Lumentum

Refresh delta (since 2026-06-18)

What changed — six material moves

1. The stock had a 39% peak-to-trough drawdown and is now 14% below the prior dossier's price. $1,053.09 all-time-high close on 2026-05-11 → $644.56 on 2026-07-28 (-38.8%) → $833.78 as of 2026-08-10 11:35 EDT. Prior dossier's price at analysis was ~$970–982 (2026-06-05). Market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. $64.87B; shares outstanding 77.80M; TTM P/E 158.51. Still +112% YTD 2026.
2. The 2028 convertible overhang was equitized. On 2026-05-29 Lumentum signed privately-negotiated exchange agreements delivering ~5.0M common shares for ~$650.4M principal of the 0.50% 2028 Notes plus conversion value, closing on or about 2026-06-04, leaving ~$172.2M of 2028s outstanding on unchanged terms. Incremental dilutionDilutionIssuing new shares, so each existing share owns a smaller slice of the same company. attributable to the principal: ~0.8M shares. No cash proceeds; the notes are cancelled. This directly answers management question #10 from the prior dossier.
3. Washington is drafting a ban on Chinese optical transceivers — the single most thesis-relevant new fact. Reuters reported the Trump administration and the FCC are drafting a ban on US imports of new Chinese data-center optical transceivers, explicitly naming Zhongji Innolight (~27% of the global transceiver market). It is a draft proposal, not enacted — sources said the FCC could modify or abandon it; officials aim to finalize and enforce by year-end 2026. LITE +9% Monday 2026-08-03 and +6% Tuesday 2026-08-04 to $829.42; COHR +11%, AAOI +17%.
4. The supply-demand gap widened rather than closed — and the guided share count is bigger than the prior dossier assumed. From the newly-ingested fiscal-Q3 call: CEO Hurlston said the imbalance is "probably even higher than we reported in our last call, somewhere greater than 30%," particularly in EMLs and pump lasers; "we are significantly undershipping demand… having to make choices as to who we support." Fiscal-Q4 guidance is built on approximately 102 million diluted shares — not the ~96M the prior dossier used for its per-share math.
5. Index flows turned against the name, then AI-capex doubt hit it. Late June: removed from multiple Russell benchmarks (Russell 2500, several Russell 3000 / Midcap Value) at the annual reconstitution → -11.7%. Then 2026-07-28: -9% to $644.56 after Alphabet printed Q2 capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs. of $44.92B (double the year-ago figure) with -$5.86B free cash flow and a FY2026 capital budget of $195–205B. The Nasdaq-100 inclusion (2026-05-18) that the prior dossier counted as a tailwind has been more than offset.
6. Two management items and a competitive one. (a) Vincent Retort, EVP Global Reliability & Quality, notified the company on 2026-07-27 of retirement effective October 2026, with a two-year consulting term under which his outstanding equity awards keep vesting. (b) Insider selling has grown to $38.8M over 90 days from the ">$23M over three months" the prior dossier cited. (c) TrendForce projects combined monthly EML + CW-DFB capacity roughly doubling to ~50.7M units/month in 2026 — the bear's "supply catches up" leg now has an industry number attached to it.

What held

The structural thesis stands, and on the operating side it strengthened. Every load-bearing claim in the prior dossier survives contact with the new material:

• The chokepoint is real and got tighter, not looser. Undershipping demand by >30%, worse than the prior call's 25–30%. EML shipments +100% YoY; management expects EML units +>50% from December 2025 to December 2026.
• Operating leverage is intact and still building. Non-GAAP gross margin 47.9% (+540bps QoQ, +1,270bps YoY), non-GAAP operating margin 32.2% (+700bps QoQ, +2,140bps YoY), guided to 35–36% for the June quarter.
• The second and third acts are ramping. OCS backed by "a recently announced multiyear, multibillion-dollar purchase agreement"; internal CW laser production to supply ~20% of transceivers in the June quarter, a margin tailwind as insourcing scales.
• The NVIDIA lock is unchanged and the Greensboro fab — the fifth InP fab, acquired from Qorvo, converting from gallium arsenide — is proceeding, but management was explicit that its capacity is not in revenue projections, with line contributions "six or so quarters away".

What weakened. Not the business — the per-share arithmetic and the duration of the moat. The share base is ~6% larger than the prior dossier modeled, FY26 non-GAAP EPS is landing ~12% below its estimate, and the industry is adding EML capacity on a schedule that now has a public number. Separately, a secondary source argues Coherent's proprietary 6-inch indium-phosphide platform yields more die per wafer at lower cost than Lumentum's smaller-wafer lines — flagged, not adopted: this is a substack claim, not a filing or a company statement, and Lumentum's own answer to capacity (a fifth fab) is on the record. It belongs on the watch list, not in the thesis.

Active-call review finding

Call under review: HIGH · "Lumentum: The Optics Play That Bridges Power and Memory."

• Supports the core premise. The optics leg of the call is better supported than at the last dossier: the supply gap widened past 30%, EML units are guided +>50% into December 2026, non-GAAP operating margin is guided to 35–36%, the OCS lane now sits on a multiyear multibillion-dollar purchase agreement, and a draft US ban on Chinese transceivers would hand share directly to non-Chinese suppliers.
• Weakens one leg — the "bridges power and memory" framing. Nothing in the new primary material evidences a memory bridge. The new disclosures cluster around lasers, pump lasers, ultra-high-power lasers, optical circuit switching and external laser sources for CPO — interconnect and photonics, not memory. The pump-laser and "multi-rail" commentary is the closest thing to a power-adjacent read, and it is about optical amplification, not electrical power delivery. If the call's thesis leans on a memory linkage, that leg is unsourced and should be either evidenced or dropped at the next review.
• Breaks nothing. No fact found in this refresh falsifies the call. The genuine new risks are (i) a ~6% larger diluted share base than modeled, (ii) an industry EML/CW-DFB capacity roughly doubling in 2026, and (iii) demonstrated 39% peak-to-trough beta to AI-capex sentiment that has nothing to do with Lumentum's own execution.
• Timing note for the reviewer: the fiscal-Q4 print lands 2026-08-11 after the close. A HIGH-conviction call on a name at ~103x the fiscal year it is about to close should be re-read against that print before any position action is contemplated. This is a review finding, not a trade proposal.

---

Phase A — Understand the business

Company Overview — carried, one amendment

Unchanged from the previous dossier. One amendment: the prior dossier's product map should now read five indium-phosphide fabs, not four — the Greensboro, North Carolina site acquired from Qorvo is explicitly "a fifth indium phosphide fab," currently converting from gallium arsenide, with the transferred workforce included in the purchase and no revenue contribution in current projections ("six or so quarters away").

Supply Chain — unchanged, see the previous dossier

Competitive Advantages — carried, with two amendments

The moat description (process + IP in the laser chip; vertical integration; OCS/MEMS; the NVIDIA lock) stands. Two things the new material changes:

• Vertical integration is now quantified and inflecting. Internal CW laser production is expected to supply ~20% of Lumentum's own transceivers in the June quarter, with management naming it a margin catalyst — and, critically, necessary because of external supply constraints. The prior dossier described vertical integration structurally; it now has a percentage and a margin mechanism.
• The scarcity clock has a public number on it. The prior dossier's honest read — "Lumentum holds the technical whip hand today because capacity is scarce; that flips if EML supply catches up to demand" — now has an industry forecast attached: combined EML + CW-DFB monthly capacity roughly doubling to ~50.7M units/month in 2026. Against that, Lumentum is still described as the only supplier shipping 200G-per-lane EMLs in volume and the gap widened this quarter. The two facts are not contradictory — capacity is doubling into demand that is growing faster — but the bear now owns a date-stamped supply curve, which it did not in June.

---

Phase B — Measure performance

Segments — carried, now backed by CSV

No new quarter. The product-type split is unchanged from the prior dossier and is now written to the research layer for the first time (our figures, six rows, three quarters):

What the newly-ingested call adds underneath the split: within Components, narrow-linewidth laser assemblies +>120% YoY and pump lasers +80% YoY — the growth is broader than "EMLs into 1.6T," which matters because it means the franchise is not a single-SKU story. Within Systems, cloud transceivers +>40% sequentially, with 1.6T production stepping up in the June quarter at margins "better than 800-gig" but transceiver segment margin overall "still a bit challenged."

Earnings Result — no new print; the print lands 2026-08-11

This is the honest state: the latest reported quarter is still fiscal Q3 FY26 (ended 2026-03-28), the same quarter the prior dossier analyzed. Nothing in Lens 5's reported figures has changed. What has changed is that the guidance is now written to the research layer and resolves tomorrow.

The bar being cleared tomorrow:

At the guide midpoint, FY2026 revenue = $2,007.7M (nine months, GAAP) + $985M = $2,992.7M, which ties to the street's $2.99B almost exactly. That agreement is worth noting: the sell side is modeling the guide, not above it.

Balance-sheet flags — what to check in the 10-K, and why. As of 2026-03-28 the balance sheet carried $3,238.6M of debt in current liabilities because every convertible series had breached its 130%-of-conversion-price trigger, against $2,617.8M cash and $554.5M short-term investments; total debt $3,281.8M; total equity $2,973.4M. Since that balance-sheet date two equitization events have settled: the April 2026 exchange of ~5.7M shares for portions of the 2026 and 2029 Notes (disclosed as a subsequent event in the 10-QForm 10-QThe quarterly version of the annual report. Lighter, and not audited.), and the June 4 exchange of ~5.0M shares for $650.4M of the 2028s. The FY26 balance sheet should therefore show materially lower total debt and a much smaller current-debt reclass. If it does not, that is a flag — it would mean the equitization did not do what the 8-KForm 8-KA filing for something that happened between reports and matters enough to tell shareholders now. described.

Four more things to verify in the print, ranked by how much they would move the thesis:

1. Diluted share count. Guidance was built on ~102M; fiscal-Q3 non-GAAP used 95.2M and GAAP diluted was 96.2M. Every per-share number in this dossier is sensitive to this line.
2. Customer concentration. The prior dossier's 26% / 38% (top one / top two, fiscal Q3) is the live number and is unrefreshed. A move up is the bear case arriving.
3. Inventory. $632.8M at 2026-03-28, +$62M sequentially, against the auditor's only Critical Audit Matter (excess-and-obsolete). It has grown slower than revenue every quarter of FY26 — watch for that to break.
4. First FY2027 framing. The street is at $18.73 FY27 EPS, +128% on FY26. That is the number the multiple rests on.

Market reaction, refreshed. The tape's behaviour since June confirms the prior dossier's read that LITE no longer trades on its own print. It fell 5.62% on the day of the fiscal-Q3 beat-and-raise. Since then the moves that mattered were an index reconstitution, an Alphabet capex line, and a Reuters policy leak — none of them Lumentum's own numbers.

Earnings Calls (sentiment trend) — upgraded from to primary

The prior dossier ran this lens entirely on second-hand summaries because transcripts/ was empty. It is no longer empty: transcripts/2026-q1.md (the 2026-05-05 fiscal-Q3 FY26 call, Fool transcript, 55.9KB) is on the shelf.

Tone across the last four calls (FQ4 FY25 → FQ3 FY26): escalating, and specifically escalating on scarcity, not on demand. The revenue path is $480.7M → $533.8M → $665.5M → $808.4M → guided ~$985M. But the sentence that defines this call is a supply sentence, not a demand one:

Hurlston: "We are significantly undershipping demand. And we're having to make choices as to who we support." The imbalance is "probably even higher than we reported in our last call, somewhere greater than 30%."

That is a deterioration in the prior call's number (25–30%) presented as good news, and it is the correct way to read it — but it is also a management team publicly admitting it is rationing its own customers. That has a second-order cost the bulls do not price: a rationed customer qualifies a second source.

What is new in the language versus the prior three calls:

• "Tight rope." Management described OCS and pump-laser production as being on a tight rope, with constraints that arrived "somewhat unanticipated" and "relatively suddenly". This is the first call where the constraint is framed as a risk to the ramp, not just as evidence of demand.
• Commercial terms are being renegotiated in the company's favour. Management is "negotiating prepayment, take-or-pay, and price-increase terms with key customers to secure capacity and offset capital expenditures." That is the single most bullish structural sentence on the call — take-or-pay and customer prepayments convert a cyclical hardware business into something with contracted revenue and customer-funded capex. It is also, read the other way, the leash the prior dossier's contrarian section warned about, extended to more customers.
• A financial target with a clock. $1.25B/quarter at ~35% operating margin near-term, and $2.0B/quarter at 40% non-GAAP operating margin on an 18–24-month timeline. Management has put its own number on the bull case.
• What they stopped saying remains what the prior dossier found: no destocking, no underutilization, no "normalization." Add to that — they have stopped talking about Industrial Tech almost entirely, consistent with the one-segment re-report.

Comps — unchanged, see the previous dossier (one refreshed data point)

The peer table is not re-sourced here. One point worth carrying forward: as of 2026-08-04 the two closest comps traded at COHR TTM P/E ~127.17x and LITE ~128.24x — i.e. the ~8-point premium LITE carried on forward multiples in June has compressed to roughly parity on trailing. LITE's own TTM P/E is quoted at 158.51x by a different aggregator on 2026-08-10. Conflict surfaced, not resolved: the two TTM figures are ~24% apart and I could not reconcile which trailing EPS base each uses. Treat trailing multiples for this name as unusable and work forward.

Stock-Price Catalysts — refreshed for the 53-day window

The prior dossier's three-regime history (2021–22 consolidation, 2023–mid-24 telecom trough, late-2024→mid-2026 AI re-rating) stands. What the last 53 days added is a fourth phase the prior dossier could not see: the first real AI-optics de-rating. Every move below is >5% and dated [all web]:

Peak-to-trough: -38.8% in eleven weeks, on a business whose own numbers improved throughout. What the pattern reveals, updated: the marginal buyer is an AI-thematic fund and the marginal seller is now an index rebalance and a hyperscaler capex headline. The prior dossier called this vulnerability; the last 53 days priced it. The one genuinely idiosyncratic driver in the whole window — the Chinese-module draft rule — is worth +15% across two sessions, which is a fair estimate of what a final rule would be worth again, and of what abandoning it would cost.

---

Phase C — Judge people & books

Management — refreshed

The prior dossier's read on Hurlston (proven datacom-optics operator, hired-closer profile, ~16 months in — now ~18) is unchanged and unchallenged by anything in this window. Three updates:

• A senior departure, handled quietly and generously. On 2026-07-27 Vincent Retort, EVP Global Reliability & Quality, notified the company of retirement effective October 2026, with a two-year consulting term during which his outstanding equity awards continue to vest on their existing schedules provided he remains a service provider. Honest read: this is a low-signal event with one nuance. Reliability & Quality is not a P&L seat, and retirement with a consulting tail is the standard way to retain institutional knowledge. The nuance is which function: in a company whose entire moat is InP epitaxy yield and reliability at 200G/lane, and which is simultaneously converting a GaAs fab to InP and standing up a fifth site, the reliability/quality function is closer to the moat than its title suggests. The two-year consulting term reads as the board pricing exactly that. Not a red flag; a thing to watch alongside any yield commentary in the print.
• Insider selling continued and grew. ~$38.8M sold over the trailing 90 days as of 2026-07-19, versus ">$23M in the past three months" in the prior dossier. Caveat the prior dossier did not apply and should have: a material share of reported "disposals" at this company are tax-withholding events on RSU vesting, not open-market sales — e.g. CEO Hurlston's 2026-05-15 Form 4Form 4The filing an insider must make within two business days of buying or selling their own company’s shares. was 5,438 shares withheld by the company at $970.70 to satisfy vesting taxes, not a sale. The signal is real (no insider purchases surfaced) but weaker than the headline dollar figure implies. Do not treat $38.8M as $38.8M of conviction-selling.
• Capital allocation, updated. The June equitization is the first capital-structure action of this management team that reduces rather than adds leverage, and it was done without cash. Combined with the take-or-pay / prepayment negotiations described on the call, the direction of travel is: customer-funded capacity, equity-settled debt, no buyback. For a company at ~103x the closing fiscal year with a dilution history, not buying back stock is the correct call, and they are making it.

Forensic Red Flags — refreshed

The prior dossier's forensic read stands: clean audit (Deloitte, unqualified on FY25 financials and ICFR, single Critical Audit Matter on inventory E&O), a convertible stack as the dominant flag, a ~37% GAAP/non-GAAP wedge driven by SBC and acquired-intangible amortization, an FY25 "profit" that was entirely a $198.0M tax benefit, and one-offs (a $34.9M facility gain in FY25, a $27.5M escrow settlement in FY26) that must be stripped for a clean run-rate. All of that is now encoded in our figures note fields rather than living only in prose.

What changed:

• The dominant flag is materially smaller. Total debt was $3,281.8M at 2026-03-28 with $3,238.6M sitting in current liabilities on the 130% conversion trigger. Two equitizations have settled since (April: ~5.7M shares for portions of the 2026 + 2029 Notes; June 4: ~5.0M shares for $650.4M of the 2028s, leaving ~$172.2M). The 10-Q's stark warning that mass conversion "could adversely impact our ability to continue as a going concernGoing concernAn auditor’s formal doubt that the company can fund itself for the next year. It is a warning about survival, not about performance." is being defused deliberately and in equity, not cash. The remaining stack is dominated by the $1,265M 0.375% 2032 Notes (partially hedged by a $102.0M capped call).
• The cost is where you would expect it: shares. Common outstanding went 71.7M (2026-03-28) → 77.80M (2026-08-10). Add the 2,876,415 Series A preferred held by NVIDIA, convertible one-for-one at $695.31 — in the money at $833.78 — and the as-converted common base is ~80.7M. Management's own guided fully diluted count is ~102M. Any per-share figure on this name that uses 71M, 78M or even 96M is wrong.
• Regulatory: unchanged, re-verified. regulatory/regulatory-findings.md was re-fetched on 2026-08-10 covering 2021-08-10 → 2026-08-10 across SEC Litigation Releases and AAERs: total_sec_findings: 0 — no LR, no AAER naming the issuer. The prior dossier's two web-sourced items (the insider-trading ring against a former CISO in which Lumentum was the victim, and a $7.8M export-control settlement charged in fiscal Q2 2023) are unchanged and remain the only "findings." Nothing new surfaced in this window.
• A new regulatory item on the other side of the ledger. The draft FCC/administration ban on Chinese optical transceiver imports is a regulatory event that would benefit the issuer. It is not a compliance exposure; it is a policy option written on Lumentum's competitive position, and it is unfinalized.

---

Phase D — Project & stress-test

Forward Projection — rebuilt, with a correction to the prior dossier

Model status — read this before any number below. model.xlsx was built for the first time this run (our model, MODEL_VERSION 2.0, 7 quarters 2024-Q3 → 2026-Q1, opening balance sheet sourced from our figures, not hand-typed). Reading it back returns: "No computed values. The workbook has 51 formulas and no cached results." Share price and CAPM inputs are blank; the balance check cannot be read. Therefore no valuation output from the workbook is cited anywhere in this dossier. The workbook is a scaffold on disk awaiting an Excel pass, and every figure below is an explicit `` with the arithmetic shown.

Correction to the prior dossier — two errors, both in the direction of flattery.

1. FY2026 non-GAAP EPS is ~$8.1, not the ~$9.0–$9.5 the prior dossier estimated. Actual reported quarters: Q1 $1.10, Q2 $1.67, Q3 $2.37 = $5.14 nine-month non-GAAP. The prior dossier estimated the nine months at "~$6.5" — a $1.36 overstatement that flowed straight into its full-year figure. Adding the Q4 guide midpoint $2.95 gives FY26 ≈ $8.09; street is $8.23. (Summing quarterly EPS is approximate because the diluted share count rose every quarter — 95.2M in Q3, ~102M guided for Q4 — which is most of the gap to the street's $8.23.)
2. "~52x FY26" was never FY26. At the prior dossier's own ~$970 price and its own ~$9.3 EPS, the multiple was ~104x, not 52x. The 52x figure came from an aggregator's forward P/E, which is a next-twelve-months blend. Same today: at $833.78 the quoted forward P/E of 52.21 implies an EPS base of ~$15.97 — a blend of the June quarter and the first three quarters of FY27, not a fiscal-year figure. Conflict surfaced, resolved with arithmetic rather than by averaging.

Where the multiple actually is, at $833.78:

Three-year non-GAAP EPS path — rebuilt on the guided ~102M share base.

Shared arithmetic for every cell: EPS = ((revenue × non-GAAP operating margin) + net other income) × (1 − 15% tax) ÷ diluted shares. Net other income is taken as +$0.08B/yr. Tax at 15%. Diluted shares 104M FY27, 106M FY28.

Two sanity checks on that table. (i) My base FY27 of $17.6 sits just under the street's $18.73 — the sell side is modeling either slightly more revenue or slightly fewer shares; the two are within a rounding error of each other, which is reassuring for the base but means the base case is the consensus case, and consensus is what a 44.5x multiple is already paying for. (ii) The bull case is management's own stated target: $2.0B/quarter at 40% non-GAAP operating margin on an 18–24-month timeline from March 2026 annualizes to $8.0B at 40% → (8.0×0.40 + 0.08) × 0.85 ÷ 106 = $26.3, i.e. 31.7x at today's price. That is the cleanest single frame for this stock: you are paying ~32x for management hitting its own publicly stated 18–24-month plan, and ~103x if it stops here.

Downside arithmetic. A de-rate to 30x on bear FY2028 EPS of $11.7 is ~$351, i.e. -58% from $833.78. The prior dossier's bear landed at ~$380 on different inputs; the two agree that the downside is roughly a halving and is larger than the street's ~32% upside to a $1,105 consensus target. Nothing about the 14% price decline since June has changed that asymmetry — because the earnings base fell about as much as the price did.

No Brier forecast logged — our model create is out of scope for an unattended refresh. The natural one to log, unchanged from the prior dossier and now sharper: "LITE FY27 non-GAAP EPS ≥ $17, resolves ~2027-08-15."

Bull vs Bear — updated

Bull case, updated. The toll-booth thesis is intact and the toll got more expensive to bypass. Lumentum is undershipping demand by >30% and rationing customers; EML units are guided +>50% into December 2026; narrow-linewidth assemblies are +>120% YoY and pump lasers +80%, so this is a franchise, not a SKU; non-GAAP operating margin went 32.2% → guided 35–36% with 40% as the stated destination; OCS sits on a multiyear multibillion-dollar purchase agreement; internal CW lasers hit ~20% of own-transceiver content this quarter with margin upside as it scales; a fifth InP fab is being stood up with none of its capacity in guidance; and management is converting the shortage into take-or-pay, prepayment and price-increase terms — customer-funded capacity, which is how a cyclical hardware business becomes a contracted one. On top of that, Washington is drafting a rule that would remove a ~27%-share Chinese competitor from the US market. Meanwhile the price is 21% below its May high and the FY27–28 multiple has compressed.

Bear case (permanent-impairment lens), updated.

1. The per-share denominator keeps growing and the market keeps quoting the numerator. ~102M guided diluted shares against 77.8M outstanding. The dilution is not an accident of the past — it is the mechanism by which the debt is being retired.
2. The scarcity moat now has a public expiry schedule. Industry EML + CW-DFB capacity roughly doubling to ~50.7M units/month in 2026, plus a credible (if secondary-sourced) claim that Coherent's 6-inch InP platform is structurally lower-cost per die. Lumentum's own answer — Greensboro — is six quarters out.
3. The stock is a levered proxy on hyperscaler capex sentiment and just proved it. -38.8% peak-to-trough in eleven weeks on an Alphabet capex line and a Russell rebalance, with company results improving throughout.
4. ~103x the fiscal year that ends tomorrow. Even on the street's own FY27, 44.5x. There is no margin of safety at this price; there is a bet on a specific execution path.

Pre-mortem (18 months out, thesis broke — what happened?). The Chinese-transceiver rule was quietly abandoned in Q4 2026 after hyperscaler lobbying over input costs, removing the August catalyst and returning Innolight to the US market. Hyperscaler capex growth flattened in CY2027 as the -$5.86B FCF quarters that spooked the tape in July 2026 turned into board-level discipline. The >30% supply gap closed — not because Lumentum failed to add capacity, but because the industry added ~2x monthly EML/CW-DFB units and Coherent's larger-wafer platform undercut on price; ASPs fell, gross margin slipped from ~48% toward ~40%, and the take-or-pay terms Lumentum negotiated at the peak of scarcity became the thing customers renegotiated. NVIDIA, having secured its capacity, dual-sourced the next node. Greensboro's capacity arrived into a softer market. Diluted shares reached ~106M. FY28 EPS came in near $12 instead of $23, the multiple went to 30x, and the stock printed ~$350 — down 58% — with the business still fine and still the leader.

Are multiples too high? At ~103x the closing year, yes unambiguously. At ~32x management's own 18–24-month target, no. The whole question is which of those two numbers is the right denominator, and the answer arrives in installments starting tomorrow.

Contrarian view — updated, and it has moved. The prior dossier's contrarian read was "the NVIDIA $2B is as much a leash as a gift." That still holds, and the newly-ingested call extends it: management is now negotiating take-or-pay and prepayment terms with multiple key customers to fund capacity. Read one way that is contracted revenue. Read the other way, Lumentum is selling forward the exact scarcity that is producing today's 48% gross margin — locking in price and volume at the moment its bargaining power is at a cyclical maximum, with the counterparties being the largest buyers on earth. The market is scoring these agreements as pure de-risking. They are also a ceiling, negotiated by the party with more lawyers.

Devil's Advocate (short-seller) — updated

Dismantling the bull case with what the last 53 days actually produced.

• The single best short argument is now arithmetic, not narrative. The bull deck quotes "$2.37 EPS, 90% growth." The company's own guidance is built on ~102M diluted shares against 77.8M outstanding — a 31% wedge. The prior dossier, written by a careful analyst, still overstated FY26 EPS by ~12% and mislabeled a 104x multiple as 52x. If a dedicated dossier got the per-share math wrong in the bull's favour, the marginal buyer certainly has.
• "Undershipping by >30%" is not unambiguously bullish. Management is choosing which customers to disappoint. Every rationed customer has a procurement team whose job is now to qualify Coherent or a Chinese alternative. Scarcity that lasts one quarter is pricing power; scarcity that lasts eight quarters is a second-source program.
• The best new bull catalyst is a draft. The +15% across 2026-08-03/04 rests on a Reuters report of a rule the FCC "could still modify or abandon," targeted for year-end 2026. Hyperscalers — Lumentum's own customers — bear the input-cost increase and have every incentive to lobby against it. That entire +15% is a policy option, not an earnings change, and it is being carried in the price into a print.
• Capacity is arriving, and Lumentum's is arriving last. ~50.7M units/month of industry EML + CW-DFB capacity in 2026 versus a fifth fab whose "line contributions" are "six or so quarters away" and are not in projections. Between now and then, the incremental supply is somebody else's.
• The margin mix is weaker than the headline. Transceiver segment margin is "still a bit challenged" even as 1.6T ramps; the 47.9% non-GAAP gross margin is carried by Components. Systems grew 121% YoY — the fastest-growing line is the lower-margin one.
• Assumptions that must all hold for $833.78: (1) AI capex compounds through CY27–28 without a digestion air-pocket — Alphabet's -$5.86B FCF quarter says that is not free; (2) Lumentum holds its 200G/lane lead as industry capacity doubles; (3) gross margin holds ~46–48%; (4) OCS and CPO inflect on schedule despite management calling OCS "the biggest single tight rope"; (5) no NVIDIA second-sourcing; (6) diluted shares stop at ~102–106M. Six now, not five — the share count joined the list.
• Single scenario that permanently impairs: the architecture transition (CPO / LPO / silicon photonics cutting discrete-EML content per port) arriving simultaneously with the industry capacity build and a hyperscaler second-source. Plausibility unchanged from the prior dossier — moderate and rising with time; a 2027–28 risk, not a 2026 one — except that the capacity leg now has a date on it.

Management Questions — unchanged, see the previous dossier

All fifteen questions stand. Question #10 (the plan for the 2028 Notes) has been answered by events — $650.4M equitized on 2026-06-04, ~$172.2M stub remaining. Question #6 (the fully-diluted share count management is managing to) has been partially answered: ~102M. Two questions to add for the 2026-08-11 call:

1. The take-or-pay, prepayment and price-increase agreements under negotiation — what portion of FY27 capacity is now contractually committed, at what price protection, and what happens to those prices if spot EML supply loosens?
2. With industry EML/CW-DFB capacity forecast to roughly double in CY2026 and Greensboro six quarters out, what is your assumed supply-demand gap exiting CY2026 and CY2027?

---