Self-Help

Self-Help

AT&T can fund its own recovery. Net debt of about $117.9 billion sits at 2.53x adjusted EBITDA against a 3.75x covenant, with laddered maturities near $8–9 billion a year, $18.2 billion of cash and two undrawn multi-billion facilities [1] [2]. The buyback engine is real but young — token before 2025, then $4.5 billion in 2025 and roughly $10 billion planned for 2026 [3] [4]. The dividend was reset once, in 2022, and has held since. Share count is falling, not rising.

The balance sheet against the problem's duration

The dislocation here (Dislocation) is a valuation and fiber-ARPU worry, not a solvency scare — and the balance sheet is built to outlast it. At December 31, 2025, AT&T carried $136.1 billion of total long-term debt (including current maturities), of which $9.0 billion is current; the blended coupon is about 4.2%, and substantially all of it is unsecured [5]. Against $18.2 billion of cash, net debt is roughly $117.9 billion, which management reports at 2.53x adjusted EBITDA — below the 3.75x net-debt-to-EBITDA covenant in its credit agreements, and inside the "moderate" band of the framework's own leverage rule rather than the levered band [6] [7].

The maturity ladder is the reassuring part. No single year before 2031 requires more than about $9 billion, well inside the roughly $40 billion of annual operating cash flow and $16.6 billion of free cash flow the business throws off.

Source: FY2025 Annual Report (Form 10-K), Note 11 Debt, scheduled maturities table [8]. The final bar aggregates all maturities due 2031 and later.

The concentration sits in the "thereafter" stack — $106.2 billion, or about 78% of notes and debentures, matures 2031 or later, at a 4.3% weighted rate [9]. That is a long runway, but it also means refinancing is perpetual: AT&T issued $14.0 billion of new long-term debt in 2025 at about 5.0%, and another $6.5 billion in February 2026 at a 5.2% coupon [10]. Rolling 4.2% paper into 5%-plus paper lifts interest expense gradually — a headwind to free cash flow, not a threat to it.

Liquidity behind the ladder is deep. On November 3, 2025 AT&T put in place a $12.0 billion revolving credit agreement and a $17.5 billion delayed-draw term loan, neither drawn at year-end, and the company reports it was in compliance with all debt covenants [11]. The one wrinkle is the pending EchoStar spectrum purchase: management expects net leverage to rise to about 3.2x on close, then return to the 2.5x range within roughly three years [12]. That temporary step-up consumes flexibility that could otherwise front-load repurchases, but it stays inside the covenant and inside the moderate band. On the balance-sheet leg of self-help — can the company comfortably outlast the problem without capital allocation being forced toward debt paydown — the answer is yes: the deleveraging is essentially done, and cash is now free to move to shareholders.

The repurchase record — executed, not authorized

The record separates cleanly into two eras. From 2016 through 2024, buybacks were a rounding item — $0.2 to $0.9 billion in most years, with a single $5.5 billion burst in 2020 — while the dividend did all the returning. The share count did not shrink in this period; it rose, from 6.19 billion in 2016 to a peak of 7.59 billion in 2022, as the Time Warner and DirecTV-era transactions issued stock [13]. The buyback as a real capital-allocation pillar only began in 2025.

Source: cash-flow statements, FY2016–FY2025 10-Ks; FY2025 figure $4.5B [14]; series per fit_features.share_count_trend.buyback_cash_per_year.

Source: fit_features.share_count_trend; FY2025 shares of 7,179M per income.json and market-cap feature.

Two facts matter for the framework. First, the direction has turned: the share count has fallen from 7.59 billion (2022) to 7.18 billion (2025), a 5-year CAGR of about −0.8%, and the deterministic feature records rising = false — so the hard-fail condition, a share count climbing on stock-based compensation or serial M&A, does not apply today (fit_features.share_count_trend). Second, the engine is nascent and expensive to date: the 2025 repurchases retired about 168 million shares for $4.5 billion, an average of roughly $27.0 per share [15] — above today's $22.96. The buyback flywheel Ruchir looks for is only just starting to spin; it is not a decade-long habit like the levered precedents.

Management's buyback intent, from the record

Intent is where the recent record is strongest and most explicit. AT&T reached its 2.5x leverage target in mid-2025 and commenced repurchases under a $10 billion authorization, guiding to at least $3 billion completed by year-end [16]. On the Q4 2025 call CFO Pascal Desroches framed a three-year program: "we expect to return over $45 billion to shareholders during 2026 to 2028… we expect to maintain our current common stock dividend with a consistent pace of share repurchases through 2028, including approximately $8 billion of buybacks in 2026," with the Board authorizing an additional $10 billion beyond the current tranche [17].

By Q2 2026 the pace was raised. Desroches said AT&T would "further increase our pace of planned share repurchases this year by up to 25% to approximately $10 billion to capture what we see as a disparity between our operating fundamentals and the valuation of our stock" [18]. CEO John Stankey put it in valuation terms directly: "I'd like to buy more of it back because I think it's incredibly undervalued" [19]. This is exactly the price-sensitive, valuation-triggered repurchase intent the framework wants at a moment of maximum fear.

The counter-facts are worth stating plainly. The $10 billion raise is a pull-forward of buybacks already planned through 2028, not net-new capital — the $45 billion three-year envelope is unchanged. And Stankey deferred the ultimate allocation decision, not to himself but to the Board: "this Board will be deliberate. They'll spend some time in September, at their September meeting on this topic" [20]. Management has also declined to commit to a buyback floor when pressed. Intent is genuine and escalating; it is not yet locked in.

On the insider leg: there is no open-market buying alongside the corporate buyback. Every Form 4 transaction in the last cycle is an equity grant or award (SEC code A), not a purchase — including Stankey's — so directors and officers are accumulating stock through compensation, not adding conviction with their own cash (governance/insider_activity). That is common for a widely held mega-cap, but it is not the insider-buying signal that would strengthen the case.

The levered exception — does not apply

The framework tolerates real leverage only when the adjusted yield is massive (roughly 25%+) and a multi-year share-count reduction is already demonstrated and FCF and revenue are not deteriorating — all three legs, with numbers. AT&T meets none of the first leg's bar: the adjusted free-cash-flow yield cannot be computed from the feature file (adjusted FCF is not_computable — see Yield), but consensus forward free-cash-flow yield sits near 10–12% on current market cap, not 25%+ (fit_features.consensus_forward_yield). And the leverage itself, at 2.53x, is moderate, not levered. This is the default 10%-bar case, not the Charter-style levered path — so the levered exception is not the frame through which to read AT&T, and the yield question belongs to the ordinary bar in Yield.

The absurdity check

The deterministic float-retirement figure is not_computable because it requires positive adjusted FCF, which the feature file could not build (fit_features.float_retirement_years). Using reported free cash flow as the basis instead, the arithmetic is un-absurd: at a $164.8 billion market cap and $16.6 billion of 2025 free cash flow, it would take about 9.9 years of free cash flow to retire the entire float; on the 2026 guide of over $18 billion, about 9.1 years [21]. That is a stock priced at roughly ten times free cash flow — cheap for a durable oligopolist, but not the ~3-year figure that signals a price making a claim it cannot survive. The self-help math here is a steady grind, not a coiled spring.

Dividend safety

The dividend is a material part of the return — $1.11 per share against a $22.96 price is a 4.8% yield — so it earns a full look [22].

Coverage is comfortable. AT&T paid $8.18 billion in common and preferred dividends in 2025, against $16.6 billion of free cash flow — roughly 2.0x cover, and a payout of about half of free cash flow, leaving headroom for the buyback alongside [23]. Management's 2026 plan — about $18 billion of combined dividends and buybacks against an $18 billion-plus free-cash-flow guide — is essentially 100% of free cash flow, but the split (dividend fixed, buyback flexible) means a soft year pressures repurchases first, not the dividend [24].

The record through stress is the honest caveat. AT&T cut its dividend once, and recently: the payout was reset from $2.08 to $1.11 per share in 2022 — a 47% reduction — as part of the WarnerMedia separation, taking annual dividend cash from about $15.0 billion to $8.0 billion [25]. Since then it has held flat at $1.11 through 2023, 2024 and 2025, and management commits to "maintain our current common stock dividend… through 2028" [26]. So the base case for the dividend is safe — 2.0x covered, flat-to-held, explicitly prioritized over buybacks — but this is not a serial raiser with an unbroken record; it is a rebased payout three years into a hold. What would force a cut: free cash flow falling toward $12 billion or below while the deleveraging and spectrum bills still had to be paid, which the current $18 billion-plus guide leaves a wide margin against.

Sources: net debt derived from Note 11 debt of $136.1B less $18.2B cash [27]; 2.53x leverage and cash per Q4 2025 call [28]; dividend $1.11 on $22.96 and $8.18B paid vs $16.6B FCF [29].

Management credibility

Judged on promise-versus-delivery across the transcript archive, current management (Stankey and Desroches) reads as a keeper of controllable commitments rather than a promotional pattern. Five material promises from the 2023–2024 calls and their outcomes:

Source: earnings-call guidance ledger, Q3 2023–Q2 2026 (calls curation); FY2025 EPS/FCF actuals confirmed on the Q4 2025 call [30].

Four of the five landed; the miss was a Mobility EBITDA guide walked down mid-year, not a headline blow-up. Management telegraphed the one hard action it took — the 2022 dividend reset — well ahead of it, rather than defending an unsustainable payout, and the deleveraging promise that unlocked buybacks was hit on schedule. The blemishes are real but modest: a Q3 2025 deflection on CEO succession, no buyback floor when pressed, and no open-market insider buying to back the "incredibly undervalued" language.

The larger caution is corporate history, not current promises. The 2015 DirecTV (~$49 billion) and 2018 Time Warner (~$85 billion) acquisitions destroyed value, drove the share count up by roughly a fifth, and forced the 2022 dividend cut and the WarnerMedia unwind — a value-destruction record on the same board, executed by prior management. Stankey, who led the simplification, is the counter to that pattern rather than its author. On the framework's exclusion test — a promotional CEO with big claims, repeated misses and low skin in the game — AT&T does not trigger it: the numeric commitments have largely been met, the language is measured, and the one big reversal was owned rather than denied. It is a clean-enough record, with the dividend cut and the M&A history on the ledger as facts, not a promotional flag.