Yield

Yield

AT&T's reported free cash flow of $16.6 billion on today's $164.8 billion market cap is a 10.1% yield — level with the framework's 10% bar [1]. The framework's adjustment cuts that roughly in half. Subtracting stock compensation and the five-year average of acquisition spend — almost all of it spectrum — leaves adjusted FCF near $8.2 billion, a 5.0% yield, about 500 basis points short. The whole gap is the 2021–22 C-band spectrum bulge, now rolled off.

The deterministic feature file could not build any of this tab's core quantities: fit_features.adjusted_fcf.latest_adjusted, adjusted_fcf_yield, yield_baseline and balance_sheet_class all return not_computable — stock compensation is absent for the early years, no clean five-year adjustment window survives in the structured feed, and the FY2025 debt line is missing. Every figure below is rebuilt from the filed cash-flow statements, the share-based-compensation note and management's reported free cash flow, each anchored to its page.

The adjustment, line by line

Adjusted FCF removes two things the reported number keeps: the cost of paying employees in stock, and the capital that leaves the business as acquisitions — for AT&T, overwhelmingly spectrum. The three most recent years are the ones with a complete trailing five-year acquisition window inside the corpus.

Adjusted FCF = reported FCF − stock compensation − trailing five-year average acquisitions; yield on the current $164.8B market cap. Reported FCF: 2025 $16.6B [3], 2024 $17.6B [4], 2023 $16.8B [5]. Stock comp per FY2025 10-K Note 15 [6]. Acquisition averages derived from the cash-flow statements [7][8][9].

Two lines carry the weight. Stock compensation is trivial. AT&T recognized $536 million of share-based compensation in 2025, $505 million in 2024 and $479 million in 2023 — under half a percent of revenue and about 3% of reported FCF [10]. For a company this size that is a rounding item; the adjustment barely registers. This is the opposite of the software names the SBC screen is built to catch.

Acquisitions do all the work — and they are spectrum. The trailing five-year average of "Acquisitions, net of cash acquired" runs $7.9–8.4 billion across these three years, and the FY2022 10-K breaks that line into its parts: it is almost entirely spectrum, not businesses [11]. Two years dominate the window. In 2021 AT&T paid $24.7 billion for C-band licenses won in FCC Auction 107; in 2022 it paid a further $9.1 billion, largely for 3.45 GHz spectrum [12]. Every other year in the series is far smaller.

The 2021 C-band ($24.7B) and 2022 3.45 GHz ($9.1B) auctions are ~90% of the seven-year total; 2024–25 ran under $0.4B. Continuing-operations basis. 2019–2021 per the FY2021 10-K [13]; 2020–2022 breakdown per the FY2022 10-K [14]; 2023–2025 per the FY2025 10-K [15].

So the adjustment is not really a stock-comp haircut; it is a spectrum haircut, and it is large because the five-year window still contains a once-a-decade mid-band build-out. That is the fact the reader has to weigh: whether $7.9 billion a year is the right run-rate to charge against owner earnings, or a backward-looking artifact. The Normalized spectrum charge section works both readings.

The yield, three ways

Reported: $16.6B ÷ $164.8B. Adjusted current: $8.19B ÷ $164.8B. Three-year average adjusted FCF ($8.36B, FY2023–25) ÷ current cap. Sources as in the adjustment table above [16][17][18].

On reported FCF, AT&T yields 10.1% today and has sat around that level for a while; on adjusted FCF it yields 5.0%, and the three-year average adjusted yield is essentially the same at 5.1%. The adjusted figure is stable, not volatile — the spectrum charge is steady enough that the current year and the three-year average land within ten basis points of each other.

On the baseline and the "jump" test, AT&T is not the fortress signature. The framework's prized pattern is a name that yielded a stable ~3.5–4% for years and suddenly prints 8–9% — a dislocation, not a re-rating. AT&T is the reverse: its reported FCF yield has been persistently elevated for years — a high-single-digit-to-low-double-digit level that reflects durable market skepticism about $117.9 billion of net debt and cable's encroachment, not a sudden break from a calm baseline. The fit_features.yield_baseline feature is not_computable (it needs fiscal-year-end prices and same-year shares the feed lacks), so there is no clean per-year distribution to point to; that read is qualitative, drawn from the steadiness of the reported-FCF band against a market cap that has traded at a discount throughout. What has moved recently is price: the drawdown from $29.62 in September 2025 to a $20.48 trough — 31% peak-to-trough, with the stock at $22.96 now — lifted the reported yield roughly 230 basis points (from ~7.8% at the September peak market cap to 10.1% today) and the adjusted yield about 110 basis points (from ~3.9% to 5.0%). That is a cyclical lift on top of a persistently elevated level — the anatomy of it sits in Dislocation — not a fortress jumping off a stable base.

Which bar applies

The reference line comes from the balance sheet. The balance_sheet_class feature is not_computable (the FY2025 debt line was missing from the feed), so it is rebuilt here.

Net debt = total debt $136,100M (long-term $127,089M + current maturities $9,011M) − cash $18,234M, per FY2025 10-K balance sheet [19]. Adjusted EBITDA implied from management's stated 2.53× leverage; ratio per the Q4 FY2025 call [20].

Net debt is $117.9 billion [21]. AT&T states it ended 2025 at net debt to adjusted EBITDA of 2.53 times, at its 2.5× target [22]; that ratio implies adjusted EBITDA near $46.6 billion. Against the framework's rule — fortress at net debt/EBITDA ≤ 0.5×, levered at ≥ 3.0×, moderate in between — 2.53× is moderate, which selects the 10% default bar, not the ~8–9% fortress line and not the 25% levered line.

That places the read plainly: 5.0% adjusted yield against the 10% bar is 500 basis points short — the name yields about half its bar on the mechanical adjustment. Even the unadjusted 10.1% only reaches the bar because it ignores the spectrum spend the balance sheet had to fund. One caveat on the bar itself: the pending EchoStar and Lumen deals push pro-forma leverage to roughly 3.2× on close, before management guides it back toward 3.0× by year-end 2026 and to the 2.5× target within about three years [23]. At 3.2× the name would momentarily tip into the levered class — but that is a deliberate, temporary step-up for spectrum, not structural leverage, so the 10% bar is the honest reference line, with a note that leverage is heading the wrong way near-term.

Normalized spectrum charge

AT&T is not a cyclical business — its cash does not swing with a commodity or an underwriting cycle — so there is no mid-cycle earnings normalization to do on the revenue line. The normalization that matters is the acquisition line, because the mechanical five-year average is dominated by a spectrum event that does not repeat annually. This is the judgment on top of the arithmetic, stated so it can be recomputed under different assumptions.

The mechanical charge is ~$7.9 billion a year. Its components: the 2021 C-band ($24.7B) and 2022 3.45 GHz ($9.1B) auctions supply roughly 90% of it, while the ordinary years — 2019, 2020 and 2023 — ran $1.6–2.9 billion and 2024–25 ran under $0.4 billion [24][25]. A defensible mid-cycle charge treats the C-band super-cycle as non-repeating, but does not zero out spectrum — because it does recur. In August 2025 AT&T agreed to buy 600 MHz and 3.45 GHz licenses from EchoStar for approximately $23 billion, closing in early 2026 [26]. Amortized over a ~10-year license life that is ~$2.3 billion a year; add ordinary auction and small-M&A cadence of ~$1.5–2 billion, and a mid-cycle charge lands near $4 billion a year.

All on FY2025 reported FCF $16.6B less $0.54B stock comp less the stated spectrum charge, over the current $164.8B cap. Sources as above.

The point of showing all three: under every acquisition assumption from mechanical to generous, the current adjusted yield sits below the 10% bar — 500 basis points short on the mechanical charge, 270 short at a mid-cycle $4 billion, and still ~90 short even if you treat the pending $23 billion EchoStar purchase as a one-off and charge only $1 billion a year. The reported 10.1% clears the bar only by pretending spectrum is free. It is not: the balance sheet funded $34 billion of it in two years and is about to fund $23 billion more.

The consensus check and the forward path

The current-year read is below the bar on adjusted FCF. The forward question is whether it reverts above it, and here consensus is genuinely constructive — but on the reported line, which is the distinction the adjustment exists to draw.

Consensus reported-FCF yields from S&P Capital IQ estimates (fit_features.consensus_forward_yield, source data/sp/estimates.json): FCF means of $18.2B (2026), $19.25B (2027), $21.2B (2028), $22.6B (2029) over the current $164.8B cap. Adjusted-forward applies −$0.55B stock comp and a −$4B mid-cycle spectrum charge to each. Management's own guidance aligns: FCF over $18B in 2026, growing ~$1B in 2027 and ~$2B in 2028 [27].

Consensus reported-FCF yield already clears the bar — 11.0% in 2026 rising to 13.7% by 2029 — and the sell side and management agree on the path: FCF grows from $16.6 billion toward roughly $22 billion as the fiber build peaks, capital intensity eases from high-teens to mid-teens percent of revenue, depreciation holds near $20 billion, and $4 billion of cost savings land by 2028 [28][29]. Management underwrites the same math from the return side: over $45 billion of planned 2026–28 shareholder returns is described as "over 75% of our expected free cash flow," which implies roughly $60 billion of three-year FCF, about $20 billion a year [30].

The adjusted line tells the harder story. Charge the same mid-cycle $4 billion spectrum and $0.55 billion stock comp against consensus FCF, and the adjusted forward yield runs 8.3% in 2026, 8.9% in 2027, and crosses the 10% bar around 2028 at 10.1%, reaching 10.9% by 2029. So the mean-reversion underwrite is not "already agreed" — it is a two-to-three-year path, and it has to clear two conditions to land:

• FCF growth of ~$6 billion (36%) by 2028–29 must materialize. It is well-supported — capex intensity is falling as major projects complete by 2030, D&A is flat, and consensus and management guidance coincide — but it is not yet in the numbers.
• Spectrum and M&A must not re-accelerate beyond EchoStar. A fresh auction super-cycle on the scale of C-band would reset the trailing charge upward and push the crossing out again. The recurring nature of spectrum — the very reason the adjustment charges for it — is also the main threat to the reversion.

My read, stated once: on adjusted FCF, AT&T sits below the 10% bar today under every reasonable spectrum assumption, and reverts above it around 2028 only if consensus FCF growth is roughly right and spectrum spend normalizes toward $4 billion a year. I would put that reversion at a little better than even — call it 55–60% within three years — because the FCF path is corroborated by both management and the sell side, while the spectrum-normalization leg is the genuine uncertainty. The strongest fact against the read is that on reported FCF the name already yields 10% and consensus has it rising to 13%+ — a skeptic can fairly argue the spectrum charge double-counts capital the market already sees in the leverage. The strongest fact for it is that AT&T is, on the record, about to spend $23 billion more on spectrum, so treating that spend as real is not conservatism — it is arithmetic. What would move the read: capital intensity easing on schedule and a quiet spectrum calendar would pull the crossing into 2027; another C-band-scale auction would push it past 2029.

Conversion is holding

A deteriorating FCF-to-revenue trend would undercut both the reversion and the durability case. It is not deteriorating.

Reported FCF over total operating revenues (continuing operations): 2022 $14.1B/$120.7B, 2023 $16.8B/$122.4B, 2024 $17.6B/$122.3B, 2025 $16.6B/$125.6B. FCF per the earnings releases and calls [31][32][33][34].

Reported FCF converts 12–14% of revenue and the trend is up from 11.7% in 2022, with one honest seam: 2024's 14.4% is flattered by ~$2.3 billion of DirecTV distributions, so on the like-for-like basis 2024 was closer to 12.5% and 2025's 13.2% is the higher of the two clean years [35]. Conversion is stable-to-improving, not eroding — which supports the reported-FCF-growth story that the reversion depends on. The FCF-stability question that anchors the year-10 durability test is developed in Durability; this tab's contribution is the adjusted-FCF reconstruction beneath it. The reservation the whole tab keeps returning to is not conversion or stability — it is that the capital intensity protecting the franchise, spectrum included, keeps adjusted owner earnings well below what the reported yield advertises.