Dislocation

There is a real dislocation: AT&T fell −30.9% from a September 2025 peak of $29.62 to a July 1, 2026 trough of $20.48, and trades at $22.96. But its anatomy is a multiple de-rating, not an earnings break — over the same window forward EPS estimates rose and the company beat consensus EPS every quarter. Selling volume peaked at only 2.37× the pre-fall median. The operative trigger was narrative and mechanical — a satellite-competition scare, an analyst downgrade, and index-exclusion flow — not a guidance cut.

The drawdown — one number, two legs, a full round-trip in between

The fall spans 289 days and −30.9%, but it is not one move. The stock dropped, recovered almost the entire loss, then fell further and deeper.

Source: derived from the daily price feed; peak $29.62 (2025-09-15), trough $20.48 (2026-07-01), current $22.96 (2026-07-23) per fit_features.capitulation_gauge.drawdown. Company filings, as reported.

The shape is a W, confirmed by the moving-average signals: a death cross on 2025-11-07, a golden cross on 2026-03-25, then a second death cross on 2026-05-27.

Source: derived from the daily price feed; interim rebound high $29.10 on 2026-03-27. Company filings, as reported.

The fit_features peak-to-trough of −30.9% is arithmetically correct, but it straddles a recovery that carried the stock back to within $0.52 of the September peak by late March. The operative dislocation is the second leg — −29.6% from $29.10 (Mar 27) to $20.48 (Jul 1) — and that is where the anatomy below concentrates.

The trigger — two different stories

Leg 1 was a demanding valuation meeting a merely-fine quarter, not a fundamentals break. AT&T reported Q3 FY2025 on October 22, 2025: 405K postpaid phone net adds to 73.8M subscribers, +2.3% mobility service revenue [1], $4.9B of free cash flow [2], and full-year guidance affirmed — capital investment of $22B to $22.5B and adjusted EPS at the higher end of $1.97 to $2.07 [3]. Consensus was met, not beaten: EPS surprise +0.4%, revenue −0.6%. For a stock sitting near a 52-week high, an in-line print was not enough — shares fell −3.64% on October 23 on 107M shares, roughly triple normal turnover. The larger single-day drop of the leg, −4.40% on October 6, preceded earnings and had no company-specific catalyst in the corpus; it reads as a macro/rate move, not an event.

Leg 2 had no earnings deterioration at all. Between the March high and the July trough, AT&T reported Q1 FY2026 (Apr 22): adjusted EPS growth of +11.8% year over year, ahead of consensus [4]. Nothing in the filed record explains a −30% leg. The identifiable, dated drivers are external to the numbers and come from the market, not the company:

• A competitive-narrative scare. Through June 2026, sell-side and press coverage repriced AT&T's long-term broadband and wireless growth for the low-earth-orbit satellite threat (Starlink), questioning what the growth story is worth if satellite operators sell direct-to-device plans.
• An analyst downgrade. Oppenheimer downgraded the stock in June 2026, flagging both the satellite threat and execution risk on AT&T's fiber build — the view that penetration could disappoint and the company might stop nearer 50M locations than its 60M-by-2030 ambition, a large capex bet whose payoff the market began to doubt.
• Mechanical index selling. AT&T's removal from a Russell top-tier index at the late-June reconstitution forced passive/institutional selling into the exact window of the trough.
• A management-continuity overhang. CFO Pascal Desroches's planned departure (successor Jennifer Biry named July 6) landed days after the July 1 bottom, adding transition risk rather than causing the low.

(The four bullets above rest on press and sell-side reports — tikr.com, stockstotrade.com, ts2.tech — not the filed corpus, and are attributed as such.)

Then the fear reversed on a number. AT&T reported Q2 FY2026 on July 22, 2026 with adjusted EPS up +20.4% year over year, +2.7% service revenue and $4.7B free cash flow [5]; the stock recovered +12% off the trough into the current $22.96.

The fear gauge — orderly repricing more than panic

The measured volume spike is 2.37× — the peak 20-day average daily volume during the fall against the median daily volume over the 180 days before the peak (roughly 33M shares).

Source: fit_features.capitulation_gauge; historic spike multiples from the daily volume feed. Company filings, as reported.

Where the volume concentrated matters: it clustered at the end of the fall, not the trigger. Monthly average turnover was ~30M shares before the peak, rose to ~63M in October (leg 1), and spiked to ~60M in June and ~85M in July as the second leg capitulated. The single heaviest day was June 30, 2026: −5.13% on 126.6M shares — the day before the July 1 trough. That late clustering is consistent with forced and emotion-driven selling into the low (the index-reconstitution flow fits here). But a 2.37× peak multiple is a modest reading: this looks more like a sustained, orderly de-rating than a violent, single-day capitulation.

Who was selling — diffuse, with one mechanical hand

The seller evidence is thin and, where it exists, points away from informed distribution.

• Short interest: not available. FINRA returned no reported short-interest rows for the ticker in this run, and no days-to-cover or borrow-pressure data was staged. The short-side of the story cannot be quantified here — a genuine data gap, not a finding of low short interest.
• Insiders neither sold nor bought. Across the drawdown window (2025-09-15 to 2026-07-24) the only insider Form 4 activity was 14 routine grants/awards (code A) — zero open-market sales and zero open-market purchases. No insider distributed into the fall; equally, no insider stepped in to buy the −31% dip.
• The 5% holders are passive. The disclosed >5% owners are index managers (BlackRock, Vanguard). Their positions track index weight, which is the point: the one identifiable forced seller is mechanical — the late-June index-reconstitution flow described above — rather than an investor acting on a changed view of the business.

Net: a broad, sentiment-driven de-rating by an income/institutional holder base, with index-exclusion flow the only nameable structural seller. No evidence of informed insiders selling ahead of the fall.

Estimates vs price — the framework's signature, in its cleanest form

The framework looks for a price fall that outruns the estimate cut. Here there was no estimate cut at all: forward estimates rose while the price fell.

Source: CapIQ consensus revision series, data/sp/estimates.json (momentum snapshots); price from the daily feed. Company filings, as reported.

Over the 180 days the momentum series covers, consensus FY2027 EPS rose from $2.45 to $2.57 (+4.6%) and FY2028 EPS from $2.58 to $2.91 (+12.8%); FY2027 revenue estimates climbed ~2%. Across the same span the stock round-tripped from $23.59 up to $29.10 and down to $20.48, ending lower. Because the earliest estimate snapshot (January 2026) already sits below today's, the full −30.9% peak-to-trough came with zero downward revision to forward earnings.

The realized numbers say the same thing. AT&T beat consensus EPS in every quarter of the drawdown, and the beats widened:

Source: CapIQ actual-vs-consensus, data/sp/estimates.json (beat_miss). Company filings, as reported.

Reported adjusted EPS growth accelerated through the fall — +11.8% in Q1 FY2026 to +20.4% in Q2 FY2026 [6] [7]. Revenue surprises stayed inside ±2% throughout. On every measurable axis, the earnings line went one way and the price went the other.

Bottom line

A real dislocation exists — a −30.9% drawdown to a July 1 trough, now $22.96 — and it carries the framework's signature almost to caricature: the stock fell ~31% while forward estimates rose 5–13% and the company beat EPS every quarter. The repricing was driven by a competitive narrative (satellite), an analyst downgrade, and mechanical index-exclusion selling, not by any deterioration in the filed numbers. Two cautions keep it honest: the drawdown is a slow, two-legged de-rating spanning a full recovery, not a single-event crash, and its peak selling multiple of 2.37× is well short of AT&T's own historic panic days — orderly repricing more than capitulation. What happened to the price is clear; what it means for value is not this tab's to judge — see Damage Math and Fit.