Educational publication. Not investment advice. Neither John Gillespie nor InsightfulAgents.AI LLC is a registered investment adviser. The author does NOT hold AMD.
Data Window & Methodology Run
Run date: 2026-05-29. Jury workflow execution time: 2026-05-29 15:48 UTC. The four-lens reads consolidated from a Hybrid Valuation workflow run against TTM financial inputs. Workflow-snapshot market price: $511.89. The methodology output below is unchanged from the 2026-05-29 run; the dossier is published 2026-06-03 alongside the companion YouTube walkthrough.
Stored DCF intrinsic: $74.52/share. Conviction: 45. Valuation consensus: EXPENSIVE. Verdict: HOLD with asterisk. The asterisk preserves the cross-method disagreement on the record - DCF said EXPENSIVE 4/4; Relative Valuation said FAIR 4/4; the methodology refused to compress those into a smooth directional call.
Why this dossier exists
AMD is the boundary case that demonstrates cross-METHOD discipline. Previous Forensic Jury cycles showed lens-vs-lens disagreement inside the jury. AMD's jury was UNANIMOUS - all four lenses agreed the DCF says expensive. The disagreement lives one layer up: between the two valuation methods (DCF vs Relative Valuation) the methodology runs in parallel. When the two methods point in different directions, the methodology refuses to fake a clean answer.
For RIAs and HNW practitioners reading this: AMD is the canonical "the math says expensive but peers say fair" case. The audit-trace template at /reading-pack/ lets you reproduce this discipline on any holding.
Key Metrics
The Mode 5 Boundary Case — Cross-Method Discipline
The OCCAM'S FORENSIC JURY methodology runs two valuation methods in parallel on every company. The DCF method builds an intrinsic value estimate from the company's own cash flows. The Relative Valuation method compares the company's multiples to a peer cohort.
Both methods are accepted in equity research practice. Neither is inherently more legitimate than the other. They look at different evidence layers and can produce different verdicts on the same company at the same price.
On AMD, the two methods diverged cleanly:
| Method | Result | Reading |
|---|---|---|
| DCF | EXPENSIVE 4/4 | All four lenses concluded the DCF intrinsic value sits well below the market price. The cash-flow case is bearish. |
| Relative Valuation | FAIR 4/4 | AMD trades at the 60th percentile of its semiconductor peer cohort (ASML, MU, LRCX, ARM, QCOM). Peers don't price AMD differently than the DCF math would imply for the whole sector. The relative case is neutral. |
| Forensic Quality | CLEAR | Zombie: false. Rule of 40: 56.6. Divergence: WATCH (9.7) - not escalated to FAIL. Cash runway: 999 months (model ceiling). Operational health is not the question. |
The methodology's resolution. Four lenses agreed AMD looks expensive on cash-flow math. All four said SELL. The methodology still refused to publish SELL - because the peer-comparison check came back FAIR. Same multiples as the rest of the semiconductor cohort. The DCF problem isn't AMD's; it's the whole sector's.
So the methodology landed on HOLD with asterisk. The cash-flow case is bearish, the relative case is neutral, and compressing those into one directional call would have papered over the actual disagreement. The asterisk says: read both methods, not the headline.
Cross-cycle context: Mode 5 within the Five Modes canon
The companion essay Five Modes of Jury Disagreement walks through the methodology's five distinct disagreement modes:
- Mode 2 (MongoDB / MDB): jury split, one lens FLAGGED, override fires.
- Mode 3 (Qualcomm / QCOM): jury split, methodology holds the line, dissent stays on record.
- Mode 4 (NVIDIA / NVDA): lens-level dissent preserved through normal merge.
- Mode 5 (AMD, this dossier): jury unanimous on DCF SELL, cross-METHOD check (DCF vs Relative Valuation) refuses to publish SELL.
Mode 5 is structurally different from Modes 2-4. The earlier modes operate inside the jury - lenses disagreeing with each other. Mode 5 operates one layer up - the two valuation methods disagreeing with each other while the lenses themselves are unanimous within each method.
The Four Lenses
Four independent AI lenses ran on the same data. Each lens uses a different model provider under a persona-prompted framework. Each lens returned a verdict and confidence on both valuation methods. All four landed SELL on the DCF method and FAIR on the Relative Valuation method - a unanimous within-jury read on each side of the cross-method tension.
- The Auditor → SELL @ MEDIUM (DCF) / FAIR (Relative): Fortress balance sheet collides with a valuation that has decoupled from fundamentals. At the industry-anchor cost of capital, the stock prices in growth assumptions that contradict late-cycle positioning.
- The Architect → SELL @ MEDIUM (DCF) / FAIR (Relative): Strong operational metrics; DCF intrinsic dramatically below market; relative valuation FAIR vs peers is the only positive valuation signal.
- The Storyteller → SELL @ HIGH (DCF) / FAIR (Relative): Operational health is robust; the market is baking in a growth trajectory or profitability far more aggressive than the DCF assumptions support.
- The Sentinel → SELL @ HIGH (DCF) / FAIR (Relative): "It is mostly a price story, and the price is doing a lot of the work here." External signal: strong growth + bullish trend + overbought tape on one side; a late-expansion elevated-rate regime + DCF math that prices in far less than the market on the other.
The convergence is the story: every lens, every method, agreed within its own frame. The disagreement happened between the frames.
Lens 1 — The Auditor
The balance sheet. AMD presents a paradox. Fortress balance sheet ($5.5B cash, 999-month runway by the model ceiling), no solvency risk, and exceptional Rule of 40 execution at 56.6 (34.3% growth + 22.3% operating cash flow margin). The zombie and divergence frameworks clear easily. The forensic-quality gate is not the issue.
The valuation tension. At the industry-anchor cost of capital (Damodaran semiconductor industry CoC, around 10.5%), base intrinsic value is around $197 per share. Current price of $511.89 is around 157% above that base. Reverse-DCF at market price implies a required return around 6%, which sits well below the assumed cost of capital - untenable in an ELEVATED rate regime where the Fed holds 3.50-3.75% and the 10-year yield sits at 4.56%.
The anchor-choice question. Three of the four lenses ran the DCF at a company-specific beta-derived discount rate around 15%, which produces a stored intrinsic of $74.52 per share - around 580% overvalued by that anchor. Both anchors land on EXPENSIVE; the magnitudes differ materially. The industry-anchor case is arguably more defensible per framework discipline (Damodaran data is the public reference); the company-specific anchor is the deterministic-reference output. The verdict survives either way.
What the price requires. The market is pricing in sustained 22%+ growth with a sub-7% cost of capital, a bet that contradicts late-cycle positioning. ISM Manufacturing PMI sits at 52.7 (expansionary but with production cooling); S&P 500 reached all-time highs after an 8-week winning streak; AMD's quality composite percentile (0 vs peer cohort on ROE/ROIC) adds structural headwind.
Auditor verdict. The cash-flow math doesn't support the price at either anchor. Confidence MEDIUM because the deterministic reference uses a company-specific beta that may overstate risk, and the sensitivity bracket (low $118 / base $197 / high $296 at the industry anchor) brackets the range reasonably but late-cycle inflection risk isn't fully quantified in the model. SELL @ MEDIUM on the DCF; FAIR on Relative Valuation.
Lens 2 — The Architect
Financial health language. AMD's financial health appears robust. Cash runway of 999 months indicates formidable liquidity. Rule of 40 score of 56.6 indicates solid operational performance. The relative valuation reads FAIR compared to peers - a positive signal in the current market context.
The DCF picture. Intrinsic value calculated at $74.52 against a current market price of $511.89 - a significant premium. The discrepancy raises questions about the sustainability of the valuation, especially in an elevated rate environment. Regime context (late expansion, elevated rates) suggests caution on the forward growth profile.
Mode tension. The strong operational metrics on one side and the substantial DCF gap on the other create the cross-method tension this dossier preserves. The relative valuation FAIR call is the only positive valuation signal. Without it, the verdict resolves to SELL; with it, the methodology holds at HOLD with asterisk.
Architect verdict. Strong operational metrics; current valuation may warrant caution. SELL @ MEDIUM on the DCF; FAIR on Relative Valuation.
Lens 3 — The Storyteller
Fascinating dichotomy. Operational health is robust - the company is far from a zombie, with a massive cash runway, and the Rule of 40 score is excellent. These metrics paint a picture of a financially stable and growing enterprise.
What the valuation says. Relative valuation suggests AMD is fairly priced compared to its peers; the DCF analysis says the opposite. The deterministic DCF reference, even when adjusted for optimistic and pessimistic scenarios, yields intrinsic values dramatically lower than the current market price. The market is baking in a growth trajectory or profitability that is far more aggressive than the model's assumptions, or perhaps valuing AMD on metrics not fully captured by a standard FCF DCF.
The unsaid. The divergence flag - though only WATCH at 9.7, not escalated to FAIL - hints at some underlying disconnect that might be worth exploring further. The Storyteller's job is to surface what the data doesn't say out loud. The methodology refuses to call the cross-method gap a settled question; the unsaid is the unresolved.
Storyteller verdict. The cash-flow story and the relative-valuation story don't reconcile. The methodology shouldn't force a synthesis. SELL @ HIGH on the DCF; FAIR on Relative Valuation.
Lens 4 — The Sentinel
The external read. AMD clears the cash and operating-health screens; the company is not a zombie, Rule of 40 sits at 56.6, and the relative read is only FAIR. Then the DCF comes back with a base value around $74.52 against a market price near $512. That is not a small miss. It is an expensive stock by the model bundle, even after giving the range some room on both sides.
The internal mix. A strong growth profile, a bullish trend, and an overbought tape on one side; a late-expansion, elevated-rate regime and a DCF that prices in far less than the market on the other. The divergence flag is only WATCH at 9.7, so this is not a collapse story. It is mostly a price story, and the price is doing a lot of the work here.
Macro regime context. The Sentinel pass referenced the regime composite tag set by the macro evidence layer: elevated-rates / late-expansion / sector-dispersed / normal-vol. Primary-source anchors from the regime layer include Fed Funds 3.50-3.75% (held at the most recent FOMC), 10-year Treasury at 4.56%, ISM Manufacturing PMI at 52.7 (18 consecutive months of expansion), and Damodaran implied ERP at 4.77%. These are publicly verifiable; the dossier preserves the URL list for any reader who wants to audit the inputs.
Sentinel verdict. No fresh external bearish catalyst; no evidence the current price is supported by a new fundamental step-change outside what the dossier captured. The market is doing momentum continuation. Mostly a price story. SELL @ HIGH on the DCF; FAIR on Relative Valuation.
Primary-Source Anchors (Sentinel Pass)
Regime evidence layer - publicly verifiable URLs at workflow-run time:
- Fed Funds 3.50-3.75% (Apr 29 FOMC held): federalreserve.gov press release
- 10-year Treasury 4.56% (May 22 close): advisorperspectives.com
- ISM Mfg PMI 52.7 (Apr 2026, 18 consecutive months expansion): prnewswire.com
- S&P 500 close 7473.47; 8-week winning streak: tradingeconomics.com
- VIX 16.70 (low-normal boundary): FRED VIXCLS
- IG OAS 77 bps (May 12 vintage): investmentgrade.com
- Damodaran implied ERP 4.77% (Apr 1 2026 vintage): Damodaran NYU Stern
Stale-vintage flags worth knowing: Damodaran ERP at Apr 1 vintage (~57 days at run date; threshold ~60 days, so marginal). IG OAS at May 12 vintage. S&P EW YTD at May 15 vintage. The dossier preserves these so a reader can re-run the inputs against fresher data at any time.
The Convergence Map
The four lenses agreed within each method. The disagreement happened between the methods:
- All four agree (DCF method): AMD's DCF intrinsic value sits well below market price at either discount-rate anchor. The cash-flow case is bearish. The verdict is EXPENSIVE 4/4.
- All four agree (Relative Valuation method): AMD trades at the 60th percentile of its semiconductor peer cohort (ASML, MU, LRCX, ARM, QCOM). The relative case is neutral. The verdict is FAIR 4/4.
- All four agree (Forensic Quality): Operational health is not in question. Zombie: false. Rule of 40: 56.6. Divergence: WATCH only.
- Where the methodology diverges: The two valuation methods produce different directional reads. Compressing that into a smooth verdict would hide the disagreement. The methodology refuses; the verdict carries an asterisk; both reads stay on the record.
The methodology preserves the cross-method disagreement on the public record. A reader picking up this dossier next quarter - when earnings move the DCF inputs, or when peer-cohort multiples shift - can re-evaluate the two methods independently without rerunning the lens jury. That is the point of preserved cross-method discipline: forward-state reproducibility on the same architecture.
Bias Flags & Risk Callouts
- WACC anchor split (~480 bps gap). Auditor used the Damodaran semiconductor industry CoC (~10.5%); the other three used a company-specific beta-derived rate (~15%). Both produce EXPENSIVE; magnitudes differ materially (~157% overvalued at the industry anchor; ~580% at the company-specific anchor). The arbiter let valuation_consensus resolve at clean EXPENSIVE because both paths converge on the same verdict.
- Quality composite percentile 0 vs peer cohort on ROE/ROIC. AMD trails ASML / MU / LRCX / ARM / QCOM on the quality metric. May be a cyclical trough artifact from AI ramp investment phase, or structural underperformance. Affects whether EXPENSIVE is justified by quality premium or penalized further.
- Divergence Razor WATCH (9.7). Not escalated to FAIL. The specific divergence driver (earnings quality / revenue recognition / margin sustainability) is not surfaced in the methodology output and would benefit from manual audit.
- Regime dependency HIGH. The Auditor narrative explicitly flags HIGH regime dependency - any deceleration in growth or rise in rates compresses the valuation multiple sharply. The methodology's regime-conditioned resolution states (CHEAP_REGIME_FLAGGED / EXPENSIVE_REGIME_FLAGGED / CONFLICTED_GENUINE) did not fire to REGIME_FLAGGED on this run; verdict landed clean EXPENSIVE.
- Reverse-DCF inconsistency. At $511.89, what growth and terminal assumptions does the market appear to be pricing in? A clean reverse-DCF consensus would clarify whether the price requires implausible assumptions. The current jury reads do not converge on a single answer.
- Late-expansion cycle position. S&P 500 reached all-time highs after an 8-week winning streak (longest since Dec 2023); IG OAS at 77 bps (near historical tights). Late-cycle positioning amplifies the regime-sensitivity risk for premium valuations.
Peer Comparison Snapshot
AMD sits at the 60th percentile of its semiconductor peer cohort on relative valuation - mid-pack, not standout-expensive vs peers. That is the load-bearing fact that produces the FAIR call on the Relative Valuation method.
| Ticker | Role in cohort | Note |
|---|---|---|
| AMD | Subject company | 60th percentile on the cohort relative-valuation read |
| ASML | Lithography monopoly | Semi cap equipment; structural moat |
| MU | Memory cycle exposure | Commodity memory cycle; cyclical |
| LRCX | Etch / deposition equipment | Semi cap equipment peer |
| ARM | IP licensing model | Different business model; included for cohort breadth |
| QCOM | Communications / mobile | Companion dossier QCOM FLAGGED case |
The cohort selection reflects the methodology's peer-comparable group at run time. AMD at the 60th percentile means roughly four out of every ten peers trade richer; six trade cheaper. The relative-valuation FAIR call is the structural reason the methodology did not publish a clean SELL.
What Could Change This Read
The methodology returns HOLD with asterisk today because the cross-method tension is structurally unresolved. Five forward-state changes would materially shift the verdict:
- Earnings trajectory closing the DCF gap. If forward earnings sustain growth at levels that match the market-implied required return, the DCF intrinsic re-computes higher and the gap shrinks. Conversely, an earnings deceleration would widen the gap and pressure the relative-valuation FAIR call.
- Peer-cohort multiple compression. If the broader semiconductor cohort de-rates (peer multiples compress), AMD's relative-valuation read could move from FAIR toward EXPENSIVE - which would resolve the cross-method tension toward a clean SELL.
- Rate regime shift. A material easing cycle (Fed cuts; 10Y below 3.5%) reduces the cost-of-capital pressure that drives the DCF gap. The methodology re-runs would likely land at a smaller intrinsic-vs-market gap. The verdict could move toward HOLD without asterisk.
- Quality composite improvement. If AMD's ROE/ROIC moves from the 0 percentile toward the peer median, the quality-premium argument strengthens the relative-valuation FAIR call and weakens the DCF-EXPENSIVE conclusion (a higher quality firm can defend higher multiples).
- Cross-method canonical resolution. The methodology's regime-conditioned resolution layer can fire EXPENSIVE_REGIME_FLAGGED when the DCF says EXPENSIVE and the regime sensitivity puts pressure on the call. The current AMD run resolved to clean EXPENSIVE (no REGIME_FLAGGED), but the underlying narrative explicitly says regime dependency is HIGH. A subsequent fire under a more pressured macro setup could surface this resolution.
The methodology will re-run AMD on its standard cadence. Each re-run lands as a new dated dossier with the prior verdict preserved on the public record for trajectory tracking.