Financebotresearch desk研究台

ADI › analyze

ADI · Analyze

Semiconductors

A portfolio-specific passage was removed from the public build.

⚠️ Price moved during the analysis. All models were run at $353.37; ADI last printed $366.30. Every "distance to fair value" figure below is therefore ~3.7% worse than the model tables show. The drawdown from the $445.91 high is now −17.8%, not −20.8%.


1. Verdict First

🟡 PASS at $366.30. Watch. Conviction 5/10 — but for the opposite reason to NXPI.

Fair value $255–305 (midpoint ~$280). At $366.30 the stock is ~31% above midpoint and well above the top of the range.

ADI is a genuinely better business than my earlier compare note credited, and a worse entry. The debate round overturned two of the Moat Analyst's marquee bear claims using primary sources — the "margins never recovered post-M&A" argument is a purchase-accounting artifact, and the "ADI's moat is weakest in gluts" prediction was falsified by the 2023–25 downcycle that has already run. ADI won that natural experiment against TI decisively.

But quality was never the question. The question is price, and the price requires ADI to reproduce organically, for a decade, the growth it previously achieved with a $30B chequebook.

ADI NXPI (analyzed today)
Business quality Higher — moat confirmed, margins expanding, won the downcycle Good, narrowing
Buyback Real — 5.88× SBC coverage… but see the correction in §3 Zero net (buyback ≈ SBC)
Dividend 22-yr streak, +11.1% raise, 45% FCF payout Frozen ~10 quarters
Balance sheet Fortress — 18% debt/assets, 9.2× coverage Levered, negative tangible book
Hidden capex None — verified clean ~$583M/yr of fab-JV equity
Reported ROIC 6.1% — below WACC 12.6%
Price vs fair value ~31% above ~5–15% above
Verdict PASS — better business, worse price PASS — cheaper, weaker

Conviction is 5 because the business case is strong (I'd rate the business 8/10) and the entry case is weak (3/10). This is the framework's "Great company + Expensive → wait and watch" cell, held with more confidence than NXPI's because ADI's quality is less ambiguous.

Dimension scorecard

Dimension Grade One-line
Financial health A Fortress. 18.05% debt/assets, 1.25× net debt/EBITDA, laddered to 2054, notes trade $700M below par.
Cash generation A− Record TTM FCF $4,565M at 36% of revenue. ~$481M of FY2025 was a working-capital flatter.
Capital allocation B+ Zero platform M&A in 5 years, R&D raised through the trough to a record 16.0% of revenue. But distributing 110.5% of FCF and buying back at 43× adjusted earnings.
Dividend (§4 overlay) A 22-year streak, raises every February without a miss, latest +11.1% — above the 9.5% 10-yr average. FCF payout 44%.
Moat A− / 4.0-of-5 Confirmed under adversarial stress-testing. Won the downcycle against the cost leader.
Growth quality C+ ~4.4% organic revenue CAGR over 9 years; roughly half the 8-year revenue increase was purchased.
Returns on capital C+ Reported 6.14% is below WACC; ex-goodwill 18.9% — but that was 36–39% pre-Maxim.
Valuation D+ Requires ~13.4% FCF growth for a decade at r=9%. Demonstrated organic: 4.25–6.5%.
Sentiment 6.5/10 Beating guidance, hostile tape, ~$104M insider selling and $0 of buying in 24 months.

2. The Debate Round — Three Bear Claims Tested, Two Broken

Three analysts reached structurally incompatible conclusions from correctly-computed numbers. A debate round adjudicated using primary SEC sources. The findings changed the verdict materially, in both directions.

⭐ Q1 — "ADI monetized its moat away." RULING: arithmetically true, economically false.

The Moat Analyst's most damaging claim: standalone ADI earned 65.1% GAAP gross margin in FY2016; post-Linear/post-Maxim ADI earned 61.5% in FY2025below the pre-acquisition level. So $36B of M&A bought nothing.

Both figures verify exactly (FY2016 10-K: $2,227.2M / $3,421.4M = 65.09%; FY2025 10-K: $6,773.5M / $11,019.7M = 61.5%). FY2016 was not cherry-picked — ADI's pre-acquisition band ran 63.9–65.8% across FY2012–16.

But the comparison is not like-for-like, and the gap is fully explained by purchase accounting. ADI's own FY2025 Q4 reconciliation shows $867.6M of acquisition-related amortization sitting inside FY2025 COGS = 787bps of GAAP gross margin. FY2016 carried at most $75M of intangible amortization across the entire P&L — a ceiling of 220bps, most of it in opex.

Basis FY2016 FY2024 FY2025 Q2 FY2026
GAAP gross margin 65.1% 57.1% 61.5% 67.3%
Acquisition amortization in COGS ~0 $1,022.5M $867.6M $205.5M
Adjusted gross margin ~65–66% 67.9% 69.3% 73.0%

Like-for-like: +330 to +430bps of expansion at a comparable cycle point, and +700–790bps at peak.

And the GAAP-only proof that needs no non-GAAP charity: FY2018 GAAP gross margin was 68.3% — with Linear purchase accounting fully running at $571M/yr — which is 320bps ABOVE standalone FY2016. FY2019 (67.0%) and FY2020 (65.9%) also cleared it. The Linear deal raised GAAP gross margin. The post-FY2021 decline maps exactly onto Maxim's $28B step-up entering COGS plus the FY2024 cyclical trough.

Verdict: the deals PRESERVED and modestly EXPANDED the product-level moat. The Moat Analyst conflated two distinct claims — "the moat got worse" (false) and "the return on capital got worse" (true, and severely). ADI bought a genuinely good moat at a genuinely bad price. Those are not the same failure, and only one is a moat failure.

The number no Phase-1 analyst produced: FCF/share went $3.73 (FY2016) → $8.61 (FY2025) = +9.74%/yr over nine years — through 68% share dilution. On the user's own §0 principle ("per-share is what I own"), the M&A era worked. But it decays hard: FY2018→FY2025 is +5.73%/yr and FY2022→FY2025 is +6.04%/yr. Linear did the work; Maxim mostly did not.

⭐ Q2 — "ADI's moat is strongest in shortages, weakest in gluts." RULING: FALSIFIED. The experiment already ran, and ADI won it.

This was the Moat Analyst's central forward-looking prediction, and the 2023–25 analog glut is the natural experiment. It has concluded.

Peak → trough Trough → FY2025 FY2022 → FY2025
ADI revenue −23.4% +16.9% −8.3%
TI Analog segment −20.8% +15.2% −8.8%
TI total −21.9% +13.0% −11.7%

Revenue: a dead heat. No share shift. (TI's fiscal year ends six weeks later, which flatters TI's recovery column.)

Gross margin: ADI held, TI collapsed.

FY2022 FY2023 FY2024 FY2025
ADI GAAP GM 62.7% 64.0% 57.1% (trough) 61.5%, recovering → 67.3% in Q2 FY26
TI GAAP GM 68.8% 62.9% 58.1% 57.0%, still falling

ADI degraded 250bps peak-to-current; TI degraded 1,180bps. ADI now runs 450bps ABOVE TI on GAAP gross margin. That is the precise inverse of what a cost-advantage thesis predicts.

Free cash flow: the decisive column.

FY2022 FY2023 FY2024 FY2025 FY22→25
ADI FCF $3,776M $3,556M $3,122M $4,279M +13.3%
TI FCF $5,923M $1,349M $1,498M $2,603M −56.1%
TI capex $2,797M $5,071M $4,820M $4,550M $14.4B over 3 years

Verdict: over the tested window, product position beat cost position decisively. TI's 300mm die-cost advantage is real and the engineering description is not in dispute — but it has so far been consumed by depreciation and underutilization on a $14.4B three-year build.

Honest counterweight: TI's capex is now falling and its operating leverage on a real recovery is enormous; a 2028 rerun could invert this table. And ADI's ROE of 10% vs TI's 35% is a permanent structural gap caused by $27B of goodwill that will never earn a return. The Moat Analyst was right that ADI has a price premium rather than a cost advantage — the error was inferring that a price premium is fragile in a downcycle. It wasn't.

Q3 — Is the datacenter business moat-accretive or moat-dilutive? RULING: SPLIT — and both Phase-1 analysts were wrong about which part is which.

The discriminating fact neither found is the mix. On the Q1 FY2026 call, Roche disclosed that hot-swap/protection is "roughly one third of our data center power revenue" and DC power control "roughly one third," with power and optical "pretty balanced."

Doing the arithmetic: of ~$1.66B annualized data center revenue, only ~$280M — roughly 2.5% of company revenue — is the contested multiphase/vertical power-delivery socket. Two-thirds of the power half is protection, telemetry and hot-swap: Linear-heritage, high-ASP, high-reliability, low-competition parts that behave far more like ADI's industrial book than like an AI socket. Content per rack rises mechanically with rack power and the 800VDC transition. That is a genuinely good business and it is under-narrated.

But the strategic direction is the dilutive part, and the evidence is specific: - The margin test cuts against the contested socket. MPS — the company that actually holds the power-delivery socket — ran 55.5% gross margin, flat for four straight quarters, at the low end of its own model, while Enterprise Data grew +97.7% yoy. Its 2022 quarterly GMs ran 57.9–58.8%. For the incumbent, AI datacenter has been ~300bps gross-margin dilutive. - The Vicor/MPS precedent verifies — Vicor was designed out of the H100 and replaced by MPS on cost and manufacturability (~$50 vs $100+), with product revenue −19.7% in FY2024 while NVIDIA data center grew >100%. But it does not generalize the way the Moat Analyst claimed: Vicor is still not back at NVIDIA in mid-2026, and MPS absorbed a full sole-source→multi-source conversion and still grew +26.4%. The correct reformulation: no single design win is an annuity, but the franchises are durable. What kills you is being a single-socket, single-customer company with a fragile balance sheet. ADI is neither. - Empower is an admission, and ADI said so out loud. Roche on the Q2 call: "The reason that we acquired Empower is that there was a gap in that portfolio, and time is of the essence." $1.5B all cash, completed 2026-07-07. The press release hedges: silicon capacitors are "already in production" but IVR "programs are advancing" — the $1.5B thesis product is not shipping, no customer named, revenue explicitly deferred to 2027. Empower closed a >$140M Series D in September 2025; $1.5B eight months later is a fast, large step-up. - 🚩 ADI has not named a single data center customer across four consecutive earnings calls. The one hard datapoint is Q1 FY2026: "We shipped our Smart PowerStage to our first vertical power customer last quarter." Customer count: one, unnamed. MACOM is publicly sole-source on NVIDIA 1.6T; MPS discloses a named segment. Companies that hold sockets signal it. ADI describes weather.

What survives is narrower than the narrative: ADI owns a high-margin protection-and-telemetry franchise whose content rises with rack power. That is worth owning. It is not "ADI is an AI power leader," and the multiple is being underwritten as though it were.

⭐ Q4 — The "$20B by 2030" vision. RULING: a gradeable scorecard DOES exist, and ADI missed 4 of 5.

The Sentiment Analyst reported that no ADI Investor Day scorecard exists — the artifact that proved most predictive in the NXPI analysis. That is wrong. ADI held a public Investor Day on 2022-04-05 and published a long-term financial model.

Target (set 2022-04-05) Promised Delivered FY2022→FY2025 Verdict
Revenue CAGR 7–10% −2.84% ($12,014M → $11,020M) SEVERE MISS
Adjusted gross margin floor ≥70% 73.6 / 72.5 / 67.9 / 69.3 Floor broken 2 of 4 years
Adjusted operating margin 42–50% 49.4 / 48.9 / 40.9 / 41.9 ❌ Below range 2 of 4
FCF margin 34–40% 31.4 / 28.9 / 33.1 / 38.8 ❌ Below range 3 of 4
Capex % of revenue 4–6% 5.8 / 10.3 / 7.7 / 4.8 ❌ Breached 2 of 4
Capital return 100% of FCF 109 / 131 / 77 / 96% → 103.6% cumulative MET

1 of 5 met — and the one met was the capital-return promise, which is the easiest to keep. There has been no investor day since 2022-04-05.

The pattern that matters: management still quotes fragments of the 2022 model on calls (capex 4–6%, 100% FCF return) while never re-baselining the revenue and margin targets it missed. That is selective retention of a stale framework.

On "$20B by 2030" specifically — CFO Puccio, Q2 FY2026 call, verbatim: "we are comfortable that we have the capacity to support, you know, up to the $20 billion that we have been talking about as part of our 2030 vision." This is a fab-capacity statement, not guidance. From FY2025 it implies ~12.7%/yr, above the top of a range ADI has never hit and just missed by ten points. Grade: capacity marketing. Do not model it.

Mitigating, and worth stating fairly: FY2026 is back inside every line of the 2022 model (Q2 adjusted GM 73.0%, adjusted operating margin 49.0%, TTM FCF 36% of revenue). The misses were substantially cyclical. But a company that will not re-publish a model it missed has removed the artifact that made it accountable.

Q5 — Cybersecurity and MOFCOM. RULING: both real, both materially SMALLER than Sentiment weighted them.

Cybersecurity — LOW risk, and the item number is the tell.

The 8-K (filed 2026-07-29, event date 2026-06-23) was filed under Item 8.01 (Other Events), NOT Item 1.05 (Material Cybersecurity Incidents) — and ADI does not appear among Item 1.05 filers for the period. ADI affirmatively states it "does not believe the June 23, 2026 incident is reasonably likely to materially impact its business." Only ~10% of Item 8.01 cyber filings ever escalate to 1.05.

The IP-exfiltration framing is unsupported. ExfilSquad is a real tracked extortion group, but it claimed ~570,000 records of customer PII — not design or process data — and ADI is no longer listed on the leak site, with nothing published. ADI itself discloses no data classes, no record counts, no systems.

The precedent base rate is decisive:

A portfolio-specific passage was removed from the public build.

The strongest finding: every court-measured semiconductor IP-theft award traces to insiders walking out (ASML v. XTAL, UMC/Micron), not to network intrusion. A stolen GDSII or RTL tree without the fab, PDK, process recipes and yield history is largely inert. For a semiconductor company, cyber is a production-timing risk, not a franchise risk.

Residual genuine tail: the July 26 matter is days old and unassessed. That is the only part not yet dismissible.

MOFCOM — real, narrow, and probably leverage rather than punishment.

Initiated 2025-09-13; deadline 2026-09-13, extendable six months. Scope is a single HS code (85423990): ≥40nm CAN/RS-485 transceivers, I²C interface chips, digital isolators, gate drivers. MOFCOM's announcement names no companies — the ADI/TI/Broadcom/onsemi list is press reporting, and the 340% figure is the petitioner's allegation, not a MOFCOM finding.

No preliminary determination, no provisional duties, no cash deposits exist as of 2026-07-30. Chinese AD cases typically issue a preliminary determination within 8–10 months when Beijing intends duties. We are at T+10.5 months with nothing — strong evidence the case is being held as negotiating leverage ahead of the 2026-11-10 truce expiry.

Two facts that cut ADI's exposure hard: 1. ADI has never used the words "anti-dumping" in any SEC filing (EDGAR full-text, all filings, zero hits) — and the FY2025 10-K was filed two months after initiation. That is a considered legal judgment that the exposure is immaterial. 2. China's origin test for ICs is wafer-fabrication location. ADI sources "more than half of our wafer requirements annually from third-party wafer fabrication foundries, such as TSMC," and its remaining internal fabs are Wilmington MA, Beaverton OR, and Limerick, Ireland. Taiwan- and Ireland-fabbed parts are not US-origin.

Estimated exposure: $70–250M of revenue at pricing risk — under 2% of ADI. TI's exposure is several multiples of ADI's. Meanwhile TI is raising prices into China, which is not the behavior of a company bracing for a dumping finding.

Q6 — "Record results" vs "$104M sold, $0 bought." RULING: Sentiment is right, and understated the one item that matters — but Fundamentals overstated the buyback by ~2×.

Parsing 154 Form 4s / 382 transaction lines from raw SEC XML, with codes separated:

Window Code S (open-market sales) Code F (tax withholding — not sales) Code P (purchases)
TTM $104.2M (324,941 sh) $25.0M $0 — ZERO
24 months $159.5M (576,213 sh) $45.9M $0 — ZERO

Zero open-market purchases in 24 months verifies — not one P-coded line. Roche alone is 61% of TTM selling.

⭐ The CEO block was DISCRETIONARY — worse than Sentiment reported. Roche's 60,000 shares across 2026-05-22 and 2026-05-26 at ~$405.97 = $24,358,406. The Rule 10b5-1(c) affirmative-defense checkbox is NOT checked on any of the four lines, and there is no plan-adoption footnote. Independently corroborated: the Q2 FY26 10-Q Item 5 table lists only Ray Stata and Karen Golz as plan adopters — Roche is absent.

And it was incremental, not substitutive: he also sold his mechanical 10,000 shares on 2026-05-01, 06-01 and 07-01 under plan. His pattern since mid-2024 is exactly 10,000 shares on the first trading day of each month, unbroken. The 60,000 was a 6× departure, executed in the open window days after a record print at then-all-time highs.

Two mitigants I weight seriously: it was an exercise-and-sell off two 30,000-share option exercises at $144.06 (likely expiry-driven), and Roche's direct holdings rose from 51,039 to 137,538 shares over the period despite $90M of sales — PRSU vesting outpaces his selling. He is not net-liquidating. Heavy GRAT/foundation activity reads as estate planning.

Golz was mis-characterized by Sentiment — the 1,000 shares at $389.83 on 2026-07-13 verify, but the 10b5-1 checkbox is set with a plan adopted 2026-03-12. Plan-driven. Discount it entirely.

Other discretionary selling worth flagging: 24-month discretionary total $60.9M (38% of all selling) — Roche $33.6M, Martin Cotter (SVP) $14.6M, 100% discretionary, and Michael Sondel, Chief Accounting Officer, $6.1M, all discretionary. A CAO selling $6M outside a plan is worth noting given proximity to the numbers.


3. ⚠️ Two Corrections to the Fundamentals Case

The buyback is smaller than reported — and this is material

The Fundamentals Analyst's headline finding was that buybacks cover SBC 5.88× with $3,043M of genuine net retirement, making ADI "the mirror image of NXPI." The direction is right. The magnitude is not.

  1. The Q2 figure includes tax withholding. The 10-Q issuer-purchases table shows 2,319,488 shares at $332.09 = $770.3M — but footnote (a) discloses 381,348 of those shares were withheld for employee tax obligations. Real open-market program repurchase was ~$652M — the headline overstates it by ~18%.
  2. The per-share tailwind is roughly half what was claimed. Cover-page shares outstanding: 496,216,857 (2024-05-04) → 487,087,040 (2026-05-02) = −1.84% over 24 months ≈ −0.92%/yr. Share count was flat for a full year through FY2024–mid-FY2025 — buybacks merely offset dilution. The "+1.7pp/yr" figure is a FY2022-anchored artifact; the sustainable current-price rate is ~1.0–1.2%/yr.

ADI is still a genuine net retirer — unlike NXPI, which is not — but the per-share engine runs at about half the advertised speed. The re-rating has eaten the buyback's power. At a ~2.6% FCF yield, ADI is retiring shares at a low internal rate of return.

🚩 A $3.0B revolver nobody surfaced

8-K filed 2026-07-02: ADI entered a new $3.0 BILLION 364-day revolving credit facility with BofA as agent, expiring 2027-07-01, with an option to term out into a one-year non-amortizing loan, and a 3.00× EBITDA/interest covenant. Undrawn as disclosed; the agreement permits proceeds for stock repurchases.

A $3B 364-day facility with a term-out option is the classic M&A-bridge structure — signed six weeks after a $1.5B cash acquisition, by a company with $2.4B of cash and $4.5B of FCF that did not need it. This is the most forward-looking governance fact available and no Phase-1 analyst surfaced it. It is item #2 in §9.

Also unreported: ADI is distributing more than it earns. TTM FCF $4,565M against dividends $1,998M + repurchases $3,045M = $5,043M = 110.5% of FCF, funded partly with $1.5B of new debt. Sustainable only if FCF rises to $5.5B+ — which Q3 guidance implies, but it is not slack.


4. Fundamentals

Free cash flow — 10 years

FY 2016 2017ˢ 2018 2019 2020 2021ˢ 2022 2023 2024 2025 TTM
OCF $M 1,291 1,154 2,442 2,253 2,009 2,735 4,475 4,818 3,853 4,812 5,107
Capex $M 127 204 255 275 166 344 699 1,262 731 534 541
FCF $M 1,164 950 2,188 1,978 1,843 2,391 3,776 3,556 3,122 4,279 4,565

ˢ Stub years — never use as a CAGR endpoint. Linear closed 2017-03-10 (~8mo contribution); Maxim closed 2021-08-26 (~2mo).

FCF CAGR window Rate Usable?
5yr FY2020→FY2025 +18.4% ❌ M&A-inflated
8yr FY2017→FY2025 +20.7% Most-flattered number in the file — stub base and both deals
Cycle-neutral (avg FY16–18 → avg FY23–25, 7yr) +14.3% ⚠️ Best raw read, but embeds Linear + Maxim
Organic FY2022→FY2025 +4.25% ✅ Clean, but peak-to-recovery
TTM vs FY2022 peak (3.5yr) +6.5%/yr The fairest organic read

⚠️ FY2025 FCF was flattered by a +$481M working-capital swing (accrued comp $220M, deferred income/price-adjustment reserves $276M) → ex-WC ≈ $3,798M, ~11% lower than headline.

One-off scan: clean. No analogue to NXPI's $2.0B Qualcomm break fee. Restructuring cash drag immaterial (accrued balance $4.1M). 29 consecutive years of positive FCF.

Capital allocation

FY Capex R&D M&A cash Buybacks Dividends % of FCF returned
2016 127 654 83 370 513 75.9%
2017 204 968 9,633 (Linear) 47 602 68.2%
2020 166 1,051 0 245 886 61.4%
2021 344 1,296 (2,451) 2,605 1,109 155.3%
2022 699 1,701 0 2,577 1,545 109.2%
2023 1,262 1,660 0 2,964 1,679 130.6%
2024 731 1,488 0 616 1,796 77.2%
2025 534 1,766 46 2,165 1,924 95.6%
TTM 541 46 3,045 1,998 110.5%

ᵃ Maxim was all-stock, so ADI acquired $2.45B of net cash — an inflow, not an outlay.

  • The ~100% payout is deliberate policy dating to FY2021, not opportunism: FY2016–20 averaged 63.7% of FCF returned; FY2021–25 averaged 113.6%.
  • ADI stopped acquiring at scale in August 2021 — cash M&A was literally $0.0M in FY2022, FY2023 and FY2024. (Until Empower, $1.5B, closed 2026-07-07.)
  • R&D was raised straight through the trough to a record $1,766M = 16.0% of revenue, while capex was cut from $1,262M to $534M. That is the correct priority for a fab-lite analog franchise.

Leverage — a non-issue

FY Total debt Debt/Assets Net debt/EBITDA EBIT/Interest
2017 $7,551M 35.7% (post-Linear) 3.72×
2020 $5,145M 24.0% 1.77×
2021 $6,770M 12.9% (Maxim all-stock) 1.73×
2024 $7,582M 15.7% 1.35× 6.3×
2025 $8,664M 18.05% 1.25× 9.2×

Linear was the leveraging deal; Maxim was the de-leveraging deal (all-stock, added $30.9B of equity and zero debt). Blended cost of debt ~3.4%; only $447M matures within 12 months against $2.5B of cash; laddered to 2054; the notes trade ~$700M below par — an unrecognized asset. Leverage has crept back up since FY2021 not from acquisitions but from funding buybacks with debt.

⚠️ Goodwill $26,945M = 56% of assets, 80% of equity. With acquired intangibles, 72.8% of the balance sheet is purchase accounting. Derived tangible book value ≈ −$1,155M, or −$2.35/share.

ROIC — the real weakness

FY NOPAT (GAAP) Invested capital ROIC ROIC ex-goodwill
2018 $1,735M $16,726M 10.37% 38.78%
2020 $1,395M $16,087M 8.67% 36.61%
2022 $2,908M $41,543M 7.00% 19.88%
2023 $3,512M $41,556M 8.45% 23.98%
2025 $2,452M $39,908M 6.14% 18.91%

Reported 6.14% is almost certainly below cost of capital (beta 1.19, after-tax cost of debt ~3.4% → WACC 8–10%). Ex-goodwill 18.91% shows the operating business earns excellent returns — but it ran 36–39% in FY2018–20, pre-Maxim. The deal roughly halved return on capital, permanently.

Incremental ROIC is not computable — and that itself is the finding. Since FY2022, invested capital shrank $1.64B while $15.2B went out to shareholders. There is no incremental capital on which to measure a return.

This is a materially different and better situation than NXPI, whose incremental ROIC was negative on positive incremental capital — actively destroying value with new investment. ADI is not investing incrementally at all; it is harvesting a paid-for asset base and routing the cash out. FCF-return-on-invested-capital actually rose from 9.09% (FY2022) to 11.44% (TTM) on a shrinking base. The strategic read: ADI's problem is not "where do we earn a return" but "we have no high-return use for the cash." That justifies the payout policy — and caps the growth rate.

⚠️ The effective tax rate is structurally rising: 6.9% (FY2020) → 16.4% (FY2025) — a direct NOPAT and EPS headwind that every CAGR table masks.

Growth quality

Metric Raw CAGR Honest read
Revenue 8yr +10.1% ❌ Roughly half the 8-year revenue increase was purchased
Organic revenue, ~9yr (estimate) ~4.4% ✅ The number nobody quotes
Revenue/share 8yr +5.39% The most honest single growth number in the file
Net income 8yr +13.8% ❌ M&A-inflated

But the organic picture is better than the CAGRs imply. TTM revenue of $12,740M exceeds the FY2022 peak by 6.0%, and the Q2 run-rate annualizes to ~$14.5B, +20.7% above the prior peak. TTM net income has essentially matched the all-time FY2023 record on lower revenue. Peak-to-peak, ADI is genuinely growing organically with margins above the last peak. The cycle, not the business, makes FY22→FY25 look bad.

Cash-flow quality — verified clean

FY Capex % revenue Capex/Depreciation Net PP&E
2020 3.0% 0.71× $1,121M
2023 10.3% 3.77× $3,219M
2025 4.8% 1.31× $3,316M

Capex/depreciation of 1.31× means the asset base is still growing — the opposite of NXPI's 0.65×. Net PP&E is up 196% since FY2020 after a deliberate ~$2.7B FY2022–24 build-out. Management guides FY2026 capex to 4–6% of revenue — normalizing, not cutting.

No hidden capacity commitments. The 10-K contractual-obligations table was read line by line: no fab JV, no capacity prepayment, no take-or-pay. Only $187M of venture-fund commitments (digital biology/life sciences — strategic VC, explicitly not capacity) and $270M of inventory purchase commitments (2.4% of revenue). This is genuinely clean — and it is the sharpest contrast with NXPI, which hides ~$583M/yr of fab-JV equity in investing activities and carries a $14.1B / 37-year take-or-pay.

⚠️ Watch: depreciation is rising 12%/yr as the build-out enters service. If capex/dep drops below 1.0× for two consecutive years, revisit.


5. 📌 Dividend Grower Overlay (§4)

This is where ADI is strongest, and it is the single biggest differentiator versus NXPI.

FY 2016 2018 2020 2022 2023 2024 2025 FY26 run-rate
DPS $1.66 $1.89 $2.40 $2.97 $3.34 $3.62 $3.89 $4.40
Dividend CAGR Rate
5yr (FY2020→25) +10.14%
8yr (FY2017→25) +10.34%
10yr (FY2015→25) +9.50%

The freeze test that caught NXPI — ADI passes emphatically

Quarter DPS Event
FY23 Q2 (Feb '23) $0.86 🔼 +13.2%
FY24 Q2 (Feb '24) $0.92 🔼 +7.0%
FY25 Q2 (Feb '25) $0.99 🔼 +7.6%
FY26 Q2 (Feb '26) $1.10 🔼 +11.1% — most recent

ADI raises like clockwork every fiscal Q2 (February declaration), with no missed cycle in the observable record — the exact opposite of NXPI's ~10-quarter freeze.

And the raise rate is re-accelerating, not decelerating. The 2024–25 slowdown to ~7% coincided exactly with the FY2024 revenue collapse (−23.4%); management protected the streak with a smaller raise rather than freezing it, then restored a double-digit raise the moment the cycle turned. The +11.1% is above the 10-year average of +9.5%. That is a management confidence signal.

⚠️ A trap worth naming: the Nov-2025 declaration was still $0.99. Anyone snapshotting ADI in December would have seen four consecutive quarters at $0.99 and wrongly inferred a freeze. The cadence, not the snapshot, is the signal — the same error that would have been made in reverse on NXPI.

Streak: 22 consecutive years of increases, verifiable in XBRL back to FY2007. >$32B returned since 2004.

Coverage — stable to improving

FY Dividends FCF Div/FCF Div/GAAP EPS
2018 $703M $2,188M 32.2% 47.0%
2022 $1,545M $3,776M 40.9% 56.5%
2024 (trough) $1,796M $3,122M 57.5% 110.4% ⚠️
2025 $1,924M $4,279M 45.0% 85.2%
TTM $1,998M $4,565M 43.8% 65.4%

10-year FCF payout range 32–58%, averaging ~46%, with no secular deterioration. The FY2024 spike is the stress test: at the worst point of the worst downcycle in a decade, the dividend still consumed only 58% of free cash flow — even while exceeding GAAP EPS, because $1.6B of non-cash amortization makes GAAP EPS a badly misleading denominator here. Covered 2.1× by FCF/share.

⚠️ One watch item: recent raises (~11%) run ahead of organic FCF/share growth (~6%). At a 44% payout there is ample room, but the payout drifts up ~5pp per decade if that persists.

Current yield 1.20% (at $366.30) vs a 5yr average of 1.65%27% below its own history. Excellent dividend, expensive price. See §7.


6. Moat — 4.0 / 5

Confirmed under adversarial stress-testing, with one genuine weak spot.

Source Rating Justification
Intangibles (design IP, process know-how) 4.5/5 The real moat. Tacit precision-analog circuit design — layout parasitics, thermal gradients, matching, noise floors — is apprenticeship knowledge learned over 10–20 year careers. A 20-bit ADC is not a bigger 12-bit ADC. State capital cannot buy this directly; RISC-V is irrelevant to it.
Switching costs 4/5 Real in industrial/auto/medical/defense: requalification cost, 10–20yr designs, regulatory re-cert. Docked because 56% of revenue goes through distributors (intermediated relationships) and switching cost is ~1/5 in datacenter — the fastest-growing stream.
Efficient scale 3.5/5 75,000 SKUs is a genuine barrier no entrant will replicate. But TI, Infineon, ST, Renesas, Microchip and six Chinese firms all fit in this market — that's breadth, not efficient scale.
Cost advantage 2/5 The honest weak spot. ADI sources >50% of wafers externally (TSMC) and runs 200mm internally; TI is moving to >95% internal / >80% 300mm at ~40% lower die cost. ADI has a price premium, not a cost advantage — its opposite. (But see §2/Q2: the downcycle test says this matters less than the theory predicts.)
Network effects 1/5 Essentially absent. The software layer (CodeFusion Studio — a free VS Code fork wrapping open-source toolchains; ADI Assure) has no disclosed adoption metrics, no monetization, and competes with TI's Code Composer Studio, which has existed since 1999 and has never prevented a socket loss. Do not credit this as a moat until a number exists.

Revenue streams and evergreen ratings

Stream Evergreen Why
Precision converters / signal chain (high end) 5/5 The physical world stays analog. Forever business.
Aerospace & defense / RF 5/5 The most durable stream ADI owns — multi-decade qualification, ITAR/EAR-gated, US-domestic-preference, near-zero Chinese substitution risk. ~1,000 mil-spec parts just repriced up to +30% with zero customer flight risk.
Medical / healthcare 4.5/5 FDA design-change burden is a permanent switching tax.
MEMS / iCoupler isolators 4/5 Specialty proprietary processes. Underrated.
Factory automation / broad industrial 3.5/5 Evergreen demand, contested supply — the front line of Chinese import substitution.
Instrumentation & test (ATE) 3.5/5 Secular right now because AI chip-test intensity is exploding — but that is a derivative of the same AI capex cycle as datacenter. Investors are double-counting it as "industrial diversification."
Automotive 3.5/5 Content growth real; socket-shaped; TI/Infineon-contested. And it is +2% yoy — flat.
Power management (general) 3/5 Most commoditizable analog category; where TI's cost pressure lands.
Datacenter power / interconnect 2/5 Not evergreen: 12–18mo refresh, ~6 hyperscaler customers, cost-decided, no legacy tail.
Consumer 2/5 Cyclical, low-margin, low-loyalty.

Overall evergreen: 4/5 — but the forward-looking rating is lower than the backward-looking one, and that gap is the thesis risk. ADI's core is one of the most genuinely evergreen businesses in technology. Its growth is coming from its least evergreen segments.

Comparative: ADI ≈ TXN (4.0) > NXPI (3.5)

  • ADI over NXPI, decisively. ADI's barrier is diversification-plus-tacit-design — ~75,000 SKUs, roughly half of revenue from products >10 years old, no single design loss matters, and architectural change does not consolidate it away because signal chains persist across every topology. NXPI's is a portfolio of specific sockets protected by qualification calendars, facing structural socket-count deflation from zonal automotive architectures. This distinction survived stress-testing and is the strongest plank of the ADI case.
  • ADI vs TXN is closer than the gross-margin gap implies, and the downcycle scored it for ADI (§2/Q2). ADI has the better product moat; TI the better cost moat. They fail in opposite conditions — ADI's is strongest in shortages, TI's in gluts. The 2023–25 glut said product won. The 2028–30 glut, with TI's capex finally falling, is the rematch.

China — a real exposure the Phase-1 work under-weighted

China is 26% of revenue ($2.86B FY2025, +34% yoy) — higher than NXPI's 17%. Chinese analog vendors (SG Micro, 3Peak, Silergy at ~$550M revenue, Southchip, Joulwatt, Novosense) are established in power management and signal chain.

The comforting story — "precision analog's barrier is tacit talent, which capital can't buy" — is true and load-bearing on the wrong assumption. It holds the top of the stack for 10+ years. It does not hold the middle, and the middle is where the volume is. Three things break it: tacit knowledge is person-transferable and Chinese firms have been hiring Western analog designers for a decade; losing the bottom raises the cost of serving the top by removing the volume that amortizes fixed costs; and a tariff can buy in one year what design talent takes fifteen to earn.

The moat is not eroding; it is narrowing. ADI's blended 67% gross margin is a weighted average across precision tiers, and the weights are moving the wrong way.


7. Valuation

Model weights

Model Output Weight Rationale
Reverse-DCF / EV-FCF $175–275 30% Only clean measuring stick; book value is unusable
📌 Dividend Yield Theory $244–284 20% DYT's assumptions genuinely hold here — 22yr streak, ~10% CAGR, stable 44% coverage
Mid-cycle multiples $230–294 15% Corrects the peak-earnings error
📌 DDM (two-stage) $78–149 15% Deserves weight (unlike NXPI) but ignores the 56% of FCF not paid out — a floor-finder
Bogle expected return ~6.6%/yr 15% Multiple-change term dominates → low conviction
Graham IV $102–114 5% Out of domain — tangible BVPS is −$2.35

⭐ Reverse-DCF — the decisive calculation

What 10-year FCF growth rate does the price require (3% terminal)?

Base FCF r = 8% r = 9% r = 10%
TTM $4,565M 10.74% 13.38% 15.79%
WC-normalized $4,100M 12.11% 14.80% 17.24%
Mid-cycle $3,800M 13.08% 15.80% 18.27%

Versus demonstrated:

Benchmark Rate Requirement is…
Organic FY2022→FY2025 +4.25% 3.1× this
TTM vs FY2022 peak (fairest organic) +6.5% 2.1× this
Cycle-neutral 7yr +14.3% ≈ matches — but embeds Linear + Maxim

The conclusion is stark: $353–366 requires ADI to reproduce, organically and for a decade, the growth it previously achieved by spending $30B+ of stock on two acquisitions — while invested capital shrinks, the tax rate climbs from 6.9% to 16.4%, and reported ROIC sits below cost of capital. A company earning below its cost of capital on reported invested capital cannot compound at 13% by reinvesting; it can only do so by buying growth again. Empower at $1.5B is 0.9% of EV — a rounding error against that requirement.

Value at demonstrated growth rates ($/share):

Base FCF r g=4.25% g=6.5% g=10% g=14.3%
$4,565M 9% $176 $209 $273 $379
$4,100M 9% $158 $188 $245 $340
$4,565M 10% $150 $177 $230 $317

Only the M&A-inflated +14.3% justifies anything near today's price, and only at the lower discount rate.

Multiples

Metric TTM WC-normalized
P/E GAAP 54.6×
P/E adjusted (TTM EPS ~$8.40) 43.6×
P/E adjusted (NTM est.) 26–29×
EV/FCF ~39× ~44×
FCF yield on EV 2.5% 2.3%
EV/Revenue 13.92×
EV/EBITDA 28.9×

A 2.3–2.5% FCF yield on EV against a 10-year Treasury near 4.3% is a negative real cash yield versus the risk-free rate.

Note: fin.py reports EV of $177.4B — below the $178.4B market cap, implying net cash. That reflects ADI's short-term investment portfolio being counted as cash. On a gross-debt-less-cash basis EV is ~$184.6B. The multiples above use the conservative (higher) figure where it matters.

⭐ Adjudicating GAAP vs adjusted — the whole valuation debate

Ruling: neither number is right; use cash flow, and where an EPS is unavoidable, sit nearer GAAP than the Street does.

The bull case for adjusting is legitimate in principle — $1,592M of FY2025 acquisition-intangible amortization (14.4% of revenue) is non-cash and does not require replacement capex; valuing ADI at 54× trailing GAAP would systematically mis-value any acquisitive company.

But the bear case is stronger here, for a reason specific to ADI: the amortization is real cash, spent at prices that halved ex-goodwill ROIC from 36–39% to 18.9%. Adding it back tells you what ADI would have earned if Maxim had been free. It wasn't.

The decisive point: you cannot simultaneously claim the M&A-era growth rate (14.3%) in the DCF and add back the cost of achieving it in the EPS. Bulls do exactly that. That double-count is the entire 43× vs 54× argument.

EV/FCF of ~39× (44× normalized) is the honest headline multiple — it sits between the two camps and closer to the bears.

📌 Dividend Yield Theory — high weight here

Yield Implied FV on $4.40
1.20% (today) $367
1.40% $314
1.65% (5yr avg) $267
1.80% (7yr realized avg) $244
2.10% (trough) $210

DYT fair value $244–284. Today's 1.20% yield sits 27% below the 5yr average and near the richest end of ADI's entire dividend history — after a 18% drawdown. That is the most uncomfortable single fact for the bull case.

The honest counter: with the dividend compounding ~10%, waiting for a 1.80% yield may mean waiting for a price that never comes — the dividend can grow into the price. That is why DYT gets 20% weight, not 35%.

⚠️ Cycle position — peak-adjacent on three axes simultaneously

Indicator Reading
TTM revenue $12,740M vs FY2022 peak $12,014M +6.0% above prior peak
Q2 annualized (~$14.5B) vs prior peak +20.7% above
GAAP gross margin 67.3% record
Adjusted operating margin 49.0% management-named near-term ceiling
Capex 4.8%, capex/dep 1.31× normalized — no relief available

The growth is amplitude, not trend. Of the +$983M yoy gain, ~66% is industrial rebounding +56% off a three-year trough; datacenter is ~20%; automotive is +2% yoy and flat at 24% of revenue. ~2pts of 2026 growth is price, not volume (management's own statement), and Q2 contained a disclosed one-time channel-repricing benefit that reverses in Q3 (−50bp gross margin).

Downside case

Template: FY2024 (revenue −23.4%, FCF $3,122M at 33.1% margin).

Peak rev Trough rev FCF margin FCF/sh 25× 30× 35×
$14.7B $11.26B 30% $7.19 $180 $216 $252
$14.7B $11.26B 33% $7.91 $198 $237 $277
$15.5B $11.87B 33% $8.34 $208 $250 $292

Downside case ~$230 (band $210–250) — a ~37% drawdown from $366. Empirically corroborated by the 52-week low of $218.37, which sits inside the band.

But the floor is real. With two more raises DPS reaches ~$5.20 by FY2028; at a 1.9–2.3% trough yield that implies $226–274. The 22-year streak, 44% payout, 2.1× coverage and a genuinely non-issue balance sheet mean the dividend does not get cut in a normal downcycle. ADI is not a solvency story. It is purely a price story.

💰 Synthesized fair value: $255 – $305 · midpoint ~$280

Reconciling two independent builds: the Valuation Analyst landed at $225–285 (mid $255); the debate adjudicator, after correcting the falsified moat claims, landed at $270–320. The gap is the moat correction (pushes up) net of the buyback correction (pushes down). Splitting on the merits:

Zone Price Action
🔴 Trim / avoid > $320 Requires >11% perpetual organic FCF growth; DYT yield <1.38%
⚠️ Current: $366.30 ~31% above midpoint No action. Do not initiate.
🟡 Fair $255 – $305 Pays for demonstrated organic growth plus a genuine quality premium
🟢 Accumulate < $290 DYT yield ≥1.52%; reverse-DCF requirement falls under ~10%
🟢🟢 Strong entry < $255 Below trough-FCF fair value; reverse-DCF requirement ~9%
Downside case ~$230 (band $210–250) Where a normal downcycle prints

Three assumptions this valuation is most sensitive to

# Assumption If wrong $/share
1 Long-run organic FCF/share growth ~6.5–7.0% Each +1pt of sustainable growth ≈ +$14 to +$17. To reach $366 needs ~13.4% — ≈ +$105. At the demonstrated 4.25%, fair value falls to $158–176. This is the whole thesis.
2 Cost of equity 9–10% (beta 1.19, rf 4.3%, ERP 4.75%) Each 100bp move at g=6.5% ≈ ∓$29. A bull using 7.5% adds ~$60 and closes most of the gap — the bull case is largely a discount-rate argument in disguise.
3 Normalized FCF base ~$4,100M (TTM less the WC flatter) Each $500M ≈ ∓$23. If TTM FCF is clean and repeatable, fair value rises ~$21.

Deliberately not in the range: the cyber matters and the MOFCOM ruling. Per §2/Q5 both are smaller than first assessed, but the July 26 matter is days old. I do not haircut the range for them — I widen the margin of safety demanded, which is why the accumulate zone sits at $290 rather than at the $305 top of fair.


8. Tensions Surfaced Honestly

1. Two Phase-1 bear claims were broken by primary sources, and I have said so plainly. The "margins never recovered" argument is a purchase-accounting artifact (FY2018 GAAP GM of 68.3% with Linear amortization running is 320bps above standalone FY2016). The "weakest in gluts" prediction was falsified by the 2023–25 downcycle: ADI's gross margin degraded 250bps and FCF grew 13.3%, while TI's degraded 1,180bps and FCF fell 56.1%. The Moat Analyst's diagnosis was excellent and the verdict was built on two broken measurements.

2. One Phase-1 bull claim was overstated by ~2×. The buyback is real (unlike NXPI's) but the per-share tailwind is −0.92%/yr, not 1.7pp, and the headline Q2 figure includes $118M of tax withholding. Share count was flat for a full year.

3. The Sentiment Analyst was wrong that no scorecard exists — ADI's 2022 Investor Day model exists and ADI missed 4 of 5 targets, including revenue CAGR by ten points. It was also right, and understated, on the CEO's discretionary $24.4M block; and wrong on Golz, whose sale was plan-driven.

4. Both cyber and MOFCOM were over-weighted at Phase 1. Item 8.01 not 1.05; PII not design data; ADI delisted from the leak site; the precedent base rate for semiconductor IP breaches is near-zero realized impact. MOFCOM has issued no preliminary determination at T+10.5 months, ADI has never named it in an SEC filing, and >50% of its wafers are non-US-origin. TI's exposure is several multiples of ADI's.

5. The genuinely unresolved question — datacenter. ~$280M of the ~$1.66B is the contested vertical-power socket; the rest is Linear-heritage protection/telemetry that behaves like industrial. ADI has named zero datacenter customers across four consecutive calls and just paid $1.5B cash for a pre-production IVR product to fill a gap its CEO named out loud. Nothing available decides whether this is capability-building or empire-building. The observable that would: a named vertical-power or IVR design win at NVIDIA or a hyperscaler. One named socket by the FY2026 10-K (~late Nov 2026) resolves it bullish; continued silence through 2027 resolves it bearish.

6. The ADI-vs-TXN question is closer than either side claims. They trade at near-parity EV/sales (13.92× vs 13.10×) despite TXN earning roughly double the ROE and ROIC on a quarter of the goodwill. But TI's FCF fell 56% through the downcycle while ADI's grew 13%. You are buying TI's advantage at a discount to its steady state and ADI's advantage at a premium to its steady state. At parity, that is a genuine coin-flip, and I would not pay up for either at current prices.


9. What Would Most Change This Verdict

1. ⭐ The Q3 FY2026 print — 2026-08-19, 7:00am ET (confirmed). The highest-information event of the year; three tests land at once. - The drawdown-buyback test, finally runnable. Q2 ended 2026-05-02, before the decline, so nobody yet knows whether ADI defended the stock. >$1.0B of program repurchase (excluding tax withholding) at sub-$360 would materially upgrade the capital-allocation read and lift the buyback assumption back toward 1.5%/yr. A flat ~$650M confirms ADI buys mechanically regardless of price — which at a 2.5% FCF yield is value-neutral at best. - Any sub-segment gross-margin disclosure for datacenter, which would settle §2/Q3 definitively. - Escalation of either cyber matter to Item 1.05, or a quantified expense line.

2. 🚩 Evidence on what the $3.0B revolver is for. If ADI announces another multi-billion platform acquisition, the serial-acquirer thesis wins outright, the multiple should compress toward TI's, and fair value drops toward $250. If it is drawn to fund buybacks at 27× forward earnings, that is a different kind of bad. If it expires undrawn in July 2027, it was prudence and this goes to zero. Priority is high precisely because the base case is "nothing" and the alternative is thesis-breaking.

3. A named vertical-power or IVR design win — or its continued absence. See §8/item 5. This determines whether the $1.5B bought an option or a position, and whether the real franchise is the protection/telemetry two-thirds — which is worth owning but is not an AI growth story.

Runner-up, datable: MOFCOM by 2026-09-13. Expect a non-event for ADI and a real event for TI. Silence through August raises the extension probability.


A portfolio-specific passage was removed from the public build.

11. Watchlist Recommendation

Add to Watchlist.md — 🏛 Evergreen Compounders, out-of-zone.

Field Value
Sleeve 🏛 Evergreen Compounders — it passes the sleeve test ("would I hold this through a 40% drawdown without re-underwriting?"). Yes — the moat survived adversarial stress-testing, the dividend survived the worst downcycle in a decade at 58% FCF payout, and the balance sheet is a non-issue. This is not a re-rating play; it is a quality name at the wrong price.
Conviction [7] — business 8, entry 3. Higher than NXPI's [5] because the quality is less ambiguous.
Entry zone < $290 (strong entry < $255)
Currently $366.30 — ~31% above fair-value midpoint, out of zone
Trim zone > $320
Thesis Highest-quality precision-analog franchise: ~75,000 SKUs, half of revenue from products >10yrs old, tacit design barrier that architectural change cannot consolidate away, ITAR-gated A&D core. Won the 2023–25 downcycle against the cost leader — gross margin degraded 250bps vs TI's 1,180bps; FCF +13.3% vs TI's −56.1%. 22-year dividend streak with a re-accelerating +11.1% raise at a 44% FCF payout. Fortress balance sheet, no hidden capacity commitments. Priced at ~13.4% required FCF growth against 4.25–6.5% demonstrated organic. Buy the business when the price stops requiring the acquisition-era growth rate without the acquisitions.
Break trigger A multi-billion platform acquisition drawn on the new $3.0B revolver; OR adjusted gross margin below 68% for two consecutive quarters outside a downcycle; OR the dividend raise missing its February cadence; OR datacenter still with zero named customers by the FY2027 10-K
Upgrade trigger Q3 (2026-08-19) showing >$1.0B of program buyback at sub-$360; OR a named NVIDIA/hyperscaler vertical-power design win; OR a re-published Investor Day model with a dated revenue commitment
Concentration note 26% China revenue (higher than NXPI's 17%), 50% industrial, 24% auto. Correlates with existing semiconductor and industrial-cyclical exposure. Beta 1.19 understates realized drawdown sensitivity.

Do not initiate at $366.30. The framework's conservative bias and the arithmetic point the same way.


12. Data Quality & Limitations

Issue Impact Resolution
Price moved mid-analysis: $353.37 → $366.30 Every distance-to-fair-value figure is ~3.7% worse than modelled Flagged at the top; zones stated at live price
roic.ai gated to a Free 2-year plan No vendor-computed ratio series Rebuilt from SEC XBRL companyfacts CIK 0000006281 (19 fiscal years) + FY2016/FY2025 10-Ks. All ratios are own computations with definitions stated
Moat Analyst's "GM below pre-acquisition level" Would have been a thesis-breaking claim Overturned. Purchase-accounting artifact — $867.6M in FY2025 COGS = 787bps. FY2018 GAAP GM 68.3% with Linear amortization proves the point without non-GAAP charity
Moat Analyst's "weakest in gluts" Central forward prediction Falsified by the 2023–25 experiment. TI GM −1,180bps and FCF −56.1%; ADI −250bps and +13.3%
Sentiment Analyst's "no Investor Day scorecard exists" Would have removed the most predictive artifact Wrong — the 2022-04-05 model exists; ADI missed 4 of 5
Sentiment Analyst's read of the Golz sale Cited as discretionary selling into the drawdown Wrong — 10b5-1 checkbox set, plan adopted 2026-03-12. Discount entirely
Fundamentals Analyst's buyback figures Overstated by ~2× on the per-share effect Q2 $773M includes $118M tax withholding → real ~$652M. Share count −0.92%/yr, not −1.7%/yr
Circulating "$3.09B Q3 guidance" Would imply a sequential decline Wrong — that is Q2 adjusted EPS mis-transcribed. Actual guide is $3.9B ±$100M, verified against the primary PR Newswire release
fin.py EV of $177.4B < market cap Implies net cash Reflects short-term investments counted as cash; gross-debt-less-cash EV is ~$184.6B. Both stated
fin.py "RevGrowth 0.37 / EarnGrowth 1.10" Quarterly, not annual FY2025 annual revenue growth was +16.9%
ADI's ~50%-of-revenue-from->10yr-old-products claim Load-bearing for the catalog moat Not found in ADI primary sources — traces to third-party analysis. Directionally credible, not company-confirmed
"~100,000 customers" Cited widely Not in the FY2025 10-K. The 10-K says "several thousand analog ICs, many of which can have several hundred end customers." 75,000 SKUs is confirmed
Ex-goodwill ROIC of 18.9–23.6% Central to the "operating business is excellent" claim Own computation and a third-party series that could not be reconstructed from filings. Definitions stated; treat as approximate
Organic revenue CAGR ~4.4% Key growth input Own estimate using approximate acquired-revenue baselines. Not an ADI disclosure
Datacenter revenue ≈$416M/quarter Key valuation input ADI has never stated a dollar figure. Derived from "more than 75% of communications revenue." No 2026/2027 target exists. No customer named on four consecutive calls
TI Analog segment revenue Used in the Q2 downcycle comparison Sourced via search synthesis of TI 10-Ks, not a primary fetch — slightly lower confidence than the XBRL totals
Institutional 13F data as of 2026-03-31 Four months stale; predates Q2 earnings, Empower, the drawdown, and both cyber incidents The two Vanguard "+100%" entries are a filer re-classification artifact, not accumulation
Buyback behavior during the drawdown Unknown and highly informative Q2 ended 2026-05-02, before the decline. The Q3 print (2026-08-19) is the key datapoint
Remaining buyback authorization ~$8.5B Derived from the declining authorization balance; the ~$11.5B figure circulating is from a February 2025 release
July 26 cyber matter Genuinely unassessed Three days old; ADI is "assessing its validity." The only part of the cyber story not dismissible
CodeFusion Studio / ADI Assure adoption Rated 1.5/5 as a moat source No disclosed metrics of any kind exist
Gregory M. Bryant's departure Inferred from Form 4 absence after Nov 2024 Unverified

Generated by Financebot · framework: analysis_notes.md §0–§5 + §4 Dividend Grower overlay · agents: Fundamentals, Sentiment, Moat, Valuation, Debate adjudicator · primary sources: SEC XBRL companyfacts (CIK 0000006281, 19 fiscal years; TXN CIK 0000097476), ADI FY2016 & FY2025 10-Ks, Q2 FY2026 10-Q, 8-Ks of 2026-05-19 / 2026-07-02 / 2026-07-29, 154 Form 4 filings parsed from raw XML, ADI 2022 Investor Day model, Q1/Q2 FY2026 earnings transcripts, Yahoo Finance MCP, .mcp/fin.py

Note on process: the debate round overturned two of the three Phase-1 bear claims and one Phase-1 bull claim using primary sources. The net effect on fair value was roughly offsetting (moat correction up, buyback correction down), but the effect on confidence was large — ADI's business quality is now the least ambiguous finding in the file, and its price is the most clearly problematic.