AT&T T

Communication Services · Telecommunications Services · Synthos Deep Dive · 2026-08-04

$26.00

Buy — Tactical

—

The Overview

AT&T sells mobile phone service to about 120 million connections and, increasingly, home internet over fibre-optic cable and wireless. It is one of the two largest telephone companies in the United States.

The company's total sales grew only 2.3% last quarter, which sounds dull. But that number hides the actual story, and this year AT&T changed how it reports so the story is finally visible.

It now splits itself into three parts. The big part — 91% of the business, called Advanced Connectivity — grew 4.1%, and its profit grew 20.3% because costs barely moved and the accounting charge for wearing out old equipment actually fell. Its profit margin went from 22.2% to 25.7% in a single year. Within it, home internet revenue grew 27%.

The small part — 5% of the business, called Legacy — fell 26%, because AT&T is deliberately switching off its old copper telephone network. That decline is a choice, not a failure, and it is what drags the headline number down to 2.3%.

The shares are $23.38. The company is expected to earn about $2.34 a share this year and $2.56 next, so you pay about 10 times this year's earnings and 9 times next year's. It generated $19.4 billion of free cash last year — about 12% of the entire value of the company — and pays a dividend of $1.11 a year, a 4.75% yield, which the cash covers two and a half times over. Over the past twelve months the shares fell 15.7% while the American stock market rose 24.3%.

Two cautions. First, the data file we normally rely on is badly wrong here: it says AT&T owes $174 billion when the real figure is $144 billion, because it counts the company's building leases twice. That error makes the company look far more indebted than it is.

Second, and this is real: AT&T agreed last August to buy radio spectrum from a company called EchoStar for about $23 billion, paying with a mix of cash and new loans. That is a large commitment for a business already carrying a lot of debt, and it will push borrowings back up after several years of paying them down.

Our estimate of fair value is $26.90, about 15% above the price, plus the dividend. The average analyst says $27.13 — but the analyst community as a whole rates the shares a Hold, and six firms tell clients to sell.

Downside Risk 5/10. Enormous, stable cash generation and the least volatile share price in this group — against $23 billion of new commitments.
Growth Quality 5/10. Poor headline, genuinely good composition. Profit in the main business is growing five times faster than its sales.
Exponential Potential 3/10. A mature market. The fast-growing home-internet line is too small to change the arithmetic.

Putting a number on it: our fair-value estimate is $26.90 against a current price of $26.00 — real upside if our numbers are right.

Target entry zone $25 – $26 accumulate in this band; ideal adds on a dip toward the 200-day average near $25, keeping roughly a 3% margin below our $27 base-case fair value

Our summary metrics

Downside Risk (lower = safer)5/10Moderate

"Rated 5 — very large absolute debt against very large and very stable cash generation, with one identifiable leverage step immediately ahead. Supports: free cash flow of $19,442M in 2025 covers the roughly $7.9 billion dividend 2.5 times; operating cash flow of $40,284M on $125,648M of revenue is a 32.1% margin; beta is 0.422, the lowest in this batch; interest coverage is 3.59x; the vendor's composite rating of A- / 4 is the highest in this batch and, unusually, we do not reject it; the funded balance sheet is materially better than any screen shows once the vendor's $18,943M lease double-count is removed — funded net debt at 2026-06-30 is $126,384M, not the $155,752M implied by `bal_a`; and 158.7 million shares (2.3% of the company) were retired in six months for $3,917M. Against that: AT&T agreed on 2025-08-25 to acquire FCC licences from EchoStar for approximately $23,000M, to be funded 'using a combination of cash on hand and term loan borrowings', which on our arithmetic takes funded net debt to roughly $149 billion and leverage from about 2.60x to about 3.07x of adjusted EBITDA; capital expenditure is 17.5% of revenue and $10,577M was spent in six months, so free cash flow is structurally hostage to network spending; the Legacy segment is declining 25.9% and its operating margin fell from 43.6% to 32.0%; tangible book value per share is NEGATIVE $10.47; and the June quarter carried an asset abandonment charge attributed to 'the reprioritization of our spectrum strategy' immediately before closing a $23 billion spectrum purchase. Not a 4, because $23 billion of new obligations is about to land. Not a 6, because nothing in this file threatens the dividend or the coverage."

Growth Quality5/10Moderate

"Rated 5 — the headline growth is poor and the composition is much better than the headline, which is the single most important analytical point in this dive. Consolidated revenue rose 2.3% in the June quarter to $31,558M and 2.6% for the half to $63,064M. Underneath: the ADVANCED CONNECTIVITY segment — $28,615M, 90.7% of revenue — grew 4.1%, with wireless service up 3.3%, ADVANCED HOME INTERNET UP 27.3% to $2,926M, business fibre and advanced connectivity up 10.0%, and equipment flat; its operating income grew 20.3% to $7,345M with operating margin from 22.2% to 25.7% and EBITDA margin from 40.5% to 42.0%, helped by depreciation falling 6.9% as fully depreciated legacy assets roll off. Latin America grew 16.1% with Mexican postpaid subscribers up 20.7% to 7.457 million. The offset is deliberate: the LEGACY segment fell 25.9% to $1,632M as the copper network is decommissioned, and its operating income fell 45.5%. Consensus adjusted EPS of $2.339 for 2026 (14 analysts) and $2.558 for 2027 (14) is +9.4%, and $2.910 for 2028 (7) is a further +13.8%. A 5 rather than a 6 because consolidated revenue growth is 2.3% and nothing changes that arithmetic while Legacy is still 5% of the base; a 5 rather than a 4 because 20.3% operating-income growth in 91% of the company is not what a declining business looks like."

Exponential Potential3/10Low

"Rated 3 — a share-shift story inside a mature market, with one genuinely fast-growing line that is too small to change the answer. The fast line is real: advanced home internet revenue of $2,926M grew 27.3% year on year and 27.3% for the half, driven by fibre and fixed-wireless displacing cable and legacy copper. Business fibre and advanced connectivity grew 10.0%. And AT&T's February 2026 acquisition of Lumen's Mass Markets fibre business for $5,756M is a deliberate purchase of more of exactly that. But advanced home internet is 9.3% of revenue and the market it takes share in is not growing; the wireless service line, at 55.2% of revenue, grew 3.3%. The knowledge base's most useful contribution is the one asset-level observation, from raoul_pal at conviction 69: wireless spectrum 'bought at auction and held' has delivered double-digit annualised returns — which is a statement about an appreciating balance-sheet asset ($129.1 billion of licences at 2026-06-30, with $23 billion more contracted) rather than about an earnings slope. Against that sit two claims describing satellite direct-to-cell as a structural threat, one of which explicitly says indebted telcos 'can't build.' A 3: real operating leverage, no exponential, and a technology-substitution risk that is named in our own research and not quantifiable from any source in the contract."

Fair value$26.90 $19.00–$30.50

What does “fair value” mean?

Fair value is Synthos’s estimate of what one share is worth today, based on our model of the company’s future cash generation and the risks to it. It is not a price target or a prediction of where the stock trades next quarter — it’s the price where we think risk and reward are balanced. Above it, you’re paying for outcomes better than our base case; below it, the market is offering a margin of safety.

The model’s inputs and weightings are proprietary — they’re the product. What we publish is the output, We don’t publish a reverse-DCF cross-check for pre-profit companies — negative or missing earnings break that math — so take this number on our modeling alone.

What do the 5 tiers mean? (Core · Tactical · Watch · Hold · Avoid)

Buy — CoreOwn it as a foundation — start or add now, size it for years, let dips be gifts.

Buy — TacticalGood price + confirmed trend + a defined exit — buy the setup, not a marriage.

WatchWe want the business, just not at this price/setup — act only when the listed trigger hits.

HoldFine to keep if you own it — no reason to buy more; new money does better elsewhere.

AvoidDon't own it — the problem is the business or the expectations, so a cheaper price won't fix it.

The Road Ahead

What we expect to matter in each window, and the evidence that would prove us wrong.

Short term 0-6 months

Neutral

Driver: "The valuation is compelling and the chart is not yet confirming. At $23.38 the stock is 21.1% below its 52-week high of $29.62 and 14.2% above its low of $20.48 — the 27th percentile of its annual range — and, critically, it is 7.1% BELOW its 200-day moving average of $25.17 while 2.7% above its 50-day of $22.78. The 50-day is well beneath the 200-day. That is a downtrend in the early stage of repair, and RSI at 66.9 says the repair is already extended in the short run despite the price being a fifth below its high. Relative performance is uniformly poor and has not stopped: three-month −10.4% against SPY +7.6%, six-month −11.1% against +11.1%, twelve-month −15.7% against +24.3%. The offsetting facts are that the June quarter beat on adjusted EPS by 10.2% ($0.65 against $0.59), that five of the last five prints have beaten, and that a 4.75% dividend pays you to wait. The near-term swing factor is the EchoStar close: management expected it by the end of July 2026, the archive does not confirm it, and the market will price the funding mix when it does."
What we’re watching: "Whether the EchoStar transaction closed and on what funding mix. Management states it will use 'a combination of cash on hand and term loan borrowings' against a $17,570M cash balance the 10-Q says 'remained elevated as we anticipate the completion of our pending transaction' — the more of the $23 billion that comes from term loans rather than cash, the larger the leverage step. The 2026-10-28 print against $0.62 of adjusted EPS and $31,687.3M of revenue, and within it the Advanced Connectivity operating margin, which went from 22.2% to 25.7% year on year. Advanced home internet revenue, growing 27.3% and the only genuinely fast line in the company. Whether the Legacy decline stays near 25% or accelerates. And the asset abandonment charge disclosed in the June quarter as arising from 'the reprioritization of our spectrum strategy' — $286M of impairments, abandonments and restructuring ran through Corporate & Other, and the explanation for it sits oddly beside a $23 billion spectrum acquisition."
Confidence: Medium

Medium term 6-24 months

Tailwind

Driver: "The medium-term case is operating leverage in the segment that is 91% of the company, and it is already visible rather than forecast. Advanced Connectivity operating income grew 20.3% in the June quarter and 17.6% for the half on revenue growth of 4.1% and 4.4% — a 4-to-1 ratio — because operations and support expense grew only 1.4% while depreciation FELL 6.9% as fully depreciated legacy assets roll off the books. Operating margin went 22.2% to 25.7% and EBITDA margin 40.5% to 42.0%. That mechanism has years to run: AT&T is decommissioning a copper network whose associated depreciation is largely spent, and replacing it with fibre and fixed wireless where advanced home internet revenue is compounding at 27.3%. The February 2026 acquisition of Lumen's Mass Markets fibre business for $5,756M adds passings directly. Consensus adjusted EPS of $2.558 for 2027 and $2.910 for 2028 implies 9.4% and 13.8% growth off a 2.3% revenue base — which is precisely the margin-expansion arithmetic above, and it is corroborated by two quarters of actual results rather than by a model."
What we’re watching: "Leverage after EchoStar. Funded net debt at 2026-06-30 was $126,384M, roughly 2.60x our adjusted trailing EBITDA of about $48.7 billion; adding approximately $23 billion of spectrum funded with a mix of cash and term loans takes that to roughly $149 billion and about 3.07x. AT&T has spent the last four years deleveraging and this reverses part of it. Whether the $129,123M of licences on the balance sheet — 30.1% of total assets and rising — earns a return; raoul_pal's claim that spectrum bought at auction and held delivers double-digit annualised returns is the bull case for that asset, and it is a claim about resale value rather than about operating profit. Whether the buyback continues at the recent pace: 158.7 million shares, 2.3% of the company, were retired in the first half of 2026 for $3,917M, and the EchoStar funding is the obvious competing use of that cash. Whether direct-to-cell satellite services materially erode wireless service revenue, which is 55.2% of the company and which two claims in our knowledge base identify as the structural threat."
Confidence: Medium

Long term 2+ years

Neutral

Driver: "Long term this is a utility-like asset trading at a utility-like multiple with a genuine, if slow, technology transition inside it. The favourable version: AT&T ends the copper decommissioning with a fibre-and-wireless network of far lower maintenance depreciation, a licence portfolio of $129 billion in an asset class our knowledge base says appreciates, a 4.75% dividend covered 2.5 times by free cash flow, and a share count shrinking 2.3% a half. The unfavourable version is the one our own research library actually contains: two independent claims describing satellite direct-to-cell — Starlink specifically — as a base-infrastructure layer 'going exponential' while 'indebted telcos can't build.' The honest position is that neither has been decided by anything in the filings, and that the second is a decade-scale question rather than a quarterly one. What can be said with confidence is that the mature-market arithmetic is unforgiving: consolidated revenue has grown from $122,428M in 2023 to $125,648M in 2025, a 1.3% compound rate, and no reasonable set of assumptions makes that materially faster."
What we’re watching: "Whether wireless service revenue growth stays positive; at 55.2% of the company it is the load-bearing line and it grew 3.3%. Whether the licences carried at $129.1 billion are ever impaired — they are indefinite-lived and tested rather than amortised, and a write-down would be the clearest possible refutation of the spectrum-appreciation thesis. Whether the pension obligation, named in the one bearish name-level claim in our knowledge base, becomes a cash call; the June balance sheet shows a $8,267M postemployment benefit obligation and the half included a $100M voluntary contribution. Whether direct supplier financing — which the 10-Q discloses reduced operating cash flow by $272M in the half against $2,146M a year earlier, on payment terms extended by up to approximately 120 days — is a durable working-capital source or a borrowed one. And succession: John T. Stankey has led the company since 2020 and the file contains no succession disclosure."
Confidence: Low

Exponential Potential

Exponential Potential3/10Low

What could take this further than the base case — and where the market may be underpricing it. Full forward-growth and acceleration math in Deeper analysis, §4 below.

Deeper analysis

Technicals, fundamentals, valuation, and the full expert-claim evidence panel — the detail behind the numbers above.

Reference table

Street consensus	$27.13 (+16.0%) · median $28.00 (+19.8%) · high $31 · low $20 (−14.5%) · 1 strong buy / 25 buy / 30 hold / 6 sell across 62 analysts, consensus "Hold"
Valuation	10.25x TTM ADJUSTED EPS ($2.28) · 10.0x 2026E · 9.1x 2027E · 8.0x 2028E · 8.4x 2025 free cash flow · vendor `priceToEarningsRatioTTM` of 7.72x REJECTED — trailing GAAP is inflated by a roughly $5B one-off in the September 2025 quarter
Segment structure — REORGANISED in 2026, and the payload cannot see it	The 10-Q reports Advanced Connectivity, Legacy and Latin America, replacing the former Communications structure. `seg_prod` and `seg_geo` are BOTH EMPTY in the vendor payload, so the reorganisation is invisible from the data alone
The segment that matters	Advanced Connectivity: revenue $28,615M (+4.1%), 90.7% of the total; operating income $7,345M (+20.3%); operating margin 22.2% → 25.7%; EBITDA margin 40.5% → 42.0%; advanced home internet +27.3%; depreciation −6.9%
Debt — the vendor double-counts $18.9B of leases	`bal_a` reports `totalDebt` of $173,986M and `netDebt` of $155,752M. The 10-Q shows funded debt of $136,100M at 2025-12-31 and $143,954M at 2026-06-30 — the vendor's `longTermDebt` of $146,032M equals the filed $127,089M plus the $18,943M of operating lease liabilities that ALSO appear in `capitalLeaseObligations`. Funded net debt at 2026-06-30 is $126,384M, not $155,752M
The transaction ahead	EchoStar: approximately $23,000M for 600 MHz and 3.45 GHz FCC licences, agreed 2025-08-25, expected by management to close by the end of July 2026, funded with "a combination of cash on hand and term loan borrowings." On our arithmetic that takes funded leverage from ~2.60x to ~3.07x adjusted EBITDA
Conviction	Low — 7 KB claims from 7 different channels, so no concentration to test. Only one is name-level and it is bearish and three years old. The most useful is thematic (spectrum appreciation)
Technicals	−21.1% from the 52-week high of $29.62, +14.2% above the low of $20.48; −7.1% BELOW the 200-DMA of $25.17, +2.7% above the 50-DMA; RSI 66.9; MACD +0.36; 12-month −15.7% vs SPY +24.3%

What the experts actually said 5 traceable claims on T · showing the highest-conviction voices

“Legacy carriers and satellite ISPs (AT&T, Verizon, Hughesnet, Viasat) have been decimated by Starlink, and a direct-to-cell handset launch will further accelerate their subscriber losses.”

David Friedbergbearishconviction 602026-08-08all_in-muRIXCDw-k0:f487988f09

Every claim reconciles to a real claim_id in the Synthos knowledge base — this is the evidence the verdict is built on, not vibes. Management (the company itself) is shown but half-weighted; one cautionary voice is included on purpose.

Price & moving averages 12 months · 50 & 200-day averages · 52-week range

Solid line = price · dashed line = 50-day average · dotted line = 200-day average · the two thin horizontal lines mark the 52-week high and low. Price above both averages is an uptrend.

Bollinger Bands 20-day average ± 2 standard deviations

The shaded band widens when the stock gets more volatile. Riding the upper edge = strong momentum (sometimes stretched); the lower edge = weak / potentially oversold.

RSI (14) momentum gauge · 0–100

Above 70 (overbought zone, shaded) = overbought, below 30 (oversold zone, shaded) = oversold. Currently 69.

MACD 12 / 26 / 9 · trend & momentum

The MACD line crossing above the signal line (bars flip to the up color) = momentum turning up; crossing below (bars flip to the down color) = turning down. Bar height = the size of that gap.

Relative performance vs S&P 500 & its sector (XLC (sector)), set to 100 a year ago

Solid = T · dashed = S&P 500 · dotted = XLC (sector). A rising line means it is beating that benchmark — the sector line shows whether it is a leader or laggard within its own group.

Forward revenue & earnings actual → estimate · "FY" = fiscal year, "E" = estimate

Darker bars = actual results, brighter = analyst estimates. Taller bars to the right = expected growth.

Key stats an RIA wants

Price$26.00

Market cap$178B

P/E trailing8×

P/E FY26E / FY27E11× / 10×

EV / Sales2.5×

EV / EBITDA6.0×

Gross margin59.7%

Net margin16.9%

Dividend yield4.27%

Beta0.417

52-wk range$20 – $30

RSI(14)76

50 / 200-DMA$23 / $25

12-mo return+-10% (SPY +19%)

Street target$27 ($20–$30)

Analyst grades25 Buy · 30 Hold · 6 Sell

FMP ratingA-

Next earnings2026-10-28 (third-quarter 2026 earnings, 85 days away; vendor consensus adjusted EPS $0.62 on revenue of $31,687.3M). Second-quarter results were released 2026-07-22 and the 10-Q filed the same day, thirteen days before this dive. Ahead of that, the EchoStar spectrum transaction was expected by management to close by the end of July 2026 and the archive does not confirm whether it has; a five-tranche note issue settled 2026-08-03 and was disclosed in an 8-K filed on the dive date itself.

1. What the business is, and the disclosure change the payload cannot see

AT&T Inc., Delaware-incorporated, CIK 0000732717; chief executive John T. Stankey; approximately 133,030 employees; listed on the NYSE and the Chicago Stock Exchange (the 8-K cover page tags both exch:XNYS and exch:XCHI).

In 2026 AT&T reorganised its reportable segments, and this is the single most consequential fact in the file for anyone reading the vendor data. The 10-Q for the June quarter reports three segments — Advanced Connectivity, Legacy and Latin America — replacing the former Communications structure with its Mobility, Business Wireline and Consumer Wireline units. The goodwill note records the mechanics: "The goodwill from our previous Consumer Wireline and Mobility reporting units within the Communications segment was fully assigned to the reporting units comprising the Advanced Connectivity segment."

The vendor's seg_prod and seg_geo blocks are BOTH EMPTY. There is no segment or geographic data of any kind in the payload. Every segment figure in this dive comes from the 10-Q, and a reader working from the vendor file alone would have no way of knowing the company had reorganised at all.

Second quarter 2026 by segment (10-Q filed 2026-07-22; US$M)

	Advanced Connectivity	Legacy	Latin America	Total Segment	Corporate & Other	AT&T Inc.
Wireless service	17,413	—	780	18,193	—	18,193
Advanced home internet	2,926	—	—	2,926	—	2,926
Business fibre and advanced connectivity	1,946	—	—	1,946	—	1,946
Business transitional and other	1,042	—	—	1,042	—	1,042
Other service	151	1,632	—	1,783	87	1,870
Total service	23,478	1,632	780	25,890	87	25,977
Equipment	5,137	—	444	5,581	—	5,581
Operating revenues	28,615	1,632	1,224	31,471	87	31,558
Operations and support	16,583	1,109	997	18,689	430	19,119
Asset impairments, abandonments and restructuring	—	—	—	—	286	286
Transaction, legal and other costs	—	—	—	—	149	149
Depreciation and amortisation	4,687	—	189	4,876	90	4,966
Operating expenses	21,270	1,109	1,186	23,565	955	24,520
Operating income	7,345	523	38	—	—	7,038

The three segments plus Corporate & Other reconcile exactly to consolidated revenue of $31,558M.

Advanced Connectivity — 90.7% of the company, and the reason to own it

	Q2 2026	Q2 2025	Change	H1 2026	H1 2025	Change
Wireless service	$17,413M	$16,853M	+3.3%	$34,354M	$33,504M	+2.5%
Advanced home internet	$2,926M	$2,299M	+27.3%	$5,725M	$4,497M	+27.3%
Business fibre and advanced connectivity	$1,946M	$1,769M	+10.0%	$3,828M	$3,524M	+8.6%
Business transitional and other	$1,042M	$1,249M	−16.6%	$2,125M	$2,543M	−16.4%
Total service revenues	$23,478M	$22,334M	+5.1%	$46,341M	$44,394M	+4.4%
Equipment	$5,137M	$5,163M	−0.5%	$10,745M	$10,295M	+4.4%
Total segment revenues	$28,615M	$27,497M	+4.1%	$57,086M	$54,689M	+4.4%
Operations and support	$16,583M	$16,356M	+1.4%	$33,496M	$32,603M	+2.7%
Depreciation and amortisation	$4,687M	$5,035M	−6.9%	$9,392M	$10,008M	−6.2%
Operating income	$7,345M	$6,106M	+20.3%	$14,198M	$12,078M	+17.6%
Operating margin	25.7%	22.2%	+350 bp	24.9%	22.1%	+280 bp
EBITDA margin	42.0%	40.5%	+150 bp	41.3%	40.4%	+90 bp

Three things follow and they are the whole constructive case.

First, revenue grew 4.1% and operating income grew 20.3% — a five-to-one ratio. The arithmetic is entirely explicit in the segment table: operations and support expense grew 1.4% against revenue growth of 4.1%, and depreciation fell 6.9%. The 10-Q explains the depreciation line directly: "lower depreciation from fully depreciated legacy assets, partially offset by ongoing capital spending for strategic initiatives such as fiber and network upgrades."

Second, the fastest line in the company is home internet, and it is compounding. $2,926M, up 27.3% in the quarter and 27.3% for the half — an identical rate in both periods, which is a stable trajectory rather than a comparison artefact. This is fibre and fixed wireless taking share from cable and from AT&T's own copper.

Third, the margin expansion is not a one-quarter event. Operating margin rose 350 basis points in the quarter and 280 for the half; EBITDA margin 150 and 90. Both periods, both measures, same direction.

Legacy — a deliberate 26% decline

	Q2 2026	Q2 2025	Change	H1 2026	H1 2025	Change
Segment revenues	$1,632M	$2,202M	−25.9%	$3,400M	$4,570M	−25.6%
Operations and support	$1,109M	$1,243M	−10.8%	$2,265M	$2,592M	−12.6%
Operating income	$523M	$959M	−45.5%	$1,135M	$1,978M	−42.6%
Operating margin	32.0%	43.6%	−1,160 bp	33.4%	43.3%	−990 bp

The 10-Q is explicit that this is a choice: revenue fell "driven by lower demand for legacy services, which we expect to continue as we decommission our copper-based legacy network", and costs fell "primarily driven by lower personnel and other costs resulting from the decommissioning of our legacy network."

But note the margin, because it is the honest counterweight to the constructive case. Revenue fell 25.9% and costs fell only 10.8%, so the segment's operating margin collapsed from 43.6% to 32.0%. Legacy is being wound down less profitably than it operated, and at $1,632M a quarter there is roughly $6.5 billion of annualised revenue still to go, currently carrying about $2.1 billion of annualised operating income. The bull case requires Advanced Connectivity's operating leverage to outrun that runoff, and in the June quarter it did — $1,239M of Advanced Connectivity operating-income growth against $436M of Legacy decline — but the margin on what remains is deteriorating.

Latin America — small, growing fast, barely profitable

Revenue $1,224M, +16.1%; operating income $38M, −17.4%; operating margin 3.1%. Mexican postpaid subscribers grew 20.7% to 7.457 million while prepaid fell 9.2% to 15.829 million — a deliberate mix shift toward higher-value customers that is not yet visible in profit. At 3.9% of revenue and 0.5% of operating income, the segment does not move the answer.

2. Earnings quality — why the vendor's 7.7x P/E is wrong

The vendor reports priceToEarningsRatioTTM of 7.716x and netIncomePerShareTTM of $3.098. Both are arithmetically correct on trailing GAAP and both are misleading.

Trailing GAAP net income is $21,494M (Q3 2025 $9,314M + Q4 2025 $3,724M + Q1 2026 $3,829M + Q2 2026 $4,627M). The September 2025 quarter alone contributed $9,314M — more than double any other quarter in the series — on EBITDA of $17,670M against a run-rate of $11.2B to $13.0B. That is roughly $5 billion of non-recurring benefit in a single quarter, and it is what makes the trailing GAAP figure and every ratio derived from it unusable.

The adjusted series is clean and is what estimates are struck on. The earnings-calendar actuals — $0.54, $0.54, $0.52, $0.57, $0.65 — are adjusted EPS, and they sum to a trailing twelve-month figure of $2.28. The honest trailing multiple is 10.25x, not 7.72x. Every forward multiple in this dive uses the same adjusted basis, on which consensus is $2.339 for 2026 and $2.558 for 2027.

The same problem contaminates the effective tax rate and the leverage ratio. effectiveTaxRateTTM reads 11.66% against a normalised rate nearer 20%; netDebtToEBITDATTM of 2.704x is computed on a trailing EBITDA of about $53.7 billion that includes the September 2025 item. Our adjusted trailing EBITDA is approximately $48.7 billion, and the leverage figures below use it.

Five consecutive adjusted-EPS beats: +1.9%, +0.6%, +12.4%, +3.3%, +10.2%. The June quarter's $0.65 against $0.59 is the largest.

3. Balance sheet — the vendor double-counts $18.9 billion of leases

As at 2026-06-30, from the 10-Q filed 2026-07-22 (US$M):

	2026-06-30	2025-12-31
Cash and cash equivalents	$17,570	$18,234
Goodwill — net	$63,865	$63,425
Licences — net	$129,123	$128,148
Other intangible assets — net	$6,063	$5,254
Operating lease right-of-use assets	$22,781	$22,642
Total assets	$428,359	$420,198
Debt maturing within one year	$9,323	$9,011
Long-term debt	$134,631	$127,089
Total debt (per the 10-Q's own liquidity table)	$143,954	$136,100
Operating lease liabilities (non-current)	$18,934	$18,943
Common shares issued / treasury / outstanding	7,620,748,598 / 741,978,242 / 6,878,770,356	7,620,748,598 / 583,246,242 / 7,037,502,356

The defect, stated precisely. The vendor's bal_a at 2025-12-31 reports longTermDebt of $146,032M. The filing reports long-term debt of $127,089M. The difference is $18,943M — exactly the operating lease liability. And the vendor separately reports capitalLeaseObligations of $18,943M, then computes totalDebt as short-term debt $9,011M plus longTermDebt $146,032M plus capitalLeaseObligations $18,943M = $173,986M.

> The operating lease liability is counted twice.

> Filed funded debt at 2025-12-31: $136,100M. Vendor totalDebt: $173,986M. Overstatement: $37,886M, or 27.8%.

> Filed funded net debt at 2025-12-31: $117,866M. Vendor netDebt: $155,752M. Overstatement: $37,886M, or 32.1%.

And the file then contradicts itself in the usual way. km_ttm.enterpriseValueTTM of $305,527M less the market capitalisation of $160,209M implies net debt of $145,318M — which is exactly the 2026-06-30 funded net debt of $126,384M plus the $18,934M of non-current operating lease liabilities. The enterprise value is correct and current; the balance sheet feeding it is neither.

The rebuilt enterprise value, on two bases

> Market capitalisation $160.209B

> + total funded debt $143.954B − cash $17.570B

> = funded enterprise value $286.593B

> + non-current operating lease liabilities $18.934B

> = lease-inclusive enterprise value $305.527B (matches the vendor exactly)

	Funded basis	Lease-inclusive
Enterprise value	$286.59B	$305.53B
Net debt	$126.38B	$145.32B
Net debt / adjusted TTM EBITDA (~$48.7B)	2.60x	2.98x
EV / adjusted TTM EBITDA	5.89x	6.27x
EV / TTM revenue ($127.24B)	2.25x	2.40x

Two further balance-sheet notes.

The intangibleAssets line falls from $132,290M (2024) to $5,254M (2025) and this is NOT an economic event. The 10-Q reports Licences — net of $128,148M at 2025-12-31, separately from Other Intangible Assets — net of $5,254M. The vendor's 2024 figure included licences; its 2025 figure does not. Total non-current assets rose over the same period, confirming that nothing left the balance sheet. A reader comparing the two years without the filing would conclude AT&T had written off $127 billion of intangibles.

Tangible book value per share is NEGATIVE $10.47, correctly reported by the vendor. Goodwill of $63,865M plus licences of $129,123M plus other intangibles of $6,063M is $199,051M against stockholders' equity that the payload puts at $110,533M. For a company whose licences are indefinite-lived and tested rather than amortised, this is a description of the accounting rather than of the risk — but it is the reason priceToBookRatioTTM of 1.47x carries no information.

4. Cash flow, capital returns, and the transactions in flight

Year	Operating cash flow	Capex	Free cash flow (vendor basis)	Buybacks
2022	$32,023M	$19,626M	$12,397M	$890M
2023	$38,314M	$17,853M	$20,461M	$194M
2024	$38,771M	$20,263M	$18,508M	$215M
2025	$40,284M	$20,842M	$19,442M	$6,575M

Free cash flow of $19,442M on a $160.2B market capitalisation is a 12.1% free-cash-flow yield, and it is the most striking single number in this dive. Operating cash flow of $40,284M on $125,648M of revenue is a 32.1% margin. Capital expenditure at 16.6% of revenue is the price of that cash flow and it does not fall.

First-half 2026, from the 10-Q: operating cash flow $18,396M against $18,812M a year earlier — "with the prior year benefiting from $1,675 of cash received from DIRECTV, net of related tax payments" — capital expenditure $10,577M, vendor financing payments $643M, giving capital investment of $11,220M and free cash flow on the company's own definition of $7,176M for the half.

The buyback has become material. Treasury shares went from 583,246,242 at 2025-12-31 to 741,978,242 at 2026-06-30 — 158,732,000 shares retired in six months, 2.3% of the shares outstanding — at a treasury cost increase of $3,917M. That is a run-rate of roughly 4.6% of the company a year, and it is a change: buybacks were $194M and $215M in 2023 and 2024 before $6,575M in 2025.

Dividend. dividendPerShareTTM of $1.11 gives a 4.75% yield, and this is one of the few names in this batch where the trailing figure is the run-rate — AT&T has paid $0.2775 quarterly for several years. profile.lastDiv is null. Dividends payable on the balance sheet were $1,945M at 2026-06-30, consistent with roughly $7.7 billion annualised. Free cash flow of $19,442M covers that 2.5 times, and dividendPayoutRatioTTM of 37.3% is computed against the one-off-inflated GAAP figure — against adjusted trailing EPS of $2.28 the payout is 48.7%.

Total shareholder yield is therefore approximately 9.3% — 4.75% dividend plus roughly 4.6% of net buyback at the recent pace. That is the highest in this batch by a wide margin.

The three transactions, from the 10-Q

UScellular spectrum licences — closed January 2026, $1,018M.
Lumen's Mass Markets fibre business — closed February 2026, $5,756M (of which $1,656M in continuing-operations investing and $4,100M in discontinued-operations investing). This is a direct purchase of fibre passings feeding the advanced home internet line growing 27.3%.
EchoStar spectrum — agreed 2025-08-25, approximately $23,000M, NOT YET REFLECTED. Verbatim: "On August 25, 2025, we agreed to purchase Federal Communications Commission (FCC) licenses in the 600 MHz and 3.45 GHz bands from EchoStar Corporation for approximately $23,000, subject to certain adjustments... We signed a short-term spectrum manager lease on the 3.45 GHz spectrum, which was deployed in cell sites covering nearly two-thirds of the U.S. population. We expect to close this transaction by the end of July 2026 and will fund the acquisition using a combination of cash on hand and term loan borrowings."

The 10-Q explains the elevated cash balance in the same breath: "Our cash balance at June 30, 2026 remained elevated as we anticipate the completion of our pending transaction with EchoStar Corporation." $17,570M of cash is being held against a $23,000M obligation.

What that does to leverage, as arithmetic with the method stated. If AT&T applies, say, $12 billion of cash and borrows $11 billion of term loans, funded debt rises to about $155 billion and cash falls to about $5.6 billion, giving funded net debt of roughly $149.4 billion — approximately 3.07x our adjusted trailing EBITDA of $48.7 billion, against 2.60x today. The mix is management's choice and the file does not disclose it; the range of outcomes is 2.9x to 3.1x, and no greater precision is available.

Note also the two disclosures that sit oddly together. The June quarter carried $286M of asset impairments, abandonments and restructuring, and the MD&A attributes the increase to "an asset abandonment charge associated with the reprioritization of our spectrum strategy." Abandoning spectrum assets in the same quarter as closing a $23 billion spectrum purchase is not a contradiction — it is presumably the 3.45 GHz lease replacing owned alternatives — but the file does not say so, and it is worth flagging that the two disclosures are not reconciled anywhere.

One working-capital item worth naming. The 10-Q discloses that AT&T seeks supplier payment terms "on 90-day or greater terms" and, for handset inventory, arrangements that "extend the stated payment terms by up to approximately 120 days, with an average of 85 days outstanding." The net impact of direct supplier financing reduced operating cash flow by $272M in the first half of 2026 against $2,146M a year earlier. daysOfPayablesOutstandingTTM of 271.0 and a cash conversion cycle of −229.7 days are the balance-sheet signature of that policy. It is disclosed, it is legal, and it means a portion of reported operating cash flow is financed by suppliers rather than generated.

5. Valuation — priced in or room?

At $23.38, market capitalisation $160.21B, funded enterprise value $286.59B:

	Trailing (TTM to 2026-06-30)	2026E	2027E	2028E
Revenue	$127,239M	$129,236M (17 analysts)	$131,939M (17)	$134,728M (15)
Revenue growth	—	+2.9% (vs 2025 actual)	+2.1%	+2.1%
Adjusted EPS	$2.28	$2.339 (14)	$2.558 (14)	$2.910 (7)
GAAP diluted EPS (one-off inflated)	$3.10	—	—	—
P/E on adjusted EPS	10.25x	10.00x	9.14x	8.03x
EV / adjusted EBITDA (~$48.7B TTM)	5.89x	—	—	—
Price / 2025 free cash flow	8.24x	—	—	—
Free cash flow yield	12.1%	—	—	—
Dividend yield	4.75%	—	—	—
Total shareholder yield	~9.3%	—	—	—

Estimate coverage is the best in this batch — 17 analysts on revenue and 14 on EPS for both 2026 and 2027, and 15 and 7 on 2028. The 2029 EPS row rests on two analysts and is excluded.

est.ebitdaAvg and est.ebitAvg do not trip the standard defect signatures here, but they are still not used. The ratios are not fixed — ebitdaAvg/revenueAvg runs 53.1%, 53.1%, 53.1%, 53.1%, 53.1%, 53.1% across 2025 to 2030, which IS effectively fixed to three significant figures, and ebitAvg/revenueAvg likewise sits at 33.6% throughout. That is the fixed-ratio fabrication signature. Independently, the 2025 ebitdaAvg of $66,439.9M against our adjusted trailing figure of roughly $48.7B is 36% too high — AT&T's EBITDA has never been $66 billion. Both rows are rejected and all forward valuation runs on epsAvg.

Peer context. The vendor peer set is genuinely apposite for once: Verizon ($195.8B), T-Mobile US ($190.1B), Comcast ($88.5B), América Móvil ($74.7B), Charter ($20.6B), Rogers ($18.3B), plus Disney ($170.5B), Telefônica Brasil ($8.7B), Telkom Indonesia ($15.1B) and Telefônica ($19.7B). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The structural observation available: AT&T at $160.2B is smaller than both Verizon and T-Mobile despite similar or larger revenue, which is the market's summary judgement on relative growth and leverage.

5a. What today's price assumes (the inversion)

At $23.38 — 10.0x the 2026 consensus and 9.1x 2027 — the price embeds the following falsifiable claims, each with a number and a date:

Adjusted EPS of $2.339 in 2026 and $2.558 in 2027. (Consensus-derived; 14 analysts each.) The first half delivered $1.22 ($0.57 + $0.65), so the second half needs $1.12 — below the first half, which is consistent with the seasonal pattern. The nearest test is 2026-10-28 against $0.62, a figure the company beat by 10.2% last quarter.
Revenue compounds only 2.1% a year through 2028. (Consensus-derived; 17 and 15 analysts.) The price is NOT paying for growth; it is paying for the 4.75% dividend and the 12% free-cash-flow yield. That is why the multiple is 9x rather than 15x, and it means the thesis does not require the revenue line to do anything.
Advanced Connectivity's margin expansion continues. (Our number, from the segment table.) Operating margin went 22.2% to 25.7% and EBITDA margin 40.5% to 42.0%. This is the most fragile assumption in the price — not because it is unlikely, but because it is the only source of earnings growth in a business whose revenue grows 2%, and it depends on depreciation continuing to fall (−6.9% year on year) as copper assets finish depreciating. When that runs out, so does the earnings growth.
The EchoStar transaction closes at approximately $23 billion without a credit consequence. (Filing-disclosed amount; our assumption on the consequence.) Funded leverage moves from about 2.60x to about 3.07x. The price does not appear to contain a ratings action, and the 8-K of 2026-08-04 shows AT&T issuing five tranches of notes — including a 7.050% coupon due 2052 — on 2026-08-03 without difficulty.
The market keeps paying 9-10x forward adjusted earnings. (Our number.) At 8x the 2027 consensus the stock is $20.46 — essentially the 52-week low; at 12x it is $30.70. Every dollar of the answer is in a multiple that has ranged from roughly 8x to roughly 11.6x within the last twelve months.

5b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: adjusted EPS growth (+9.4%, from 2026E $2.339 to 2027E $2.558) + multiple drift (modest EXPANSION, from 10.0x to roughly 10.5x on the respective forward year, +5%) + shareholder yield (+4.75% dividend, plus buyback) ≈ +19% to +20%, against a base fair value that is +15.1% before the dividend.

Our base assumes MILD multiple expansion and we should be explicit, because assuming expansion is the leg we would normally flag as fragile. Today the market pays 10.0x on the current year. Our $26.90 is 10.5x the 2027 estimate. The justification is specific rather than generic: the segment that is 91% of revenue is growing operating income at 20% with margin expanding 350 basis points, the copper decommissioning that suppresses the headline revenue line is a finite process, and the shares de-rated 15.7% over twelve months while that was happening. We are assuming a half-turn of re-rating on evidence of improving earnings quality, not a re-rating on hope.

But the honest framing is that the multiple is not where the return comes from. If the multiple simply HELD at 10.0x on the 2027 consensus the price would be $25.58 (+9.4%), and the dividend takes the total to +14%. The base case adds a half turn on top. A 4.75% dividend and a 4.6% buyback pace mean that even at a flat multiple and flat earnings, the shareholder yield alone is 9.3%. That is the actual argument.

5c. Variant perception (where we differ, what would surprise)

We differ from every screen on what this company owes. The published totalDebt of $173,986M and netDebt of $155,752M double-count $18,943M of operating lease liabilities, which appear in longTermDebt and again in capitalLeaseObligations. Funded net debt at 2026-06-30 is $126,384M. Any screen ranking AT&T on leverage is ranking it 32% more indebted than it is. This correction cuts in the stock's favour and is the single largest data finding in this dive.
We differ from the headline on where the growth is, and the segment reorganisation is why. Consolidated revenue growth of 2.3% is the number in every summary. Advanced Connectivity — 90.7% of revenue — grew 4.1% with operating income up 20.3% and margin up 350 basis points, while the Legacy segment AT&T is deliberately switching off fell 25.9%. Neither figure is available anywhere in the vendor payload, because seg_prod and seg_geo are both empty. Watchable number: Advanced Connectivity operating margin on 2026-10-28.
We differ from the vendor on the trailing multiple by a third. The published P/E of 7.72x rests on trailing GAAP earnings inflated by roughly $5 billion of non-recurring benefit in the September 2025 quarter. The honest figure is 10.25x. We report this because it cuts against the stock and because a 7.7x headline is exactly the kind of number that attracts the wrong kind of buyer.
Positive surprise that would force a re-rate: the EchoStar transaction closing predominantly with cash rather than term loans, keeping funded leverage below 2.9x; or advanced home internet holding above 25% growth for a third consecutive quarter, which would begin to make the fibre transition visible in the consolidated revenue line rather than only in the segment table.
Negative surprise that would break the thesis: Advanced Connectivity operating margin falling back below 24%; depreciation ceasing to decline, which removes the mechanism producing the operating leverage; a ratings action on the EchoStar funding; the buyback stopping to fund the spectrum purchase, which removes roughly 4.6 points of the 9.3% shareholder yield; or any impairment against the $129,123M of licences, which would directly refute the spectrum-appreciation argument the knowledge base supplies.

Synthos fair values

All three anchors are multiples of the 2027 consensus adjusted EPS distribution (mean $2.558, low $2.419, high $2.767, 14 analysts), the best-covered forward year in the file.

Bear ~$19.00 — 7.4x the 2027 consensus mean, cross-checked at 7.9x the consensus LOW of $2.419; roughly 7% below the 52-week low of $20.48. The scenario: the EchoStar funding forces a ratings action, the buyback is suspended, Advanced Connectivity's depreciation tailwind ends, and a 9x multiple becomes a 7.5x multiple on flat earnings. −18.7%.
Base ~$26.90 — 10.5x the 2027 consensus MEAN of $2.558. Cross-check: 11.5x the 2026 consensus; 9.2x the 2028 consensus. Sensitivity, stated openly: 10x gives $25.58 — up 9.4% — and 11x gives $28.14. The answer lives in a 10-11x band and no finer resolution is claimed. The scenario: Advanced Connectivity's margin expansion continues at a decelerating rate, Legacy runs off on schedule, EchoStar closes at roughly 3.0x leverage, the dividend is maintained and the buyback continues. +15.1%, plus a 4.75% dividend.
Bull ~$30.50 — 11.9x the 2027 consensus mean, cross-checked at 10.5x the 2028 consensus mean of $2.910 and 11.0x the 2027 consensus high. Just above the 52-week high of $29.62. The scenario: fibre and fixed wireless carry consolidated revenue growth above 3%, Advanced Connectivity margin passes 27%, EchoStar closes with minimal incremental debt, and the market re-rates AT&T toward the multiple of a growing utility rather than a shrinking telco. +30.5%.

Base is 15.1% above spot; asymmetry approximately 1.63:1 to the upside (18.7% down, 30.5% up). That ratio, plus a 4.75% dividend and a 9.3% total shareholder yield, is enough for a Tactical position. It is not enough for Core with $23 billion of spectrum about to be funded and a Hold consensus with six sells.

6. Knowledge base — seven claims, seven channels, one about the company

Raw hits: 7 (entity matches on AT&T). Used: 7. Discarded: 0. Channels: 7 — one claim each.

There is no concentration to test. Seven claims from seven different channels is the flattest distribution in this batch, and a sensitivity test is neither possible nor informative. The problem is not concentration; it is that only one claim is a view on AT&T as a security.

A case-sensitive bare-ticker sweep on the string "T" was NOT run and would be meaningless — a single capital letter appears in thousands of contexts. This is stated as a limitation of the method on this particular name rather than as an omission, and it means the entity-extraction is doing all the work here.

The one name-level claim, and it is bearish

> 2023-08-11 · bearish · conviction 58 · skill 0.9 · channel: mike_green · SPEAKER: Michael Green

> "Bias is the recession catalyst will be a credit event; highly levered names like Verizon and AT&T with pension obligations are hollowing out business to protect profits and could hit a refinancing wall."

The only claim in the lane with a named speaker and the only one that names AT&T as a security. It is three years old and its central prediction has not occurred: AT&T priced five tranches of notes on 2026-08-03, disclosed in an 8-K filed on the dive date itself, including a 7.050% coupon due 2052 — high-cost long money, but issued without difficulty. The pension point remains live in a small way: the balance sheet carries an $8,267M postemployment benefit obligation and the first half included a $100M voluntary contribution. We record the claim as unproven-and-ageing rather than wrong, and we note that "hollowing out business to protect profits" is a fair description of what the Legacy segment's margin decline from 43.6% to 32.0% looks like from outside.

The most useful claim, and it is thematic

> 2025-03-13 · bullish · conviction 69 · channel: raoul_pal · no named speaker

> "Wireless spectrum has consistently appreciated; bought at auction and held, it has delivered double-digit annualized returns."

This is the only claim in the lane that bears directly on the largest asset on AT&T's balance sheet. Licences are $129,123M at 2026-06-30 — 30.1% of total assets — and AT&T has contracted to add approximately $23 billion more from EchoStar plus $1,018M already closed from UScellular. If the claim is right, AT&T is buying an appreciating asset at a moment of forced-seller pricing; if it is wrong, it is adding $24 billion of indefinite-lived intangibles to a balance sheet whose tangible book is already negative $10.47 a share. We use it as a framing rather than as conviction — it is a statement about an asset class, not about this company's capital allocation, and it carries no named speaker.

The two disruption claims — and a stance-tag inversion worth flagging

> 2023-03-13 · tagged BULLISH · conviction 78 · channel: real_vision · no named speaker

> "Base-infrastructure layer (5G/6G, Starlink) is going exponential; indebted telcos can't build, so Starlink owns data from space outside sovereign control; India (Reliance) steps into 5G."

> 2025-09-10 · tagged BULLISH · conviction 55 · channel: all_in · no named speaker

> "Starlink could become a comprehensive global carrier (home Wi-Fi + direct-to-cell), though existing carriers won't be put out of business."

Both are tagged bullish by the distiller and both read as mildly BEARISH for AT&T. The stance tag records the speaker's view of Starlink, not of the entity AT&T that the extractor also attached. We flag the inversion explicitly rather than counting them as support, and we note the second claim's own hedge — "existing carriers won't be put out of business" — which is the more measured of the two. The first claim's phrase "indebted telcos can't build" is the sharpest bear point available anywhere in this dive, and it is worth setting against the fact that AT&T spent $10,577M of capital expenditure in six months and bought Lumen's fibre business.

The three generic claims

> 2021-01-04 · bearish · conviction 65 · skill 1.1 · lyn_alden — "Credit isn't correctly pricing risk because the Fed handicaps it... de facto yield-curve control in riskier markets."

> 2024-03-29 · bearish · conviction 65 · skill 1.2 · arthur_hayes — "Debt-laden old-economy corporates are squeezed in a high-rate environment, unlike cash-rich tech."

> 2026-03-07 · bullish · conviction 78 · we_study_billionaires — "Structure transactions as stock swaps, spin-offs, tracking stocks, and split-offs to legally defer taxes and compound more capital internally."

All three are macro or corporate-finance principles tagged to AT&T as an illustrative entity. None asserts anything about the company. The Hayes claim is the more relevant of the two bearish ones and is worth one observation: AT&T's June-quarter interest expense was $1,883M against $1,666M a year earlier, up 13.0%, so the high-rate squeeze it describes is visible in the numbers even as the company deleverages.

Conclusion. The Synthos knowledge base has essentially no view on AT&T. It has one ageing, unproven bearish claim about telco leverage, one genuinely useful thematic claim about spectrum as an appreciating asset, and two claims about satellite substitution that are tagged bullish for the wrong reason. Conviction rating: Low. Breadth: 7 nominal, effectively 2. Net conviction: neutral-negative. No conviction premium is applied, and the emptiness of the lane is an input to the decision to stop at Tactical.

7. Data integrity — what we rejected from the vendor file and why

Eleven findings. AT&T is the most definitionally damaged file in this batch after Deere: two segment tables entirely absent, a debt figure that double-counts, and a balance-sheet line that moves $127 billion for no economic reason.

1. totalDebt DOUBLE-COUNTS $18,943M of operating lease liabilities — REJECTED and rebuilt. bal_a at 2025-12-31 reports longTermDebt $146,032M, capitalLeaseObligations $18,943M and totalDebt $173,986M (= $9,011M short-term + $146,032M + $18,943M). The 10-Q reports long-term debt of $127,089M, and $146,032M − $127,089M = $18,943M exactly — the operating lease liability, which is then added AGAIN. The 10-Q's own liquidity table states total debt of $136,100M at 2025-12-31 and $143,954M at 2026-06-30. The vendor overstates funded debt by $37,886M (27.8%) and funded net debt by the same amount (32.1%). Consequentially rejected: netDebt, debtToAssetsRatioTTM, debtToEquityRatioTTM, debtToCapitalRatioTTM, debtToMarketCapTTM and interestDebtPerShareTTM.

2. seg_prod and seg_geo are BOTH EMPTY, across a segment reorganisation. No product, segment or geographic revenue data of any kind is in the payload — and AT&T reorganised its reportable segments in 2026 into Advanced Connectivity, Legacy and Latin America, a change the goodwill note in the 10-Q documents explicitly. Every segment figure in this dive comes from the 10-Q. A reader working from the payload alone could not know the segments had changed, let alone that 90.7% of revenue is growing operating income at 20.3%.

3. intangibleAssets falls from $132,290M to $5,254M between 2024 and 2025 — a RECLASSIFICATION, not an event. The 10-Q reports Licences — net of $128,148M at 2025-12-31, on a separate line from Other Intangible Assets — net of $5,254M. The vendor's 2024 figure included licences; its 2025 figure does not. Total non-current assets rose from $363,627M to $371,466M over the same period, confirming nothing left the balance sheet. Anyone comparing the two years without the filing would see a $127 billion write-off that did not happen.

4. Trailing GAAP earnings are inflated by roughly $5 billion of one-off benefit — the trailing P/E is REJECTED. The September 2025 quarter reports net income of $9,314M and EBITDA of $17,670M against a quarterly run-rate of $3.7-4.6B and $11.2-13.0B. priceToEarningsRatioTTM of 7.716x, netIncomePerShareTTM of $3.098, effectiveTaxRateTTM of 11.66%, earningsYieldTTM of 13.25% and netDebtToEBITDATTM of 2.704x are all contaminated. We use the adjusted series on which estimates and earnings-calendar actuals are struck: trailing adjusted EPS $2.28, trailing multiple 10.25x, adjusted trailing EBITDA approximately $48.7B.

5. epsdiluted is NULL on every row and eps is the GAAP basic figure. The vendor's 2025 eps of 3.04 is GAAP; the estimate block's 2025 epsAvg of $2.059 is adjusted. The two are not comparable and the file does not say so. All EPS figures here are labelled by basis.

6. Cost of revenue is reclassified between 2024 and 2025, breaking the gross-margin series. costOfRevenue falls from $69,801M (2024) to $50,820M (2025) while revenue rises 2.7%, so grossProfit jumps from $52,535M to $74,828M — a 42.4% increase. The same break appears quarterly: Q4 2024 gross profit $12,729M against Q1 2025 $18,593M. This is a classification change, not an operating event, and every gross-margin figure in the payload is unusable across the boundary. No gross-margin figure appears in this dive.

7. est.ebitdaAvg and est.ebitAvg show the fixed-ratio fabrication signature — NOT USED. From 2025 through 2030, ebitdaAvg is 53.1% of revenueAvg and ebitAvg 33.6%, effectively constant to three significant figures across six years. Independently, the 2025 ebitdaAvg of $66,439.9M is roughly 36% above our adjusted trailing figure of about $48.7B, a level AT&T has never reported. All forward valuation runs on epsAvg.

8. enterpriseValueTTM is CORRECT and CURRENT while bal_a is neither — the file contradicts itself. $305,527M less the market capitalisation of $160,209M implies net debt of $145,318M, which equals the 2026-06-30 funded net debt of $126,384M plus the $18,934M of non-current operating lease liabilities, to the dollar-million. The derived block uses the June 2026 balance sheet; the raw block uses a double-counted December 2025 one.

9. capitalLeaseObligations of $18,943M is the OPERATING lease liability, not a finance lease. AT&T's balance sheet shows operating lease right-of-use assets of $22,781M and operating lease liabilities of $18,934M at 2026-06-30. The field name is wrong and the value is then double-counted per Finding 1.

10. lastDividend is absent; dividendPerShareTTM of $1.11 is the trailing total AND happens to equal the run-rate. profile.lastDiv is null. AT&T has paid $0.2775 quarterly, so $1.11 is both the trailing figure and the annualised rate — a rare case where the standing labelling trap does not bite. dividendPayoutRatioTTM of 37.3% is, however, computed against the one-off-inflated GAAP figure; against adjusted trailing EPS of $2.28 the payout is 48.7%, and that is the number that matters.

11. Working-capital ratios are correct and describe a disclosed financing policy rather than operational strength. daysOfPayablesOutstandingTTM of 271.0 and cashConversionCycleTTM of −229.7 days reflect the supplier-financing arrangements the 10-Q describes: payment terms of "90-day or greater" and handset terms extended "by up to approximately 120 days, with an average of 85 days outstanding." The 10-Q quantifies the effect: direct supplier financing reduced operating cash flow by $272M in the first half of 2026 and $2,146M a year earlier. The ratios are right; the interpretation must not be that AT&T collects before it pays through operating efficiency.

Not defects, correctly reported: the share count reconciles (10-Q 6,878,770,356 outstanding at 2026-06-30 against 6,852.8M implied by market capitalisation, a 0.4% difference consistent with continued buybacks); revenuePerShareTTM of $18.339 reconciles to trailing revenue of $127,239M; freeCashFlowYieldTTM of 11.02% and priceToFreeCashFlowRatioTTM of 9.07x are arithmetically sound on the vendor's free-cash-flow definition; and the vendor's composite rating of A- / overall score 4 is the highest in this batch and, unusually, we do not reject it — the sub-scores of 5 on discounted cash flow and 4 on returns are consistent with the free-cash-flow yield, and the single 1 on debt-to-equity is the double-count of Finding 1 rather than an independent judgement.

Non-equity tripwire — checked and passed, with a note. T is $1 par value common stock, listed on the NYSE and the Chicago Stock Exchange, 7,620,748,598 issued and 6,878,770,356 outstanding. AT&T also has listed preferred stock (Series A, 48,000 outstanding; Series C, 70,000 outstanding; Series B redeemed during 2026) and roughly thirty separately listed global note series, all of which appear on the 8-K cover page — so a careless ticker match on this issuer can easily return a bond or a preferred. The common stock is confirmed by beta of 0.422, a variable dividend, a 52-week range of $20.48 to $29.62 (44.6%) and volume of 50.4 million shares (roughly $1.18B of turnover). This is common equity.

8. Technicals

Price $23.38. −21.1% from the 52-week high of $29.62; +14.2% above the 52-week low of $20.48. Position within the annual range: 27th percentile.
The moving averages are the problem. −7.1% BELOW the 200-day at $25.17, which is falling, while +2.7% above the 50-day at $22.78. The 50-day is well below the 200-day. This is a stock repairing a downtrend, not one in an uptrend, and it is the weakest chart in this batch alongside Blackstone's.
RSI 66.9 — the second-highest in this batch, which is striking for a stock 21% below its high and tells you the recent bounce is already extended in the short run. MACD +0.36 — marginally positive.
Maximum drawdown from peak over the trailing year: −21.1%, i.e. the current price is close to the worst of it.
Relative performance is uniformly negative and has not inflected: 3-month −10.4% vs SPY +7.6% and QQQ +7.7% — 18 points behind; 6-month −11.1% vs SPY +11.1% — 22 points behind; 12-month −15.7% vs SPY +24.3% and QQQ +30.8% — 40 points behind. Unlike Blackstone, where the last three months turned, AT&T is still underperforming on every horizon.
Sentiment: 1 strong buy, 25 buy, 30 hold, 6 sell across 62 analysts — consensus "Hold". Target $27.13 (+16.0%), median $28.00 (+19.8%), high $31 (+32.6%), low $20 (−14.5%). 62 analysts is the widest coverage in this batch, and the median target 20% above spot sitting alongside a Hold consensus is a genuine tension: the sell side thinks the shares are worth more and will not tell clients to buy them.

Today's move

T closed 2026-08-04 at $23.38, down $0.21 or 0.89% from $23.59. It opened at $23.35, traded $23.02 to $23.46, and closed near the upper end of that range on 50,412,988 shares — roughly $1.18 billion of turnover, by far the heaviest volume in this batch. The one piece of company news dated 2026-08-04 is in the archive: an 8-K filed that day disclosing that AT&T issued five tranches of global notes on 2026-08-03 — 3.600% due 2030, 4.150% due 2034, 4.550% due 2038, 5.050% due 2045 and 7.050% due 2052 — through Barclays, Citigroup, Goldman Sachs and Wells Fargo as representatives. The principal amounts are not in the extracted text and we do not estimate them. A 7.050% coupon on 2052 paper is expensive long money and is worth recording as the observable cost of AT&T's capital at the long end.

9. Insiders — eight identical accruals and nothing else

All eight transactions were filed on 2026-08-04, the dive date, for transactions dated 2026-07-31, and all eight are A-Award:

Person	Role	Security	Shares	Price	Holding after
John T. Stankey	Director, CEO & President	Common stock	927.915	$23.25	78,672.116
Pascal Desroches	Sr. EVP and CFO	Common stock	3,487.472	$23.25	134,269.572
Jeffery S. McElfresh	Chief Operating Officer	Common stock	2,690.053	$23.25	183,027.747
Lori M. Lee	Global Marketing Officer & SEVP Intl	Common stock	499.616	$23.25	9,926.969
Matthew K. Rose	Director	Deferred stock units	4,238.2967	$0	359,338.827
Cindy B. Taylor	Director	Deferred stock units	2,666.9601	$0	226,114.9709
Beth E. Mooney	Director	Deferred stock units	2,340.15	$0	198,406.7752
Luis A. Ubinas	Director	Deferred stock units	869.0834	$0	73,684.1809

The reading, and it is short. Every entry is an award, every one is indirect, and the four officer entries are priced at $23.25 on fractional share quantities — the signature of dividend equivalents accruing on deferred compensation, not of purchases. The four director entries are deferred stock units at zero cost.

What the file contains is: zero open-market purchases and zero open-market sales. That is a neutral signal and it is reported as neutral. The one observation worth making is the residual holdings, which are small for a company of this size: the chief executive holds 78,672 shares — $1.84 million at the closing price — against a $160 billion market capitalisation. The chief financial officer holds more than the chief executive (134,270 shares) and the chief operating officer holds more than both (183,028). We record that without inferring anything from it, beyond noting that insider ownership at AT&T is not economically meaningful in either direction.

10. Verdict, kill-criteria and flip conditions

Buy — Tactical.

What is genuinely good, and almost none of it is visible in the vendor payload: Advanced Connectivity — 90.7% of revenue — grew 4.1% while its operating income grew 20.3%, its operating margin went from 22.2% to 25.7% and its EBITDA margin from 40.5% to 42.0%, with advanced home internet up 27.3% and depreciation down 6.9% as a fully depreciated copper network rolls off; $19,442M of free cash flow, a 12.1% yield, covering a 4.75% dividend 2.5 times; 158.7 million shares — 2.3% of the company — retired in six months for $3,917M, a roughly 4.6% annual pace; funded net debt of $126,384M, not the $155,752M every screen shows; beta of 0.422, the lowest in this batch; the highest vendor composite rating in this batch (A- / 4), which we do not reject; and five consecutive adjusted-EPS beats, the latest by 10.2%.

What keeps this Tactical rather than Core: a ~$23 billion EchoStar spectrum purchase funded with "cash on hand and term loan borrowings" that takes funded leverage from about 2.60x to about 3.07x; a Legacy segment falling 25.9% with its operating margin down from 43.6% to 32.0%, so the runoff is getting less profitable; capital expenditure at 17.5% of revenue with no relief in sight; consolidated revenue growth of 2.3% that no amount of segment analysis changes; a Hold sell-side consensus with six sell ratings; a price 7.1% below its falling 200-day moving average; and a knowledge-base lane whose only name-level claim is bearish.

The distinction that matters. The market is pricing AT&T as a melting ice cube and the segment disclosure says it is not one — but the reason the market cannot see that is the same reason we can: the company reorganised its segments this year and the data vendors have not caught up. That is a genuine, temporary, mechanical information advantage, and it is the reason for a position. It is not a reason for a large one, because the thing the market IS pricing correctly — $23 billion of new spectrum obligations landing on a balance sheet with $144 billion of debt and negative tangible book — is also real and is about to happen. Buy the yield and the operating leverage; size it for the leverage step.

Pre-registered KILL criteria — what would take this to Watch or Avoid:

Advanced Connectivity operating margin below 24% in any quarter. It is 25.7% and the entire earnings-growth case rests on it.
Segment depreciation ceasing to decline. The −6.9% year-on-year fall is the mechanism producing 20% operating-income growth on 4% revenue growth; when it stops, so does the growth.
A ratings action or a funded-leverage ratio above 3.3x after the EchoStar close.
The buyback suspended to fund the spectrum purchase. That removes roughly 4.6 points of the 9.3% total shareholder yield and would be the clearest signal that $23 billion was too much.
Any impairment against the $129,123M of licences, which would directly refute the only genuinely useful claim in our knowledge base.
Advanced home internet growth below 15%, from 27.3% — the only fast-growing line in the company.

Pre-registered FLIP conditions — what would take this to Buy — Core:

The EchoStar transaction closing predominantly with cash, keeping funded leverage below 2.9x. This is the single most likely path and it may already have happened — management expected a close by the end of July 2026 and the archive does not confirm the outcome.
A third consecutive quarter of Advanced Connectivity margin expansion on 2026-10-28, confirming the operating leverage is structural rather than a depreciation artefact.
Consolidated revenue growth above 3%, which would mean Advanced Connectivity is finally outrunning the Legacy runoff at the consolidated line and would justify a multiple above 11x.
A price below approximately $21 (roughly 8.2x the 2027 consensus, near the 52-week low), which would put the base case at +28% with the dividend yield above 5.2%.
A recent, named-speaker, name-level knowledge-base claim. The lane's only name-level entry is a bearish claim from 2023 whose central prediction has not occurred.

Where T fits in the Synthos Framework Portfolio. The defensive-income sleeve, at 0% today with a 2-3% target initiated here. On batch overlap: AT&T is the lowest-beta name in this batch after TJX and the highest-yielding by a wide margin (9.3% total shareholder yield against 3.0% for BlackRock, 3.8% for Blackstone and 1.0% for Eaton). It is genuinely uncorrelated with the growth names in this batch and is the only position here whose return does not require an equity-market or capital-expenditure cycle to cooperate. That is a portfolio-construction argument as much as a security argument, and it is stated as such. Logged as a tracked Synthos call (Buy — Tactical) as of 2026-08-04 at $23.38.

Single biggest risk: the EchoStar leverage step, arriving as the Legacy business shrinks. Approximately $23 billion of FCC licences, agreed 2025-08-25, expected to close by the end of July 2026, funded with "a combination of cash on hand and term loan borrowings" — on a company that already carries $143,954M of funded debt, spends 17.5% of revenue on capital expenditure, and has negative tangible book value of $10.47 a share. On our arithmetic that is roughly 3.07x adjusted EBITDA against 2.60x today, and the price does not appear to contain a credit consequence. The mitigating fact is that AT&T issued five tranches of notes on 2026-08-03 without difficulty; the sobering one is that the longest tranche carries a 7.050% coupon.

Provenance & disclosures

Traceability: 7 tagged knowledge-base claims name AT&T across 51,928 distilled claims (raw hits 7, entity matches 7, used 7, discarded 0; 7 distinct channels — one claim each, the flattest distribution in this batch). No concentration sensitivity test was run because there is no concentration to test; that is stated as a finding rather than an omission. A case-sensitive bare-ticker sweep on the single letter "T" was NOT run and would be uninformative; this is a stated limitation of the method on this name, and entity extraction is doing all the work. Only ONE claim is a view on AT&T as a security and it is bearish: mike_green, NAMED SPEAKER Michael Green, 2023-08-11, conviction 58, skill 0.9 — highly levered names including AT&T with pension obligations hollowing out business and potentially hitting a refinancing wall; three years old, its central prediction has not occurred, and AT&T priced five note tranches on 2026-08-03 without difficulty. The most useful claim is thematic: raoul_pal, 2025-03-13, conviction 69, no named speaker — wireless spectrum bought at auction and held has delivered double-digit annualised returns, which bears directly on $129,123M of licences and approximately $24 billion of contracted additions. Two claims describe satellite substitution (real_vision 2023-03-13 conviction 78 — "indebted telcos can't build"; all_in 2025-09-10 conviction 55) and BOTH are tagged bullish by the distiller because the stance records the speaker's view of Starlink rather than of AT&T — a stance-tag inversion flagged explicitly and NOT counted as support. Three are generic macro or corporate-finance principles tagged to the entity without asserting anything about it (lyn_alden 2021-01-04, arthur_hayes 2024-03-29, we_study_billionaires 2026-03-07). No claim carries a management speaker_role; six of seven carry no named speaker. All quotes are verbatim from the stored claim text.
Data as-of: fundamentals — second-quarter 2026 income statement and balance sheet as at 2026-06-30, from the 10-Q filed 2026-07-22; annual figures from the 10-K filed 2026-02-09 (fiscal 2025) · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873819 = 2026-08-04T20:03:39Z ($23.38, −0.89%; 50-DMA $22.78; 200-DMA $25.17; RSI 66.9; MACD +0.36) · knowledge-base claims 2026-08-04. Fiscal year = calendar year. All figures come from the Synthos vendor data file for T or from the SEC filings in the T archive; no figure comes from memory, recall or external retrieval.
Filing archive contents: 10-K filed 2026-02-09 (fiscal 2025, 472,382 characters, financial-statement tables preserved); 10-Q filed 2026-04-27 (March 2026 quarter); 10-Q filed 2026-07-22 (June 2026 quarter, 163,848 characters — the primary source for this dive, including the reorganised segment tables, the EchoStar disclosure and the liquidity discussion); 8-K filed 2026-07-22 (Item 2.02, second-quarter results — cover only, Exhibits 99.1 and 99.2 not extracted); 8-K filed 2026-07-28 (Item 7.01, a furnished press release — cover only, Exhibit 99.1 not extracted, contents unknown); 8-K filed 2026-08-04 — the dive date — disclosing a five-tranche global note issue settled 2026-08-03 (3.600% due 2030, 4.150% due 2034, 4.550% due 2038, 5.050% due 2045 and 7.050% due 2052) under an underwriting agreement dated 2026-07-27; the principal amounts are not in the extracted text and are not estimated here.
Where the filings contradicted or corrected the vendor (detailed in Section 7): totalDebt shown to DOUBLE-COUNT $18,943M of operating lease liabilities — the vendor's longTermDebt of $146,032M equals the filed $127,089M plus the lease liability, which is then added again via capitalLeaseObligations, overstating funded debt by $37,886M (27.8%) and funded net debt by 32.1%; funded net debt rebuilt from $155,752M to $126,384M at 2026-06-30 using the 10-Q's own total-debt table ($143,954M) and cash ($17,570M); seg_prod and seg_geo recorded as BOTH EMPTY across a 2026 segment reorganisation into Advanced Connectivity, Legacy and Latin America, with every segment figure in this dive taken from the 10-Q; intangibleAssets falling $127 billion between 2024 and 2025 identified as a reclassification of the Licences line ($128,148M at 2025-12-31, shown separately in the 10-Q) rather than an economic event; the trailing GAAP P/E of 7.72x rejected because the September 2025 quarter contains roughly $5 billion of non-recurring benefit, with the honest adjusted trailing multiple of 10.25x used instead, and effectiveTaxRateTTM, earningsYieldTTM and netDebtToEBITDATTM rejected for the same reason; epsdiluted NULL on every row and eps identified as GAAP basic while est.epsAvg is adjusted; the gross-margin series rejected across a 2024-to-2025 cost-of-revenue reclassification that moved $19 billion between lines; est.ebitdaAvg/ebitAvg rejected on fixed revenue ratios of 53.1% and 33.6% across six years and a 2025 EBITDA estimate 36% above any figure AT&T has reported; capitalLeaseObligations identified as the OPERATING lease liability; and the working-capital ratios explained by reference to the 10-Q's disclosed supplier-financing arrangements rather than treated as operational strength. Where vendor and filing AGREED — worth recording: enterpriseValueTTM of $305,527M reconciles EXACTLY to the 2026-06-30 funded net debt plus non-current operating lease liabilities and is current rather than stale, even though the balance sheet feeding it is not; the share count (10-Q 6,878,770,356 at 2026-06-30 against 6,852.8M implied, a 0.4% difference consistent with continued repurchases); revenuePerShareTTM; the free-cash-flow series; and dividendPerShareTTM of $1.11, which is BOTH the trailing total and the annualised run-rate — a rare case where the standing labelling trap does not bite.
Basis note: AT&T reports GAAP results and an adjusted EPS measure excluding, among other items, merger-related amortisation and significant gains and losses. Analyst estimates and the earnings-calendar actuals are both struck on ADJUSTED EPS, verified against the reported figures ($0.54, $0.54, $0.52, $0.57, $0.65). The GAAP-to-adjusted gap is unusually large in the trailing period — $3.10 GAAP against $2.28 adjusted — because of a roughly $5 billion non-recurring benefit in the September 2025 quarter, and that gap is the reason the vendor's headline P/E of 7.72x should not be used. All forward multiples use adjusted EPS; the trailing multiple is shown on both bases.
Estimate coverage: 17 analysts on 2026 revenue and 14 on EPS; 17 and 14 on 2027; 15 and 7 on 2028; 11 and TWO on 2029; 7 and 5 on 2030. This is the best-covered near-term estimate set in this batch. The fair value is struck on the 2027 distribution (14 analysts on EPS). The 2029 EPS row rests on two analysts and is excluded from every conclusion.
Peer note: the vendor peer set is genuinely apposite — Verizon ($195.8B), T-Mobile US ($190.1B), Comcast ($88.5B), América Móvil ($74.7B), Charter ($20.6B), Rogers ($18.3B) — alongside Disney ($170.5B), Telefônica ($19.7B), Telkom Indonesia ($15.1B) and Telefônica Brasil ($8.7B), which are less relevant. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn.
Fair-value caveat: the $19.00 / $26.90 / $30.50 anchors are multiples of the 2027 consensus adjusted EPS distribution — 7.4x the mean of $2.558 for the bear (cross-checked at 7.9x the low of $2.419), 10.5x the mean for the base, and 11.9x the mean for the bull (cross-checked at 10.5x the 2028 mean of $2.910). Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 10x gives $25.58 and 11x gives $28.14. The base assumes MILD multiple EXPANSION from 10.0x to 10.5x, which is the one leg of this thesis that requires the market to change its mind, and the justification — 20.3% operating-income growth in 91% of the company with 350 basis points of margin expansion — is stated rather than assumed. At a flat multiple the twelve-month return is +9.4% of earnings growth plus a 4.75% dividend, which is the fallback case.
Timing: second-quarter 2026 results were released 2026-07-22 and the 10-Q filed the same day, thirteen days before this dive. The next print is 2026-10-28, 85 days away — consensus $0.62 of adjusted EPS on $31,687.3M of revenue. Two unscheduled events sit inside that window and one may already have occurred: management expected the ~$23 billion EchoStar spectrum transaction to close BY THE END OF JULY 2026 and the archive does not confirm the outcome; and a five-tranche note issue settled 2026-08-03 and was disclosed in an 8-K filed on the dive date itself. 2026-08-04 was a −0.89% session on 50.4 million shares, the heaviest volume in this batch.
Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
Version: 2026-08-04-full.