DIOD › analyze
DIOD · Analyze
A portfolio-specific passage was removed from the public build.
Verdict: WATCH — conviction 5.5/10. Good business, wrong price. The cyclical recovery is real, better-evidenced than I expected, and the China bear case largely failed on the evidence. But the recovery has already happened — Q3 guidance is 97.8% of the all-time record quarter — and at $100.41 the market is paying a growth multiple for a thin-moat cyclical at what may be a peak inflated by a competitor's government seizure. Add to the watchlist. Do not buy here.
0. Knowledge check
python .mcp/kb.py find DIOD → no matches. Fresh coverage. Four existing notes apply:
| Note | How it applies |
|---|---|
| [[pitfall-multiple-trim-inverts-on-peak-cycle-cyclicals]] | Fires directly. DIOD must NOT get a Trim NNx fwd. See §7. |
| [[pattern-net-margin-above-operating-margin-is-a-tripwire]] | Fires on Q2 2026 — GAAP EPS $1.00 exceeds non-GAAP EPS $0.70. |
| [[pitfall-yahoo-insider-purchases-counts-rsu-grants]] | Fires, and reconciles exactly. See §3. |
| [[principle-down-a-lot-is-not-cheap]] | Inverted here — up a lot is not expensive either, but §4 concludes it is. |
1. Fundamentals
1a. 🚩 Three corrections to the framing before any analysis
| Common framing | Reality (primary source) |
|---|---|
| "FY2025 revenue $1.48B, down from $2.00B in 2022 — a company in decline" | FY2024 was the trough at $1.311B. FY2025 already grew +13.0%. 2026 is a boom: Q1 $405.5M (+22%), Q2 $445.5M (+21.7%), Q3 guided ~$510M (+30% YoY). |
| "R&D rising while revenue falls" | True through 2025. In 2026 R&D is flat — H1 2026 $81.3M vs $79.2M (+2.7%) while revenue grew +21.9%. The step-up is now converting to leverage. |
| "Margin collapse driven by fab underutilization" | The 10-K does not say that. DIOD's own attribution for the FY2025 decline is "product mix and slower growth in the industrial end market… average unit cost increased 1.3%… lower margin/lower cost products, as well as raw material price increases, including gold." Mix, industrial weakness, and gold — not utilization. With ~50% of wafers outsourced, utilization is structurally a weaker lever here than at a pure IDM. This changes how the recovery should be modeled. |
The consequence: the premise "depressed earnings are a trough to be bought" is obsolete. They are not depressed. The live question is how much of this peak is real.
1b. Health scorecard (analysis_notes §1)
| Metric | FY2022 | FY2023 | FY2024 | FY2025 | Q2 2026 | Read |
|---|---|---|---|---|---|---|
| Revenue | $2.00B | $1.66B | $1.31B (trough) | $1.48B (+13.0%) | $445.5M (+21.7%) | 🟢 recovering |
| Gross margin | 41.3% | 39.6% | 33.2% | 31.2% | 33.1% | 🟡 recovering |
| Operating income | $404.5M | $250.1M | $51.4M | $36.4M | — | 🔴 −91% peak-to-trough |
| Operating margin | 20.2% | 15.1% | 3.9% | 2.5% | ~7.5% | 🔴→🟡 |
| Net income | $331.3M | $227.2M | $44.0M | $66.1M | $46.6M | 🟡 |
| FCF | $180.8M | $130.2M | $46.4M | $137.2M | $34.8M | 🟡 |
| ROIC | 19.0% | 11.1% | 2.1% | 1.5% | ~7.3% (2026E) | 🔴 §2 |
| Cash | $336.7M | $315.5M | $308.7M | $367.2M | ~$442M | 🟢 |
| Debt | $213.7M | $98.4M | $91.7M | $95.6M | ~$40M | 🟢 |
| Debt/Assets | 9.3% | 4.2% | 3.8% | 3.90% | ~1.6% | 🟢 fortress |
| Diluted shares | 46.04M | 46.31M | 46.41M | 46.41M | 46.4M | 🟢 +0.3%/yr |
| Revenue/share | $43.46 | $35.89 | $28.25 | $31.93 | — | 🟡 |
| FCF/share | $3.93 | $2.81 | $1.00 | $2.95 | — | 🟡 |
The balance sheet is the least ambiguous thing here: ~$442M cash against ~$40M debt, net cash ~$402M ($8.70/share), current ratio 3.17. Bankruptcy is unthinkable through any cycle. DIOD survived a 91% operating-income collapse while increasing R&D 28% and running $78M/yr of capex. That is durable-franchise balance-sheet behavior.
⚠️ Yahoo's EV of $3.69B implies ~$920M of net cash, against the 10-Q's ~$402M. Yahoo appears to be sweeping in long-term investments. I use EV ≈ $4.21B. Flagged for the KB.
1c. 🚩 The Q2 GAAP/non-GAAP inversion
A portfolio-specific passage was removed from the public build.
GAAP EPS exceeds non-GAAP EPS — the [[pattern-net-margin-above-operating-margin-is-a-tripwire]] signature. GAAP EPS was essentially flat YoY ($1.00 vs $0.99) while non-GAAP more than doubled ($0.70 vs $0.32). The non-GAAP line is the honest operating number here, which is the opposite of the usual direction.
Two useful sub-findings: there is no separate non-GAAP gross margin — all adjustments sit in opex, so GAAP GM = non-GAAP GM = 33.1%, and any source quoting a different "adjusted gross margin" is fabricating it. And the trailing P/E of 54.3x on TTM GAAP EPS of $1.85 straddles trough quarters and this gain; it is not usable.
1d. Capital allocation
| FY2025 use of cash | Amount |
|---|---|
| Capex | $78.4M (5.3% of revenue; target band 5–9%) |
| Buybacks | $33.8M (first ever) |
| Acquisitions | $5.3M |
| Dividends | $0 |
| Net debt change | +$1.2M |
Q2 2026 capex ran 7.5% of revenue — the upper half of the band, funding the 6-inch→8-inch migration and selective back-end capacity. No total capex program size or multi-year capex guidance was given. Flagged as unquantified.
Pending: ElevATE Semiconductor — signed 2026-07-10, $250M cash base + up to $50M earnout, fabless ATE (automated test equipment) ICs, San Diego. ~$50M first-12-month revenue,
20% CAGR claimed, gross margin "significantly higher than corporate average," immediately accretive. Closing 2H 2026, not in Q3 guidance. At ~5x forward revenue that is a full but defensible price for a fabless above-corporate-margin asset — it consumes over half the cash balance. Strategically it is a better-hedged AI play than DIOD's direct content story: more AI silicon produced means more test equipment needed, regardless of which chipmaker wins.
Fundamentals verdict: 🟡 Recovering strongly from a genuine trough, on a fortress balance sheet — but through-cycle returns on capital are poor (§2).
2. Moat & competitive position
Aggregate: NARROW-to-NONE, 2/5.
DIOD is a broad-line catalog supplier of low-cost, high-volume analog, discrete, logic and mixed-signal parts, sold 65% through distributors. 28,000 SKUs, 45 billion units shipped in 2025 — a blended ASP of $0.033 per part. Its own 10-K concedes competitors "have greater financial, marketing, distribution, brand name recognition, research and development, manufacturing, and other resources than we do," and the defense it offers — "product focus, packaging expertise and our flexibility and quick adaptability" — is a list of service attributes, not a moat.
| Moat source | Rating | Assessment |
|---|---|---|
| Network effects | 0/5 | No mechanism exists. |
| Efficient scale | 1/5 | Fragmented — top five ≈ 45% of a ~$35B global discrete market. DIOD is a ~4–5% player. |
| Cost advantage | 2/5 | 150mm/200mm legacy fabs are subscale against TI's 300mm. Owned capacity is a liability at low utilization and a benefit at high — see the incremental-margin table below. Real China-based low-cost manufacturing and 75% internal back-end are worth something. Not a cost leader, not a laggard. |
| Switching costs | 2.5/5 ↑ | Genuine on the ~35% automotive/industrial book — AEC-Q100/Q101, IATF 16949, 2–4 year design-in, 5–8 year program life. Effectively zero on the ~56% computing/consumer/comms catalog book. Currently deepening on shortage-driven requalification. |
| Intangibles | 2/5 | Brand is worth little in catalog parts. But the Pericom-legacy timing/signal-integrity portfolio — PCIe 6.0/7.0 mux-buffers, eUSB repeaters, ultra-low-jitter clocks — is genuinely differentiated and is winning AI-server sockets. This is the ~10 margin points DIOD holds over Vishay, its closest structural analogue. |
The returns-on-capital verdict is unambiguous
| 2022 | 2023 | 2024 | 2025 | 2026E | Mgmt's own $2B target | |
|---|---|---|---|---|---|---|
| ROIC | 19.0% | 11.1% | 2.2% | 1.5% | ~7.3% | ~9.4% |
A business whose management's own three-year target implies ~9.4% ROIC earns approximately its cost of capital at mid-cycle. That is the arithmetic definition of a thin moat. The 19% of 2022 was a shortage; the 1.5% of 2025 was a trough; neither is the truth.
The fixed-cost operating leverage, measured in both directions
| Period | Δ Revenue | Δ Gross profit | Incremental gross margin |
|---|---|---|---|
| FY2024 → FY2025 | +$171.0M | +$26.6M | 15.5% ← the price/mix give-back year |
| Q2 2025 → Q2 2026 | +$79.3M | +$32.2M | 40.6% |
| Q1 → Q2 2026 (seq) | +$40.0M | +$18.8M | 46.9% |
| Q2 → Q3 2026 (guided, seq) | +$64.5M | +$30.9M | 48.0% |
Same fabs, same cost base, opposite outcomes. Downside leverage is equally brutal and the risk factors state the mechanism plainly: "the costs associated with this excess capacity are expensed immediately and not capitalized into inventory." Peak-to-trough operating leverage of roughly 3.5x.
🔴 The China bear case — tested, and it substantially FAILED
This was expected to be the central bear case. The evidence contradicts it.
DIOD's China revenue is at a record, not in decline:
| Ship-to-China revenue | 2023 | 2024 | 2025 | H1 2025 | H1 2026 |
|---|---|---|---|---|---|
| $M | 704.8 | 589.5 | 660.3 | 317.8 | 396.1 (+24.6%) |
| Q2 alone | — | — | — | 164.4 | 219.1 (+33.3%) |
China was DIOD's fastest-growing geography in Q2 2026 at +33% YoY, running above the 2023 level. A company being displaced by state-backed domestic substitution in its largest market does not post its fastest growth there.
The regional data locates the real weakness — Europe, not China. Europe fell −35.6% from 2023 to 2025 and was the only region still declining in 2025. The structural hole was European automotive and industrial demand.
And DIOD is itself substantially a Chinese manufacturer — the point most China-threat analyses of DIOD miss. Wafer fabs in Shanghai and Wuxi; assembly/test in Shanghai, Chengdu, Wuxi; 1.65M sq ft in Shanghai and 581K in Wuxi; three plants hold HNTE status (15% preferential tax); 50-year land rights to 2056–2065. A "buy domestic" mandate that displaces TI (fabs in Texas and Utah) does not cleanly displace parts fabbed in Shanghai and packaged in Chengdu.
Direct competitor research confirms the threat is narrower than assumed. Of the four main Chinese discrete players — Yangjie, NCE Power, Silan, CR Micro — only Yangjie has meaningful Western exposure (23% overseas FY2025, ~30% in Q1 2026, via its US-based MCC brand), and Yangjie was listed in the EU's 20th Russia sanctions package on 2026-04-23 with only a 9-month exemption running to ~March 2027. Silan is 96.2% domestic; CR Micro is a central SOE earning 2.91% ROE; NCE is a fifth of DIOD's size with gross margin down 7.8 points. Silan and CR Micro threaten DIOD's China revenue, not its global franchise — and they are barely profitable doing it.
Most importantly, the Chinese price war has stopped and reversed. All four raised prices twice in 2026 (+10–20%), citing wafer, packaging and precious-metal cost inflation. This mechanically narrows the price gap DIOD faces. Management confirmed the environment on the call — asked directly about pricing pressure, SVP Emily Yang: "In Q1, what we've seen pricing really, really stabilized, and it's mainly driven by the product mix change."
Where the China bear case survives: the 13-billion-unit volume loss since 2021 (58B → 45B, −22%) in low-ASP consumer/computing parts. Some was deliberate exit; some was taken by Chinese commodity houses. It is not coming back. The communications segment — smartphone-heavy, China-heavy — is the only segment still shrinking (−3% YoY in Q2 2026).
🎯 Cyclical vs structural — the decisive question, decided
Verdict: predominantly CYCLICAL, ~80/20. The decisive evidence is a dataset most analysis never reaches: DIOD discloses its own weighted-average selling price change every year in MD&A.
| Year | ASP change | Volume change | Units (B) | Revenue ($M) | ASP index (2021=100) |
|---|---|---|---|---|---|
| 2021 | — | — | 58 | 1,805 | 100.0 |
| 2022 | +28.5% | −13.8% | 50 | 2,001 | 128.5 |
| 2023 | −2.7% | −16.0% | 42 | 1,662 | 125.0 |
| 2024 | −14.9% | −7.3% | 39 | 1,311 | 106.4 |
| 2025 | −1.7% | +15.0% | 45 | 1,482 | 104.6 |
(Chain-multiplying the disclosed ASP and volume changes reproduces reported revenue to within 1% every year — the series is internally consistent.)
DIOD's 2025 ASP is 4.6% ABOVE its 2021 level. The entire "commodity price erosion" narrative is the unwinding of a +28.5% shortage price spike. Prices never fell below the pre-shortage baseline. That is textbook cyclical give-back, not structural destruction.
The peer test confirms it. Across nine peers (onsemi, STMicro, ROHM, AOSL, Vishay, TI Analog, Infineon, Littelfuse), all nine declined — a share loser needs a market that did better than it did, and there wasn't one. DIOD's revenue fall was the deepest (−34.5%) but its gross-margin damage was mid-pack (−1,015bps), milder than onsemi (−1,588), STM (−1,400), ROHM (−2,150), AOSL (−1,141) and Vishay (−1,090). A company surrendering price to hold sockets bleeds margin faster than peers. DIOD bled slower. And DIOD is now at ~102% of its prior peak run-rate versus onsemi at 77%. Structural losers do not outrun the field in the up-leg.
What is genuinely structural (the ~20%): the 13B-unit volume loss; gross margin not yet back to the 35–37% pre-shortage baseline, with the gap being GFAB (Greenock) and SPFAB (South Portland) underutilization that management says persists into 2027–28; and discounting intensity still elevated (distributor allowances 18.3% of gross sales in 2025 vs 13.7% in 2023).
Adversarial stress-test
As TI (58% GM, 300mm fabs, ~40% die-cost advantage): I can undercut DIOD on any part I choose — but my marginal 300mm wafer earns far more on a $2 signal-chain part than on a $0.033 diode. I attack DIOD's analog IC growth vector and ignore its 28,000-SKU discrete base. Result: I cap DIOD's margin ambition; I do not take its business.
As onsemi: I am going the opposite direction — I sold DIOD the South Portland fab in 2022 and am divesting commodity discretes for SiC. I am vacating DIOD's space and I handed it capacity.
As a state-backed Chinese entrant: I have already taken what is easy to take — that is the 58B→45B unit decline. The next tranche is much harder: DIOD manufactures in Shanghai and Wuxi with HNTE status, and I cannot easily obtain AEC-Q100 qualification and Tier-1 design-in at Western automakers.
Synthesis: DIOD's commodity flank has already been breached and the damage is in the numbers. What remains is more defensible than what fell — but it is defended by friction, not by advantage.
Evergreen assessment
Will DIOD exist in 10 years? Yes, high confidence. Discretes and standard analog are the picks-and-shovels of every electronic system; net cash; 28,000 SKUs across thousands of customers.
Will it matter? Yes, as a mid-tier supplier — not a category shaper. Structural ceiling is a ~35–38% gross margin, ~10–14% operating margin, ~10–13% ROIC business. Management's $2.5B/40% GM ambition exceeds anything DIOD has held outside a global chip shortage and should be treated as aspiration.
3. Sentiment & the Q2 2026 print
The results and the guide
| Metric | Q2 2026 | QoQ | YoY | vs consensus |
|---|---|---|---|---|
| Revenue | $445.5M | +9.9% | +21.7% | +2.0% |
| Gross margin (GAAP = non-GAAP) | 33.1% | +130bps | +160bps | — |
| Non-GAAP EPS | $0.70 | +63% | +119% | +11% |
| Opex (GAAP) % revenue | 25.6% | −130bps | −330bps | — |
| Q3 2026 guidance | Implied | vs consensus | |
|---|---|---|---|
| Revenue | ~$510M ±3% | +30% YoY, +14% QoQ | +8.2% |
| GAAP gross margin | 35.0% ±1% | +190bps QoQ | — |
| Non-GAAP EPS | $1.05 ±$0.10 | 2.8x YoY | +30% |
The guide is the story, not the quarter. A 2% revenue beat produced a +17% move because the Q3 revenue guide is 8% above consensus and the EPS guide 30% above.
✅ The strongest datapoint in the print — channel inventory
Emily Yang (SVP WW Sales): "Our channel inventory decreased both in terms of dollars and weeks again this quarter with the weeks lower than our normal range of 11 weeks to 14 weeks." … asked directly about double-ordering: "we definitely don't see the double booking… we try to balance the ship through at this moment, but we're not there."
Channel inventory below the normal band, falling in both dollars and weeks, alongside record global POS (sell-through), means DIOD is under-shipping consumption. That is the opposite of a cycle top and it is the condition under which guidance is most likely to be beaten. "We're not there" is an admission they are still shipping less than end demand.
Supply is becoming the binding constraint, not demand — Yang referenced "pockets of areas [where] supply is a little bit constrained" and invoked a COVID-era analogy. That is a pricing-power signal.
End-market mix — the automotive record is real
| End market | Q2 2026 | FY2025 | FY2024 | FY2023 | YoY | Read |
|---|---|---|---|---|---|---|
| Computing | 28% | 27% | 25% | 23% | +33% | 🟢 AI servers, datacenter, storage |
| Industrial | 23% | 23% | 23% | 27% | +24% | 🟢 recovering |
| Automotive | 21% (record) | 19% | 19% | 19% | +37% | 🟢 breaks a 3-year plateau |
| Consumer | 17% | 18% | 19% | 18% | +17% | 🟡 |
| Communications | 11% | 13% | 14% | 13% | −3% | 🔴 only shrinking segment |
Automotive at 21% is a genuine record — pinned at exactly 19% for three consecutive years, now breaking with +37% YoY growth. It is mechanically margin-accretive. In dollars: 2021 ~$217M → Q2 2026 annualized ~$374M, +72%.
⚠️ The AI narrative outruns the disclosure
DIOD disclosed ZERO AI revenue — not a dollar, not a percentage. The 10-Q reports a single operating segment with disaggregation only by geography and channel.
What was disclosed is a self-estimated content opportunity of ~$267 per platform across "combined AI application areas," versus ~$109 on the narrower AI-server basis. Read that carefully: the 2.4x increase is substantially a definitional broadening, not a demonstrated content win. There is no attach rate and no disclosure of what share DIOD captures.
The one real, verifiable AI hook is the timing/clock franchise — the only line with cited design wins ("multiple strategic server platform design wins"), active ramp ("now ramping into the latest AI server platforms") and backlog. Three independent forms of evidence. Everything else is content-attach mapping. DIOD's AI revenue is bounded above by the 28% computing segment and is certainly well below it.
The automotive story is better-evidenced than the AI story, and the market is paying for the AI one.
🚩 Insiders — zero open-market buying, every named officer sold
The [[pitfall-yahoo-insider-purchases-counts-rsu-grants]] trap fired and reconciles exactly. Yahoo reports 15,300 shares "purchased" in 6 transactions. Line by line: five $0.00 stock-award grants of 3,000 shares each on 2026-05-11 (Bull, Su, Ritter, Chen, Yu) plus a 300-share gift — 15,300 in 6 transactions, 100% grants and gifts. Zero open-market purchases.
Actual selling: ~187,228 shares across 18 sales.
| Cluster | Shares | Price range | Notable |
|---|---|---|---|
| May 2026 | ~61,844 (~$6.6M) | $96.81–$112.04 | CTO Tang $1.72M, CFO Whitmire $1.71M, CEO Yu, SVP Yang, SVP Zhao, SVP Tsong |
| Feb 2026 | ~130,589 (~$8.9M) | $59.19–$71.89 | Director/former CEO Keh Shew Lu: 111,000 sh, $7.52M |
Balanced read: a 216,000-share grant to five officers on 2026-01-30 explains the mechanics, and two Form 144s indicate pre-planned sales. But every named executive officer sold and not one bought, the May cluster executed at $96–112 (above today's price), and Lu's single trade was ~9.5% of total insider holdings. Net: neutral-to-mildly-negative. No post-print Form 4s exist yet — a purchase at $100 would be a genuine signal.
🔴 Analyst coverage is two people who disagree by 44%
| Firm | Rating | New PT | Prior | Implied vs $100.41 |
|---|---|---|---|---|
| Baird (Gerra) | Outperform | $192 | $120 | +91% |
| Truist (Stein) | Buy | $133 | $139 | +32% |
⚠️ Yahoo's targetMeanPrice $126.50 is arithmetically the pre-print state ($133 + $120)/2 and
has not ingested Baird's raise. Current 2-analyst mean is $162.50. And Yahoo's
recommendationKey: "strong_buy" with recommendationMean: 1.0 is two opinions averaged —
numberOfAnalystOpinions: 2. It carries no aggregation value whatsoever.
The most interesting datapoint in the whole print: Truist cut its target into a guide that beat EPS consensus by 30%. Both analysts were on the call; both maintained Buy; they moved targets in opposite directions by a combined 64 points. Truist's analyst was the one who probed fab utilization and got no answer — see below. This is not a consensus; it is a two-person disagreement about whether DIOD is mid-cycle or late-cycle, and the stock sits between them.
⚠️ Utilization is not disclosed — management explicitly declined
Truist: "where things stand on utilization…?" Gary Yu (CEO): "Usually, we don't provide this kind of P&L for that particular wafer fab."
No number, no target, no timeline. Anyone modeling a margin recovery off a utilization curve is modeling a number the company does not publish. Book-to-bill was likewise not disclosed and not asked.
🚩 The nearest-dated, least-discussed risk: HNTE
The FY2025 10-K states that three of four China facility HNTE approvals cover 2024–2026 (one 2025–2027), and that losing High and New Technology Enterprise status would raise the statutory rate for those facilities from 15% to 25%. Against a Q2 effective rate of 12.3% and full-year guidance of ~18%, that is direct mechanical damage to non-GAAP EPS, landing at the end of 2026. DIOD says it "expects to continue to meet HNTE requirements."
Also unmentioned on the call by anyone: tariffs — which the FY2025 10-K elevated to the first-listed risk factor, ahead of pandemics and fixed-cost operating leverage. Silence on the #1 disclosed risk is itself a datapoint.
🚩 A meaningful share of the boom is a competitor's catastrophe
Management has now volunteered this twice, unprompted: - Q1 2026: "We also continue to benefit from the market supply disruption." - Q2 2026: "The supply disruption I've mentioned on previous call continues."
That disruption is the Nexperia/Wingtech seizure — the Dutch government invoked the Goods Availability Act in September 2025; China then blocked exports from Nexperia's China packaging plants. Nexperia was the one discrete competitor demonstrably gaining share (9.7% of served markets in 2024, up from 8.9%, ~60% automotive). It became un-qualifiable for Western auto OEMs precisely as demand recovered. Wingtech is seeking $8B in arbitration damages.
A meaningful slice of the +30% Q3 guide is a windfall that reverses when Nexperia is rehabilitated. This is the single most important qualifier on the earnings base.
4. Valuation
Which earnings number to use — the whole question
| Basis | EPS | Multiple at $100.41 | Verdict |
|---|---|---|---|
| TTM GAAP EPS | $1.85 | 54.3x | Void — straddles trough quarters + a $20M investment gain |
| FY2025 non-GAAP | ~$1.43 | 70x | Trough |
| Q3 2026 guide annualized | $4.20 | 23.9x | Peak/near-peak run-rate |
| Yahoo "forward EPS" | $5.85 | 17.16x | 🔴 Reject — see below |
| Normalized (my estimate) | ~$3.50 | 28.7x | Primary anchor |
🔴 Reject the 17.16x forward P/E. It rests on a $5.85 forward EPS that exceeds the company's own ">$4.00" three-year target, and comes from a two-analyst panel that disagrees by 44%. It is the most quoted and least defensible number on this name.
Are the "3-year targets" a floor or a ceiling? — the most actionable finding
From the FY2025 10-K, Item 1: "3-year interim financial targets… achieving $2.0 billion in annual net sales with approximately $700 million in gross profit, or 35% plus, in gross margin", plus ">$4.00 non-GAAP EPS" (CEO, Q2 call).
| Interim target | Target | Q3 2026 guidance annualized | Status |
|---|---|---|---|
| Revenue | $2.0B | $510M × 4 = $2.04B | ✅ met at run-rate |
| Gross margin | 35%+ | 35.0% guided | ✅ met |
| Non-GAAP EPS | >$4.00 | $1.05 × 4 = $4.20 | ✅ met at run-rate |
All three "3-year" targets set in early 2026 are met on Q3 2026 guidance — two-plus years early. Either they were sandbagged, or 2026 is an overshoot. Management did not address this on the call and no analyst asked. Watch for a target reset at Q3 or an analyst day; its absence would be a tell.
My normalization
Stripping the Nexperia windfall and the AI-spike overshoot:
| Assumption | |
|---|---|
| Sustainable revenue | $1.90B (vs $2.04B run-rate) |
| Gross margin | 34% → gross profit $646M |
| Opex | 24.5% of revenue = $466M → operating income $180M (9.5%) |
| Net interest income | ~$20M → pretax $200M |
| Tax @ 18% | → net $164M |
| Normalized non-GAAP EPS | ~$3.50 on 46.4M shares |
Models applied conditionally (analysis_notes §3)
Graham's Intrinsic Value — genuinely applicable here, unlike DT. DIOD is profitable, asset-backed, with real tangible book (BVPS $41.13, of which only $183M — ~$3.95/share — is goodwill).
| EPS basis | Graham IV = √(22.5 × EPS × $41.13) |
|---|---|
| TTM GAAP $1.85 | $41.38 |
| Normalized $3.50 | $57.16 |
| Peak run-rate $4.20 | $62.34 |
Even on peak earnings, Graham says $62 against a $100.41 price — a 61% premium. For a value play with real book value, analysis_notes weights Graham heavily. It is a clear caution.
Bogle's Expected Return — the number that decides this.
| Component | Assumption | Contribution |
|---|---|---|
| Dividend yield | none | 0% |
| Earnings growth | From a normalized base: ~5–7% revenue growth with mix-driven margin expansion | +8 to +12%/yr |
| P/E change | From 23.9x (on peak run-rate) toward a normalized 16x over 5 years | −7.6 to −9.5%/yr |
| Expected annual return | ≈ 0% to +4%/yr |
That is the crux of the entire case. The business can grow and the shareholder can still earn nothing, because the multiple compression cancels it. Buying a cyclical at a peak multiple on peak earnings is the specific way this asset class destroys returns.
DYT and DDM: N/A. No dividend.
Fair value range
| Scenario | Assumptions | EPS | Multiple | Value |
|---|---|---|---|---|
| Bear | Cycle rolls over, Nexperia rehabilitated, back to ~$1.6B / 32% GM | ~$2.20 | 14x | $31 |
| Base | Normalized $1.90B / 34% GM; auto mix holds at 21%; thin moat caps the multiple | $3.50 | 16–19x | $56–67 |
| Bull | $2.3B / 37–38% GM sustained; ElevATE accretes; GFAB/SPFAB fill by 2028 | ~$6.00 | 18x | $108 |
Fair value: $60–85. Central estimate ~$72. Price $100.41.
The stock trades ~18% above the top of the base case and ~39% above its midpoint. It is only justified in the bull case, which requires a gross margin DIOD has never held outside a global chip shortage.
Sanity check: the stock is +137% off its 52-week low of $42.28 and +74% over 52 weeks (S&P +19.9%). Baird's $192 requires roughly 22x on the bull-case EPS — i.e. a compounder's multiple on a peak cyclical's earnings on a 2/5 moat. I do not underwrite that.
5. Conflict resolution — the debate round
Tension 1 (the real one) — Moat says "the risk is paying a growth multiple for a cyclical peak inflated by Nexperia." Sentiment says "channel inventory below the normal band and falling with record POS — this is a waypoint, not a peak."
Both are supported by hard, specific, primary-source evidence, and they resolve on time horizon, not on truth:
- The channel data governs the next 2–3 quarters. Under-shipping consumption with below-normal channel weeks means Q3 and Q4 are more likely beaten than missed. Sentiment wins the near term outright.
- The Nexperia windfall governs 2027+. Management admitted it twice. When Nexperia is requalified, a slice of the base leaves. Moat wins the medium term.
Neither resolution justifies $100.41. A name that is safe for two quarters and structurally uncertain thereafter, trading 39% above base-case fair value, is a watch — not a buy. This is the honest "close call because…" the framework asks for, and it lands on price, not on quality.
Tension 2 — the China bear case failed, so should conviction rise? Yes, and it did: the business quality assessment is better than the entry premise. DIOD's China revenue is at a record, its Chinese manufacturing base blunts domestic substitution, and Chinese competitors are raising prices. Conviction on the business moved up. But the moat is still 2/5 and through-cycle ROIC is still ~10% — a better cyclical is still a cyclical, and it does not earn a growth multiple.
Tension 3 — Truist cut its target into a 30% EPS beat while Baird raised 60%. With two analysts, the "consensus" is noise. I side closer to Truist's direction (multiple compression on cycle-peak concern) while noting Truist gave no public rationale. My $60–85 is below both.
A portfolio-specific passage was removed from the public build.
7. Verdict
WATCH — conviction 5.5/10
Fair value $60–85 · Entry $60–72 · Trim: explicit instruction, NOT a multiple
Why 5.5 and not lower: the cyclical recovery is real and the evidence for it is unusually strong — the disclosed ASP series showing 2025 prices above 2021 kills the structural-decline thesis; the channel-inventory disclosure is a hard, falsifiable, bullish datapoint; the automotive mix broke a three-year plateau; the balance sheet is a fortress; and the China bear case, the one I expected to confirm, substantially failed.
Why 5.5 and not higher: the moat is 2/5, through-cycle ROIC is roughly the cost of capital, management's own $2B target implies ~9.4% ROIC, the 40% gross-margin ambition exceeds anything achieved outside a shortage, utilization is undisclosed, coverage is two analysts, every officer sold and none bought, three of four China HNTE tax approvals expire at end-2026, and a meaningful slice of the boom is a competitor's government seizure that management has flagged twice.
Why WATCH and not AVOID: this is a genuinely good cyclical business at the wrong price, not a bad business. It belongs on the watchlist with a real entry zone, because the next downcycle will hand it back at a price worth paying.
🔴 Trim convention — the multiple form is deliberately NOT used
Per [[pitfall-multiple-trim-inverts-on-peak-cycle-cyclicals]]: site.py computes
trim $ = multiple × (price / Yahoo forwardPE). For DIOD that implied EPS is
$100.41 / 17.16 = $5.85 — a number this analysis explicitly rejects as exceeding the
company's own target and sourced from a two-analyst panel. Even a conservative Trim 20x fwd
would render $117, instructing a hold through the top of the cycle.
Instruction instead: trim on strength above ~$110, or on any quarter where GAAP gross margin
plateaus at 34–35% while revenue sets records (that combination marks price surrendered to
hold sockets — structural margin loss, not cyclical give-back). Re-set at each /analyze; this
is the cost of not using the multiple form.
Key risks, named
- Nexperia rehabilitation removes a windfall management has admitted to twice. The single biggest qualifier on the earnings base.
- HNTE expiry at end-2026 — 15% → 25% on three China facilities. Concrete, near-dated, undiscussed.
- Cycle-peak risk — all three "3-year" targets met at run-rate, two-plus years early.
- Geographic concentration — ~45% of sales to China, majority of manufacturing there, 76.4% of net PP&E in Asia, with the 10-K contemplating "forfeiture of our assets."
- Two-analyst coverage makes every consensus number on this name structurally unreliable.
- GFAB/SPFAB underutilization to 2027–28 — the most credible remaining margin lever and the most credible reason to doubt the 40% target.
Watch triggers
| Date/event | What to watch | Why it matters |
|---|---|---|
| Q3 print, ~early Nov 2026 | Does GAAP gross margin land at/above 35.0%? | The whole margin thesis in one number |
| Q3 print or an analyst day | A reset of the 3-year targets | A reset is bull confirmation; silence is a tell |
| FY2026 10-K (~Feb 2027) | HNTE renewal disclosure | Mechanical EPS damage if not renewed |
| Any quarter | Channel inventory returning into the 11–14 week band | The below-band read is the bull case; normalization ends it |
| 2H 2026 | ElevATE close + first consolidated quarter | ~$50M revenue at above-corporate margin; watch for guidance re-basing |
| Ongoing | Nexperia/Wingtech resolution | Removes the windfall |
| Ongoing | Post-print Form 4s — any open-market insider buy | Would be the first in the record and a real signal |
| Ongoing | Any third analyst initiating | Fixes the structural weakness in every consensus number |
Data quality notes for the KB
- Yahoo
PE(fwd)17.16 rests on a $5.85 EPS that exceeds the company's own ">$4" target, from a 2-analyst panel. Unusable. - Yahoo
targetMeanPrice$126.50 is the pre-print state ($133+$120)/2. Post-print mean is $162.50.recommendationKey "strong_buy"= two opinions. - Yahoo
EV$3.69B implies ~$920M net cash vs the 10-Q's ~$402M — appears to sweep in long-term investments. Use ~$4.21B. - Yahoo
heldPercentInstitutions101.98% — impossible, known artifact. Two Vanguard entities each showing exactly +100% QoQ is a reclassification artifact, not accumulation. - Yahoo
insider_purchases15,300 shares = 100% grants and gifts, reconciled line-by-line. [[pitfall-yahoo-insider-purchases-counts-rsu-grants]] verified for DIOD. ir.diodes.comandsec.govreturn HTTP 403 to WebFetch. The reliable path to DIOD primary data is roic.ai MCP for the transcript +curl -A "<name> <email>"againstsec.gov/Archives/.agent-browsertimed out on diodes.com.- Secondary-source errors caught: MarketBeat reported ElevATE adding "~$15M in first-year revenue" — the CEO said ~$50M, wrong by 3.3x. "Sixth consecutive quarter of double-digit growth" — the transcript says fifth.
- DIOD publishes its own weighted-average ASP change annually in MD&A. This is a reusable technique for separating price from volume on any cyclical component supplier, and it is what decided §2. Worth a Playbook note.