Open in interactive viewer → charts, metric popovers & call review

SBA's Raise Came From the Balance Sheet, Not the Antennas

A share-count-and-interest beat, churn with an end date, and a spectrum cycle management itself keeps pushing toward 2028
SBAC · Earnings Call · 2026-11-02

The raise that wasn't organic

SBA Communications reported third-quarter 2026 results on November 2 and delivered what the slide deck promised: AFFO per share of $3.19, company-wide tower cash flow margins of 80%, an adjusted EBITDA margin of 69%, and roughly $430 million returned to shareholders in a single quarter. Management nudged the full-year AFFO midpoint up by $2 million and eight cents a share. Then look at what produced the raise. “These increases in [AFFO] and [AFFO] per share are mostly driven by lower net cash interest expenses, lower nondiscretionary capital expenditure, along with the benefit of a lower share count from our third quarter share buyback.” — Marc Montagner, Chief Financial Officer · 2026-11-02 CFO Marc Montagner was equally plain that absent FX and the exit from Chile and Ecuador, the full-year outlook for site leasing revenue, tower cash flow and adjusted EBITDA would have been essentially flat. This was a balance-sheet and share-count beat, not an operating one. The underlying financials bear that out. Funds from operations came in at $233 million, down 9% year over year, and net income fell 12% to $199 million. The net profit margin of 27.8% is down 4.5 percentage points from a year ago. And the culprit is the one management cannot argue with: interest expense rose 41% year over year to $59 million, the arithmetic cost of rolling one-handle legacy coupons into a five-handle world. Unlike most of this call, that is not a story about demand. It is a story about the bond market.

Spectrum: the first genuinely dated catalyst in years

What is new — and genuinely new, not a re-run — is the specificity around new spectrum. Brendan Cavanagh laid out a calendar rather than a vibe: the upper C-band auction begins in April 2027 and the FCC is required to complete it by July, releasing 160 megahertz between 3.98 and 4.14 gigahertz and, combined with existing lower C-band, creating 440 megahertz of contiguous mid-band. Beyond that, rulemaking on 2.7 and 7 gigahertz is expected soon, and the 4 gigahertz gigahertz band is now being studied for full-power commercial licensing.

In terms of the spectrum, though, that's already in the hands of some of the parties like the 600 megahertz that AT&T has or the C-band that T-Mobile has, I expect that we'll definitely start to see those before that time frame. Hopefully, some of that will start to see next year.

Brendan Cavanagh, President and Chief Executive Officer · 2026-11-02
That is a materially firmer answer than the same analyst got a year ago, when the reply was a diffuse "several years to get it cleared… probably into the next decade." The precision matters because SBA's growth algorithm is a spectrum-deployment derivative — every auction eventually becomes equipment on a tower and a new lease or amendment. But here is the twist Cavanagh volunteered, and it cuts against the bull case: because the AT&T master agreement caps how much SBA can earn from AT&T's 600-megahertz rollout, and that MLA expires in roughly eighteen months, he noted that "the longer it takes, maybe the better from our perspective." A CEO openly comfortable with delay is not a growth-flexing posture. It is a hedged one.

Churn finally gets arithmetic, and an end date

The other change is that the two structural churn problems now have numbers rather than adjectives. Sprint churn was $16 million in the quarter, tracking to $55–56 million for the year, with less than $20 million remaining in total for 2027 and beyond. EchoStar-related churn was $14 million in the quarter with $56 million expected for full-year 2026 — almost exactly the ~$55 million annual DISH revenue figure management disclosed in August 2025. The feared number arrived precisely on schedule, and the market has had a year to price it. Better, SBA now owns the debris. “The equipment, as I said earlier, is now ours and within our control. And so anybody that wants to use that would obviously have to deal with us on that.” — Brendan Cavanagh, President and Chief Executive Officer · 2026-11-02 That follows two prior calls of stonewalling on the identical question — “David, first of all, you don't know what we're doing and what we're not doing.” — Brendan Cavanagh, President and Chief Executive Officer · 2026-02-26 Roughly 3,000 terminated DISH leases and the gear sitting on them now represent a free option with, per management, several interested parties. Internationally, Cavanagh for the first time put a normal level on churn — 2% to 2.5% — after years in which the theme dominated the company's keyword set without a terminal value. Elevated through 2027, normalizing in 2028, with the portfolio review now having sold roughly 1,800 South American sites for about $43 million of annual leasing revenue and $20.5 million of annual adjusted EBITDA. Pair that with U.S. churn converging toward 1% (against 1.4% ex-Sprint/EchoStar today), and the recovery arc the CFO sketched in February — “we'll be moving partially back there in 2027 and I think more fully there as you get into '28 and '29” — Marc Montagner, Chief Financial Officer · 2026-02-26 — is at least internally consistent.

What the cash says that the adjectives won't

On capital allocation the actions are louder than the language. SBA repurchased 1.7 million shares at an average of $179.68 for $300 million, has $800 million left on authorization, and lifted the dividend 13% year over year to $1.25 a quarter — about 41% of the AFFO per share midpoint. The investment grade transition is now operational: an inaugural $3.5 billion bond in July, a $2.5 billion commercial paper program stood up on September 25, and a full refinancing of $1.165 billion of ABS at an estimated 5.25%, with leverage at 6.4x net debt to adjusted EBITDA, below the 6–7x target midpoint. The rationale for the buyback is not growth optimism; it is relative price.

When we look at private assets in the U.S. specifically today and granted, they're smaller and so perhaps can be more competitively bid after. The valuations are much, much higher than where SBA and the other public companies, for that matter, are valued.

Brendan Cavanagh, President and Chief Executive Officer · 2026-11-02
That grievance is a year old and unresolved — “I would say that our valuation is way too low is really the issue.” — Brendan Cavanagh, President and Chief Executive Officer · 2025-11-03 Two consecutive calls of the same complaint, now converted into a $300 million check, is the signal. Management said plainly that if asset valuations stay this stretched, buybacks will outrun acquisitions. Note also what management declined to claim. Asked whether SBA would chase the 200–300 basis points of margin expansion by 2030 that peers have published, Cavanagh pushed back: “We've actually operated as efficiently as anybody. Despite being the smallest company in the industry, we have been actually the most efficient from an overhead perspective.” — Brendan Cavanagh, President and Chief Executive Officer · 2026-11-02 Translation: there is no self-help margin story left to tell. Any re-rating has to come from the multiple, not the P&L.

Edge compute, swimming against the tape

The newest narrative thread is edge compute. Several sites are live with companies building distributed AI inferencing applications — typically 200–300 kilowatts, some at a megawatt or more, behind the meter, using vacant shelter space and excess power at minimal incremental capital. It is a real graduation from talk; in February Cavanagh himself said edge compute was something "we've been talking about for a decade, and we haven't really seen much of it." But revenue is explicitly modest and near-term, and the customer base is mostly neoclouds and startups. That lands into a hostile tape. Globally, the AI infrastructure complex is among the worst-performing theme clusters over the last 30 days: AI data center exposure is down more than 20%, with data center capacity names off roughly 34% and critical-IT-load stories down more than 36%. SBA is pitching the capital-light, distributed version of an idea whose capital-heavy version is being de-rated. It may be the better business and still get no credit for a year. The setup, then: a company with contracting near-term earnings, an investment-grade balance sheet, a shrinking share count, churn that peaks in 2026 and normalizes by 2028, and a spectrum deployment cycle that begins — maybe — in 2028. The stock sits 52.5% below its December 2021 peak of $389 and has gone essentially nowhere over ninety days. Management says the shares are cheap and is putting $300 million behind that claim. The honest read is that this is a wait-for-2028 equity trying to pay you a rising dividend and retire stock while it waits.