three).

Risk of catastrophic loss? Very low — diversified, irreplaceable assets, investment-grade balance sheet, ~98% retention. Downside is to the multiple and the AFFO trajectory, not the franchise.

Chance of total loss? Very low — the lowest-risk of the three by balance sheet and diversification.

Recent News & Events

Has the business environment changed recently? Yes: CEO transition (Bartlett → Vondran, Feb 2024); the India exit (Sept 2024, ~$2.5B); dividend reset/pause to delever (2024); deleveraging to ~5.0x and a fully-unsecured IG balance sheet; buybacks initiated (Q4-25/Q1-26); the DISH default (Jan 2026, ~400bps 2026 AFFO churn); and continued CoreSite/AI expansion.

Significant acquisitions? None recently (divestitures instead). Telxius/CoreSite (2021) is the historical binge.

Change in accounting policies? No material change; India in discontinued operations.

Recent changes — new markets, facilities, management? Exited India/Australia/NZ/SA fiber; expanded CoreSite (+200MW); new CEO/strategy; ROIC-linked comp; buybacks initiated.

APPENDIX B — Source Appendix

American Tower Corporation (NYSE: AMT). Report date: 2026-06-13. Primary sources prioritized over secondary; management commentary treated as hypothesis and validated against filings and external evidence.

Primary — SEC filings (EDGAR, CIK 0001053507)

• American Tower FY2025 Form 10-K — segment/geographic detail (US & Canada, Latin America, Europe, Africa & APAC, Data Centers), site counts (~149,700 total + CoreSite), tenant concentration (T-Mobile ~18% / AT&T ~17% / Verizon ~14% / Telefónica ~10%), India discontinued operations, organic growth, churn, debt schedule (senior unsecured IG notes, maturities/coupons), goodwill/intangibles (~$27B), equity. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001053507&type=10-K
• American Tower Q1-2026 Form 10-Q — Q1 organic growth by region, DISH default/churn, FX, leverage, balance sheet. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001053507&type=10-Q
• AMT 2026 DEF 14A (proxy statement) — compensation design (annual: Total Property Revenue 30% + Adjusted EBITDA 50% + operational; LTI PSUs 40% AFFO/share + 30% average ROIC + 30% relative TSR), board, security ownership.
• Q4-2025 earnings release (8-K EX-99) and FY2026 guidance — AFFO/share guidance (~$10.78–10.95), Adjusted EBITDA, organic growth by segment, dividend, CoreSite capex.
• Q1-2026 earnings release (8-K EX-99) — raised guidance, organic growth ex-DISH/FX, CoreSite interconnection inflection. https://www.sec.gov/Archives/edgar/data/0001053507/000105350726000094/pressreleaseq12026.htm
• 8-K material-event record — CEO transition (Bartlett → Vondran, Feb 2024), India sale (Sept 2024), DISH default (8-K, Jan 2026), frequent IG bond issuances, dividend declarations.
• Form 4 insider-transaction corpus (CIK 0001053507) — transaction-code tally since 2024 (grants/exercises/sales dominate; two small director open-market purchases — Reilly ~$994K Nov-2025, Kalathur ~$495K Mar-2026; insiders <1%).

Primary — Company disclosures and transcripts

• Earnings-call transcripts (public): Q1-2026, Q4-2025, Q2-2025 — management framing of organic growth by region, CoreSite/AI-interconnection, the flat-AFFO drivers (India, FX, DISH, refi) and the 2027 re-acceleration, the dividend reset and deleveraging, buyback initiation, and the developed-markets tilt.
• India divestiture disclosures (Sept 2024) — ~$2.5B sale of ATC India (~76,000 sites) to a Brookfield-led consortium (Data Infrastructure Trust), including ~$320M of Vodafone Idea debentures.
• Telxius (2021, ~$9.4B) and CoreSite (2021, ~$10.1B) acquisition history — the cycle-top binge underlying the goodwill/leverage and subsequent repair.

Primary — Peer comparables

• SBA Communications (SBAC) — FY2025/Q1-2026 results and SEC filings, for the three-way peer comparison (highest ROIC/lowest payout/cheapest-vs-history).
• Crown Castle (CCI) — FY2025/Q1-2026 results and SEC filings, for the three-way comparison and the US tower-industry framing.
• Equinix (EQIX) — FY2025/Q1-2026 results and SEC filings, for data-center/interconnection context relevant to CoreSite.

Secondary — Trade press and financial media

• Coverage of the AMT India sale to Brookfield/Data Infrastructure Trust (2024); the Bartlett → Vondran CEO transition; the 2024 dividend pause/reset; CoreSite AI/interconnection expansion; and the DISH/EchoStar default.

Quantitative cross-checks

• Third-party financial-data aggregators — income statement, balance sheet, cash flow, profitability/credit ratios, and valuation multiples for AMT, SBAC, and CCI (reconciled to filings; SEC filings remain primary). Note: some aggregators report an erroneous AMT enterprise value that undercounts its ~$45B debt; the report uses a hand-computed EV of ~$131B (market cap ~$87B + net debt ~$43.5B).
• Public market-data sources — live price ($187.18), market cap (~$87.2B), share count (~466M), enterprise value (~$137B), debt/cash, 52-week range ($165–234).
• Own-history valuation percentiles (third-party): composite 44.4th, P/E 6.1th (REIT-noise), P/B 82.4th (elevated on a small positive book), P/S 44.7th — “fairly valued, mid-range.”