AMD 18 min read

Advanced Micro Devices (AMD) — Investment Tree v1

Stage 7 final essay. Bilingual companion: tree_v1_zh.md. Date: 2026-05-29 · Anchor price: $516.10 (NASDAQ close 2026-05-29) · Evidence mode: R2-VERIFIED — FY2025 10-K + Q1 2026 10-Q (SEC EDGAR XBRL, Tier A); price/multiples Tier B (3-source) Archetype: Confirmed three-platform AI compounder priced for perfection — the disruptor-from-below thesis WON, and the stock has already re-rated to reflect it

SOURCE QUALITY: R2-verified (2026-05-29). Replaces the 2026-05-10 GENERATE-mode scaffold, which anchored at $105† — a training-knowledge guess equal to AMD's 52-week low. The real anchor is $516.10 (52-wk range $108.62–$527.20). Segment revenues, margins, FCF, balance sheet, and Q1 2026 Data Center are now Tier-A from the FY2025 10-K (accn 0000002488-26-000018, filed 2026-02-04) and Q1 2026 10-Q (accn 0000002488-26-000076, filed 2026-05-06). Price, forward multiples, and consensus estimates are Tier-B (3-source). Only forward scenario targets remain analytical (†). See update_2026-05-29.md for the full R2 delta record.

0. Company Fundamentals — what AMD is and how it earns

Figures FY2025 (ended ~Dec 27, 2025) unless noted.

What it is & how it earns. AMD is a fabless semiconductor designer — it engineers CPUs (EPYC servers, Ryzen client), GPUs (Instinct MI-series for AI, Radeon for gaming), and FPGAs/adaptive-SoCs (Xilinx Embedded), then outsources fabrication to TSMC. It sells to hyperscalers, OEMs (Dell/HPE/Lenovo), and enterprise/industrial customers who buy on performance-per-watt and chiplet flexibility against Intel (CPU) and NVIDIA (AI GPU). FY2025 revenue was $34.6B across three reported segments: Data Center $16.6B (48%), Client and Gaming $14.6B (42%) — Ryzen $10.6B plus Radeon/semi-custom $3.9B — and Embedded $3.5B (10%), the highest-margin segment (36% op margin). Data Center carries customer concentration toward a handful of hyperscalers.

Cash-flow anatomy. Cash generation inflected sharply with the AI Data Center ramp:

GAAP earnings quality lags non-GAAP because the $49B Xilinx deal layers large non-cash acquisition-related amortization onto the P&L — FY2025 GAAP diluted EPS was $2.65 vs non-GAAP $4.17. (FY2025 OCF includes ~$1,216M from discontinued operations; continuing-ops OCF ~$6,493M.)

Balance sheet & capital allocation. Fortress sheet: cash + short-term investments $10.6B vs total debt $3.2B → net cash ~$7.3B, investment-grade. AMD pays no dividend and runs a modest buyback (historically ~$1B/yr against a $30B authorization — roughly stock-comp-offsetting, not wealth-creating). R&D is heavy at ~$8.1B, ~23% of revenue (FY2024 $6.5B). ROIC is only ~7–8%, around or just above WACC — adequate, not yet a wide compounding spread, weighed down by the Xilinx goodwill base.

What drives it. The swing factor is the Data Center / Instinct MI-series AI GPU ramp layered on durable EPYC CPU share — the question the tree resolves is not whether the business inflected (it did) but whether the rich forward multiple already prices it in.

I. One-sentence verdict

AMD's three-platform thesis was right and is now confirmed — Data Center inflected to $5.8B/quarter (Q1 2026, +57% YoY), EPYC's structural CPU share is locked, the balance sheet is net-cash $7.3B, and Embedded is the only laggard — but at $516 the stock trades at ~70× forward earnings, ~21% ABOVE Wall Street's $405–410 fair value, so the favorable mispricing the original thesis was built on no longer exists; this is a WAIT — a great business at a full-to-excessive price, with no position justified at spot, a re-engagement zone near ~$410 (Street fair value / the SOTP-plus-growth line), and an explicit add only on a multiple reset toward 45–50× or evidence that consensus FY2026 EPS materially understates the AI-inference ramp.

II. Company snapshot

Advanced Micro Devices (AMD) is a fabless semiconductor company headquartered in Santa Clara, California, designing CPUs, GPUs, FPGAs, and adaptive-SoCs for data-center, client-computing, gaming, and embedded markets. Founded 1969. Under CEO Dr. Lisa Su (since October 2014), AMD executed the most dramatic competitive turnaround in semiconductor history — and as of FY2025 the market has fully recognized it: the stock is up ~5× from where the scaffold mis-anchored it.

Three reportable segments (FY2025 10-K — note: Client and Gaming were MERGED into one segment vs the prior 4-segment structure):

FY2025 GAAP diluted EPS $2.65 (non-GAAP $4.17). Cash + STI $10,552M; total debt $3,222M; net cash $7,330M. FY2025 FCF $6,735M (OCF $7,709M − CapEx $974M). No dividend.

Market data (2026-05-29): price $516.10; market cap $841.6B; ~1.63B shares; forward P/E ~70× on FY2026 consensus non-GAAP EPS $7.41; trailing ~124× non-GAAP / ~195× GAAP. FY2026 consensus revenue $50.0B (+44% YoY). Wall Street 12-month target ~$405–410 (Buy, 34 analysts) — ~21% below spot.

III. The five facts that drive everything

1. Data Center already inflected — the scaffold's "central unknown" is resolved. Q1 2026 Data Center revenue was $5,775M (+57% YoY vs $3,674M), already +44% above the $4.0B threshold the scaffold treated as an unresolved July-2026 falsification gate. FY2025 DC was $16,635M. The GPU thesis is no longer "pre-inflection" — it printed. ✅A

1. At $516 the re-rating is complete. Forward P/E ~70× on FY2026 consensus EPS $7.41; the stock sits above Wall Street's $405–410 fair value. The "single ~22× blended multiple discounting two businesses to zero" condition the original thesis rested on no longer exists — the market now pays a full AI-growth premium. ✗ (valuation) A/B

1. EPYC structural CPU durability is confirmed and is the quality floor. Data Center segment op margin 21.7% and the EPYC franchise underpin it; Intel's execution gap persists. This is the most durable leg — but it is already in the price. ✅A

1. The balance sheet is a fortress, not a question mark. FY2025 net cash $7,330M, FCF $6,735M — the scaffold had zero Tier-A support here and guessed "FCF ≥$3B would be achievable." Reality is 2× that. Capital-return / reinvestment optionality is real. ✅A

1. Embedded is the laggard, not the hidden compounder. Q1 2026 Embedded revenue $873M (~$3.5B annualized) is flat-to-FY2025 and below the scaffold's "$1.3B/quarter by Q3 2026" bull marker — it would need +49%. High-margin (36%), but not the 2H-2026 earnings inflection the bull case needed. ✗A

IV. The H-0 thesis

H-0 (one sentence) — REVISED post-R2: AMD's mis-categorization as a "GPU runner-up with permanent CUDA deficit" has already been corrected by the market: the three-platform compounder thesis (AI accelerator GPU inference beachhead + EPYC server CPU + Xilinx Embedded) is confirmed on the business — Data Center inflected, balance sheet is net-cash, EPYC is durable — but at $516 / ~70× forward / ~21% above Street fair value, the favorable mispricing is gone and the residual edge is negative, making AMD a high-quality business with an unfavorable risk/reward at spot (EV −13%, asymmetry 0.41× unfavorable).

The thesis split — business confirmed, valuation inverted:

• H-0A — "AMD wins inference-layer GPU share at scale" → CONFIRMED (was ~55%; now ~85%). Q1 2026 DC $5.78B proves ROCm is production-sufficient and the inference beachhead is real. This leg won.

• H-0B — "Intel's structural failure locks in EPYC share through 2028" → CONFIRMED (~80%, unchanged). DC segment economics + share durability hold. This leg won.

• H-0C — "Xilinx Embedded is a 2H-2026 earnings compounder" → WEAKENED (was ~55%; now ~35%). Embedded $873M Q1 is a slow recovery, not an inflection. This leg lagged.

• H-0D (NEW — the binding leg) — "The market still under-prices AMD's three platforms" → FALSIFIED. At ~70× forward with Street PT below spot, the market fully credits (arguably over-credits) the thesis. The mispricing is the inverse of what the scaffold claimed.

Why the verdict is WAIT despite three confirmed business legs: an investment thesis needs BOTH a right business AND a favorable price. AMD now has the first and not the second. The scaffold conflated "the business will inflect" (true) with "the stock is cheap" (false — it had already re-rated 5× before the scaffold was even written, but the scaffold never checked the price).

Mispricing anatomy — retired: the three-layer mispricing the scaffold described (time-lag on ROCm, cognitive bias on EPYC, structural-neglect on TAM) was real in ~2023–2024 and has since been arbitraged away by a 5× re-rating. There is no current exploitable mispricing to the upside; if anything the structural-neglect has flipped to structural-enthusiasm (the AI-capex premium).

V. Tree — five branches

H-0 (revised): AMD's three-platform thesis is CONFIRMED on the business
     but the stock at $516 (~70x fwd, >Street fair value) has fully
     re-rated — favorable mispricing is gone; residual edge is negative
│
├── L1A — GPU Accelerator Platform  ✅A CONFIRMED (Data Center inflected)
│   ├── 1.1 MI300X/MI350 HBM bandwidth lead durable     ✅A real moat; DC revenue proves it
│   ├── 1.2 ROCm 7.0 inference-workload parity w/ CUDA  ⚠️A 5-15% gap remains, but "good enough" at scale
│   └── 1.3 Data Center revenue ≥$4.0B/qtr             ✅A RESOLVED — Q1 2026 $5.78B (+44% past gate)
│
├── L1B — EPYC Server CPU Platform  ✅A STRONGLY CONFIRMED (Intel structural decay)
│   ├── 1.1 EPYC ≥30% server CPU share + margin stable  ✅A DC op margin 21.7%; franchise durable
│   ├── 1.2 Intel 18A fails competitive Xeon pre-Q4'27  ✅A execution gap persists
│   └── 1.3 Dell/HPE/Lenovo EPYC Turin design wins locked ✅A platform lock-in structural
│
├── L1C — Xilinx Embedded Platform  ✗A LAGGARD (recovery slow, not inflecting)
│   ├── 1.1 Embedded YoY growth by Q3 2026              ⚠️A $873M Q1 2026; ~flat, slow recovery
│   ├── 1.2 Embedded ≥$1.3B in Q3 2026                 ✗A MISS trajectory — needs +49% from Q1 run-rate
│   └── 1.3 No Xilinx goodwill impairment in FY2025     ✅A none in FY2025 10-K
│
├── L1D — Capital Allocation & Balance Sheet  ✅A CONFIRMED (fortress balance sheet)
│   ├── 1.1 FCF ≥$3B FY2026                            ✅A FY2025 FCF $6.7B — cleared 2x
│   ├── 1.2 Share count flat/declining                 ✅A ~1.64B diluted; roughly flat
│   └── 1.3 No liquidity crisis through FY2027          ✅A net cash $7.3B
│
└── L1E — Valuation & Scenario Architecture  ✗A UNFAVORABLE (re-rating complete)
    ├── 1.1 SOTP floor provides meaningful downside cushion ✗A floor ~$60 is 8x below spot — irrelevant
    ├── 1.2 Favorable upside from anchor                ✗A bull +36%, base −21%, bear −42% from $516
    └── 1.3 Asymmetry ratio ≥1.5x favorable             ✗A now 0.41x UNFAVORABLE (EV −13%)

Total: 9 ✅ / 2 ⚠️ / 4 ✗ / 0 ⊗ across 15 leaves
H-0 verdict: BUSINESS CONFIRMED, VALUATION UNFAVORABLE — net WAIT

The verdict shape tells the whole story: more ✅ than the scaffold (9 vs 7) because the business legs confirmed, but 4 new ✗ because the entire valuation branch inverted plus Embedded missed. Great company, expensive stock.

VI. Key findings

Finding 1 — The thesis was right; the stock already paid for it

Every business leg the scaffold bet on resolved in its favor: Data Center inflected hard ($5.78B/qtr), EPYC durability held, the balance sheet proved to be net-cash $7.3B with $6.7B FCF. This is a thesis vindication. The problem is purely that the scaffold anchored at $105 (a training-knowledge error = the 52-week low) and never checked that the market had already re-rated AMD 5× to $516 — fully pricing the very inflection the scaffold was "waiting for." The lesson is the entire reason for the R2 discipline: a screening memo that skips the primary-source price check can recommend accumulating a stock that has already won and is now expensive.

Finding 2 — At ~70× forward, the market prices more than even the bulls model

FY2026 consensus is non-GAAP EPS $7.41 on $50.0B revenue (+44% YoY). At $516 that is ~70× forward. Wall Street's own 12-month target is $405–410 — ~21% below spot. So the market is not merely crediting the three-platform thesis; it is paying a premium above the sell-side's already-optimistic fair value. To justify $516 as fair value (not a 12-month target) you need FY2027 EPS approaching ~$10–11 and a sustained ~50× multiple — i.e., the AI-inference ramp must materially exceed consensus AND the multiple must not normalize. That is the bull case, and it is a minority outcome.

Finding 3 — Embedded is the one leg that disappointed

The scaffold's "hidden $49B Xilinx compounder" is the laggard. Q1 2026 Embedded revenue was $873M — ~flat to the FY2025 quarterly run-rate and well short of the "$1.3B/quarter by Q3 2026" bull marker. It is a high-margin (36% OM) steady business, not a 2H-2026 earnings inflection. The Xilinx thesis is "fine, slow," not "hidden rocket." This removes one of the three upside legs the bull SOTP relied on.

Finding 4 — Durability is genuinely high — which is exactly the trap

AMD scores 22/25 on the long-term durability test with 0 fatal flags, now Tier-A-confirmed on the balance-sheet and cash-flow questions (net cash $7.3B, FCF $6.7B). The business is durable, the moat (CPU) is widening, and it will matter in 2036. None of that makes it a buy at 70× forward. High durability + high price is the classic "wonderful company, terrible entry" setup — durability protects the business, not the multiple. The 22/25 supports a high eventual position cap (5–7%) at a reasonable price; it does not justify entry at a ~21%-premium-to-fair-value spot.

VII. Valuation analysis & three scenarios

Current multiple: ~70× forward P/E (FY2026 consensus non-GAAP EPS $7.41); ~124× trailing non-GAAP / ~195× GAAP. This is a premium to NVIDIA's forward multiple, on a business with lower AI-GPU share. The "55–60% discount to NVIDIA" the scaffold described is gone — AMD now trades richer than NVIDIA on forward earnings.

Scenario architecture (anchor $516.10, 12-month):

Probability-weighted 12-month expected return: −12.9%† Asymmetry: 0.41× — UNFAVORABLE (weighted upside 8.9% / weighted downside 21.8%)

Bull ($700, 25%): the AI-inference TAM expands faster than consensus, AMD's MI400 series takes share, DC compounds toward $30B+/year, FY2027 non-GAAP EPS reaches ~$10–11, and the market keeps paying ~65–70×. This is the "consensus is too low" case — possible (AMD has beaten before), but it requires the multiple not to normalize even as growth matures, which is historically rare.

Base ($410, 45%): the most likely path. AMD executes, FY2026 EPS lands ~in line ($7.40), but the forward multiple compresses from ~70× toward the 45–55× that mature-hypergrowth semis command, landing the stock at Wall Street's current fair-value zone (~$405–410). A −20% drawdown from spot that is simply the multiple normalizing, not the business failing.

Bear ($300, 30%): the AI-capex cycle hits an air-pocket (see overlay below). 2-of-4 hyperscaler-capex triggers fire, DC growth decelerates, and AMD — first-order exposed (ai-capex-high) and lacking NVIDIA's CUDA-lock cushion — de-rates hard to ~33× on ~$9 EPS. −42% from spot. The net-cash balance sheet and EPYC floor prevent a worse outcome.

The asymmetry is the point: at $516 you are risking ~42% to make ~36%, with the central tendency a 21% loss. That is the inverse of the scaffold's claimed +2.5× favorable asymmetry — and it flipped purely because the anchor was wrong by 5×.

VIII. AI-capex air-pocket overlay (correlated drawdown)

Reference: dashboards/ai_capex_cycle_overlay.md · cycle_exposure: ai-capex-high

AMD is a first-order seller into hyperscaler capex — the ai-capex-high band. The correlated-drawdown scenario fires when 2 of 4 overlay triggers hit in a single quarter (single-hyperscaler capex cut >15%; aggregate guide-down >10% YoY; NVDA QoQ Data Center decline; hyperscaler RPO flat/declining 2 quarters). At $516 vs the scaffold's phantom $105, AMD's air-pocket exposure is far more dangerous than the scaffold modeled: a multiple compression from ~70× operates on a much larger base. AMD lacks NVIDIA's installed-base CUDA cushion, so in a total-pie compression AMD's multiple de-rates harder. This overlay is the spine of the Bear scenario. K.4 correlated-factor rule: if held alongside other ai-capex-high/mid-s-curve names (NVDA, TSM, AJNMY, TOTDY) and the combined bucket exceeds 15% of portfolio, the position cap drops 2pp — and at spot the position is 0% anyway.

IX. Why to wait (and what would change the call)

Why wait: the business won, but the price already reflects it and then some. Buying at ~70× forward / ~21% above Street fair value, into a downside-skewed (0.41×) distribution, with the AI-capex cycle mid-to-late, is paying a full premium for a confirmed story with negative expected 12-month return. There is no asymmetry to harvest at spot.

What would change the call to a buy:

1. Price reset to ~$410 or below (Street fair value / base case) — re-engage with a starter; the business quality (22/25 durability) supports building toward 3–5% at a reasonable multiple.
2. Evidence consensus FY2026 EPS is materially too low — e.g., two consecutive quarters of DC revenue beating by >15% with raised guidance, implying the bull EPS path. This would re-rate fair value up toward spot.
3. Multiple normalization without a business stumble — if AMD grows into the multiple (EPS rises while price stalls) toward ~50× forward, the risk/reward turns favorable on fundamentals.

What would deepen the avoid: any AI-capex air-pocket trigger firing (2-of-4) while the position is being considered → stand fully aside; the bear path is live.

X. Long-term holdability verdict (5–10 year horizon)

Hold for 5–10 years? The BUSINESS, yes. At THIS PRICE, no new capital.

AMD's durability (22/25, 0 fatal flags, now Tier-A-confirmed on the balance sheet) confirms a business relevant through 2036 with a widening CPU moat and a real GPU inference franchise. An existing holder with a low cost basis has a genuinely high-quality compounder and should not sell purely on valuation (the business keeps compounding). But new capital at $516 is buying a great business at ~70× forward with negative 12-month expected return — the entry, not the business, is the problem.

A 5-year hold initiated at spot is predicated on the Bull case (25%) being right — that AI-inference growth so exceeds consensus that today's ~70× is retroactively cheap. That is a minority outcome to underwrite with new capital. The disciplined path is to own the watchlist slot, not the position, until price meets quality.

Correlated exposure (K.3.4): load-bearing factor is AI-infrastructure semiconductor capex. AMD + NVDA + AJNMY + TOTDY combined should not exceed 20% of portfolio; AMD is ai-capex-high, the most exposed band.

Sell/avoid discipline:

• New capital at >$450: do not initiate (above the base-case fair-value zone)
• Re-engage starter (1–2%) only on a reset toward ~$410 with the business legs still intact
• Existing holders: trim on any 2-of-4 AI-capex air-pocket fire; otherwise hold the compounder
• Hard avoid-add if forward multiple >65× AND a hyperscaler cuts forward capex guide

XI. The 2-minute pitch

"AMD is the thesis that worked — and that's exactly why you don't buy it here.

Two years ago the bet was that the market mis-saw AMD as a GPU runner-up and ignored two hidden businesses. That bet won. Data Center just printed $5.8 billion in a single quarter, up 57% — the AI-inference franchise is real. EPYC has Intel structurally beaten. The balance sheet is net-cash $7.3 billion with $6.7 billion of free cash flow. Three legs, and the two that mattered both confirmed.

Here's the catch: the stock is $516. That's about 70× forward earnings — richer than NVIDIA — and Wall Street's own price target is $405 to $410, roughly 21% below where it trades. The cheap-optionality the thesis was built on has been arbitraged away by a 5× re-rating. Run the scenarios off the real price and you're risking 42% to make 36%, with the most likely outcome a 20% drawdown as the multiple normalizes to fair value. That's negative expected return and unfavorable asymmetry.

This is a wonderful business at a demanding price. The discipline is to put it on the watchlist, not in the portfolio — re-engage near $410 if the multiple resets, or if AMD proves consensus earnings are too low. Owning a great company is not the same as buying it at any price. Right business. Wrong price. Wait."

XII. Investment Scorecard (per MANUAL_en.md Part K.6)

15-question scorecard (analytical-tree Q-list, Format B per K.3.5)

Tally: 9 ✅ · 5 ⚠️ · 1 ✗

K.3.5 Weighted-score derivation

Applying Format B tier mapping (per MANUAL §K.3.5) — verdict values: ✅ = 1.0, ⚠️ = 0.5, ✗ = 0.0:

77% = high structural quality — but the score measures the BUSINESS, not the entry price. The single ✗ (Q5 valuation, load-bearing) is the binding constraint that the weighted score under-weights: a 77% structural score on a stock trading 21% above fair value is precisely the "wonderful company, demanding price" signature. Per the system's design, xii_score (structural, 77%) sits above h0 (thesis-confidence in a favorable outcome, 62%) — and the gap names the problem: the quality is real, the favorable-edge confidence is not.

K.3.1 fatal-flag override: 0 fatal flags. Business durability intact. The 77% structural score stands — but it does not translate to a buy, because the verdict incorporates price and asymmetry, which the scenario analysis (EV −13%, asymmetry 0.41×) drives.

Evidence tier: R2-VERIFIED (Tier-A on 10-K/10-Q financials; Tier-B on price/multiples). The Tier-C scaffold haircut is retired — this is no longer a screening memo.

Final verdict: WAIT — RIGHT BUSINESS, WRONG PRICE

One-line verdict: Confirmed three-platform AI compounder, fully-to-excessively valued at $516 (~70× forward, ~21% above Street fair value); 0% new capital at spot; re-engage starter near ~$410 or on evidence consensus EPS is too low.

Why not "Buy": EV is −13% and asymmetry is 0.41× (unfavorable) at spot. The market trades AMD above Wall Street's own fair value, into a downside-skewed distribution, mid-to-late in the AI-capex cycle. There is no asymmetry to harvest. Buying a confirmed thesis at a full premium is paying for certainty that is already in the price.

Why not "Avoid" (permanently): the business is genuinely high-quality (22/25 durability, net-cash, widening CPU moat, real inference franchise). This is a price problem, not a business problem. At ~$410 (base/fair value) or on a proven consensus-too-low EPS path, AMD becomes a legitimate 3–5% build. It is a WAIT, not a write-off.

Position posture:

• New capital today: 0% — above fair value, unfavorable asymmetry
• Re-engage trigger: price ≤ ~$410 (Street fair value) with business legs intact → 1–2% starter
• Scale path (only after re-engagement): toward 3–5% as multiple/fundamentals align; hard cap 5–7% (22/25 durability) at a reasonable multiple
• Existing holders (low basis): hold the compounder; trim only on 2-of-4 AI-capex air-pocket fire

Correlated exposure note (K.3.4): ai-capex-high. If combined AI-infrastructure exposure (AMD + NVDA + AJNMY + TOTDY) exceeds 20% of portfolio, AMD stays at 0% regardless of price until the bucket is trimmed.

Risk types present (per MANUAL Part K.4)

1. Valuation / multiple risk (primary) — at ~70× forward, a normalization to 45–55× alone is a −20% to −30% move with no business stumble
2. Cyclical / correlated risk — ai-capex-high; first-order exposure to a hyperscaler capex air-pocket, no CUDA cushion
3. Competition risk — hyperscaler custom ASIC (TPU/Trainium/MTIA/Maia) substitution in inference; NVIDIA Blackwell/Rubin supply normalization
4. Execution risk — sustaining DC growth to justify consensus FY2026 $50B revenue
5. Regulatory risk — BIS China GPU export restrictions (bounded)

When NOT to buy (per MANUAL Part K.5)

1. Because "the thesis is confirmed" — confirmation is why it's expensive; you are late, not early
2. Because "AMD is cheaper than NVIDIA" — it is not; AMD trades at a premium to NVIDIA on forward earnings
3. Because the stock keeps going up — momentum into a 70× multiple above fair value is the textbook late-entry trap
4. Because the business is high-quality — quality protects the business through a drawdown; it does not protect your entry multiple
5. If you already hold NVDA + AJNMY + TOTDY — adding AMD stacks ai-capex-high correlation; check K.3.4 before any re-engagement

All financial figures are R2-verified Tier-A (FY2025 10-K accn 0000002488-26-000018; Q1 2026 10-Q accn 0000002488-26-000076, SEC EDGAR). Price/multiples Tier-B (2026-05-29, 3-source). Forward scenario targets (†) are analytical constructs. Supersedes the 2026-05-10 GENERATE-mode scaffold; see update_2026-05-29.md.