American Tower's Third Raise Is a Data Center Story Wearing a Tower Costume
AMT sold Canada at 26x, bought back stock, and is dangling Starlink optionality—while US towers are still flat ex-DISH.
AMT · Earnings Call · 2026-10-27
American Tower's third 2026 guidance raise landed with a familiar REIT rhythm: a solid quarter, a bigger data-center number, a smaller services number, and a board that would rather buy its own stock than chase tower M&A. The headline was upbeat. “The momentum we are seeing across the company combined with the benefit of our share repurchase activity, gives us confidence to raise our full year outlook for the third time this year.” — Steven O. Vondran, President and CEO · 2026-10-27 But the composition is the story. The raise is not being driven by the US tower business, which was flat organically in Q3 and only about 4% ex-DISH. It is being driven by CoreSite, buybacks, and a Canada exit.
A raise carried by data centers, not towers
We are raising guidance for property revenue and AFFO per share while maintaining our adjusted EBITDA outlook.
That sentence is the tell. Property revenue guidance rose by $55 million at the midpoint and AFFO per share by $0.02, but adjusted EBITDA was held flat; a $15 million services shortfall offset data-center outperformance. The data center platforms story is real—CoreSite revenue growth guidance went to 16%, up from 15% last quarter and 13% at the start of the year—but it is now carrying a raise that the tower business cannot. US and Canada organic growth was flat, or about 4% after excluding DISH churn; Africa and APAC grew over 8%, Europe over 4%, and Latin America declined about 2%. Rodney called 2026 a trough for both organic tenant billings and Attributable AFFO per share. That trough language is not new: in July he had already called 2026 an inflection year. “we do view this as an inflection year where we will be driving higher AFFO per share growth going forward” — Rodney Smith, Executive Vice President, CFO and Treasurer · 2026-07-28 The market has heard the promise before; what it has not yet seen is the US tower acceleration.
The tape is making a distinction. The global AI data center basket is down about 5.6% over the past 30 days on a 6-to-48 positive-to-negative ticker split, while AMT is up 4.2% over 90 days. American Tower is not being bought as a pure AI data-center proxy; it is being bought as an interconnection-rich landlord with tower cash flow and a lower-beta data-center kicker. CoreSite's record leasing activity and rising cross-connects support that, but management still insists CoreSite is a core asset rather than a candidate for separation. In July, Steven said, “We certainly consider CoreSite to be a core asset” — Steven Vondran, President and CEO · 2026-07-28 On this call he added that AMT is exploring new markets and inorganic opportunities, but the strategic answer has not changed.
Canada: selling high, buying low
The bigger new fact is the Canada portfolio sale. AMT agreed to sell its Canadian towers for about $2 billion, or roughly 26 times 2026 expected EBITDA, with closing expected in Q4. The proceeds are earmarked for share repurchases, debt reduction, and higher-growth investments. The board added $2 billion to the buyback authorization, bringing remaining capacity to about $3.1 billion; AMT repurchased $315 million in Q3 and another $68 million through October 20. This is a quiet arbitrage: sell towers at a private-market multiple and buy AMT stock at 21.7x trailing net income, versus a 96.8x peak in 2024. With leverage at 4.9x and a BBB+ credit rating, management has the flexibility to keep doing it.
But the same discipline explains why material M&A is still missing. Steven repeated the three-part test from April: “you've got to have a willing counterparty, a constructive regulatory environment and the economics have to make sense” — Steven Vondran, President and CEO · 2026-04-28 The private-public multiple gap that made AMT a seller of Canada also makes it a reluctant buyer of other tower portfolios. That is coherent capital allocation, but it leaves the inflection in growth leaning on buybacks, CoreSite, and the 2027 tower recovery rather than a transformative deal.
DISH is now an asset; Starlink is the option
The DISH overhang is no longer just a churn line. DISH-related churn is about 400 basis points of headwind, and management again asked investors to look through it. But AMT now owns the DISH equipment left on its towers after the bankruptcy abandonment order. “yes, we own that equipment” — Steven O. Vondran, President and CEO · 2026-10-27 Steven said AMT can provide access to interested parties. That turns a dead customer into a potential re-leasing asset—and it sits beside the fresher Starlink optionality.
The stock jumped 8.6% on October 9 when SpaceX agreed to spend $8 billion on an 800-MHz spectrum portfolio, a deal that makes a terrestrial network more plausible. On the call, Steven was careful but excited: “We are really excited about the prospect of maybe having a new customer at scale in the US.” — Steven O. Vondran, President and CEO · 2026-10-27 He would not discuss specific Starlink talks, but added, “anybody who has wireless ambitions is probably somebody that we are talking to” — Steven O. Vondran, President and CEO · 2026-10-27 In July, the same logic was already there: “if they decide to go that route, we are confident that we would be a good partner for them” — Steven Vondran, President and CEO · 2026-07-28 What is new is that the market now has a public catalyst—and AMT has equipment and spectrum-adjacent sites that could matter.
The risk is that this remains a range of outcomes rather than a number. Management is not guiding any satellite revenue, and DISH mediation plus the AT&T Mexico arbitration still lack resolution. The 2026 raise is real, but it is narrow. The 2027 inflection is still a promise. If Starlink or DISH equipment monetization becomes tangible, the upside is not in the current outlook; if it does not, AMT is a low-growth tower REIT with a data-center kicker and a buyback. The next proof points are the Canada close, CoreSite leasing, Starlink engagement, and any DISH resolution. That is enough to keep the story alive—but not yet enough to call the tower cycle fixed.