Quick ReadPlain-language overview · read this first
Lumentum is a photonic-device and module maker on the AI data-center optical-interconnect chain, at $890.09 with a diluted market cap of $85.6 billion, rated Avoid.
The Components and Systems ends grew 70.4% and 76.8% year over year in the first nine months of FY2026, with TTM revenue of $2.49 billion and net income of $440 million; NVIDIA injected $2 billion at $695.31/share with a multi-year purchase commitment. The moat rests on technology and qualification, with no pricing power; two customers account for 24% and 16% of revenue, FCF is only $114 million, ROIC is 6%-8%, EV/Sales 34x, and P/FCF 750x.
Owner Earnings discounting gives a conservative $60-90, neutral $130-220, and optimistic $250-400, with the current price above the upper edge. Ideal buy $60-120; if AI capex slows on top of ASP declines and dilution, a permanent drawdown of 60%-80% is not unimaginable — a good company, not a good price.
LeadA photonic-device maker on the critical AI data-center optical-interconnect chain, with a clearly improving fundamental picture but historically volatile cash flow; at roughly $890 per share the stock already prices in years of flawless execution. Rating: Avoid.
Prices in the article are as of publication; see the valuation band above for the live price.
This report is written primarily from the company's official filings and verifiable public sources: Lumentum's FY2025 Form 10-K, the Form 10-Q for the period ended March 28, 2026, the 2025 Proxy Statement, the 2023 annual report (used to fill in 2021–2023 historical financials), the 2021/2022 official earnings press releases, and the 8-Ks and joint press releases covering the strategic partnership and preferred-stock financing between the company and NVIDIA; the current share price and some peers' market caps/PEs use real-time market-data tools.
Conclusion First
Investment rating: Avoid. More precisely: the company is worth ongoing observation, but at its current price it is not a fit for "Buffett-style value investing." Lumentum is not a bad company; it sits on an important link in the AI optical-interconnect chain, and its fundamentals have indeed recovered materially of late. But it remains a hard-tech manufacturing business with fast technology iteration, strong customer bargaining power, high capital intensity, and historically very volatile cash flow — and the valuation the market currently assigns it has already, in one shot, pulled forward a highly aggressive future.
Core judgments: First, the high-level logic of this business is understandable, but it is not "simple": at its core it sells high-performance photonic devices, modules, and systems, serving AI data centers, cloud networks, long-haul transmission, and industrial laser applications; the products matter, but both demand and profits are affected by technology roadmaps, capacity cadence, customer inventories, and ASP changes. Second, the company's latest burst of results does show that it is positioned to benefit from the AI optical-interconnect upgrade, with growth on both the components and systems ends simultaneously. Third, there is a moat, but it shows up more in technology, process, patents, and customer qualification than in consumer brand, network effects, or absolute pricing power. Fourth, capital allocation is not misguided, but it is not exceptional either: over the years it has done large acquisitions, debt management, buybacks, and now industrial financing — overall "tactically reasonable, but not yet proven to convert into high long-term growth in intrinsic value per share." Fifth, at the current share price, the margin of safety is not merely "not obvious" — it is essentially absent.
Is there a margin of safety at the current price: no. The current real-time price is about $890.09; market-data tools put the market cap at about $85.6bn. This market cap is broadly consistent with the company's diluted share-count basis for fiscal Q3 2026, not merely basic shares outstanding. In other words, the market is already pricing it in a rather "optimistic and fully diluted" way.
Suitable investor type: better suited to growth/thematic investors who can tolerate high-valuation drawdowns and are willing to bet on long-term high growth in AI optical interconnect; not suited to long-term value investors whose core criteria are margin of safety, stable free cash flow, and high predictability.
Greatest uncertainties: First, just how durable AI optical-interconnect demand really is, and whether that demand will be interrupted by customer inventories, architecture shifts, and price competition. Second, whether the new capacity, new fab, and large capex the company is pushing will ultimately generate a high ROIC, or merely buy higher revenue and higher depreciation. Third, per-share cash-flow capability after dilution: the NVIDIA preferred stock, convertible-note conversions, and share exchanges all mean that "the enterprise getting bigger" does not necessarily equal "each share being worth more."
The Business, Industry, and Moat
How this company actually makes money
Fact. In its FY2025 10-K, Lumentum defines itself as a leading supplier of "optical and photonic products," with products aimed at cloud, AI/ML, communications, consumer, and industrial end markets; as of FY2025 the company still reported under two segments, Cloud & Networking and Industrial Tech, but starting in fiscal Q1 2026 it completed a reorganization and now manages itself as a single operating segment. In other words, the high-level business framework is changing, but the underlying way it makes money has not: selling high-end photonic devices, modules, systems, and industrial lasers.
From a product standpoint, Lumentum is currently better understood as a two-layer revenue engine. The first layer is Components: laser chips, laser assemblies, optical circuit subsystems, wavelength-management systems, and so on — these are "critical parts" within a larger set of network equipment. The second layer is Systems: closer to complete functional modules or systems, such as the cloud optical-module product line and optical-circuit-switching products. In the first nine months of FY2026, components revenue was $1,356.2m and systems revenue $651.5m, up 70.4% and 76.8% year over year respectively, with the growth clearly coming from AI data-center intra-connect, data-center interconnect, and long-haul transmission.
On customers, this is a classic B2B business, with primary customers being OEMs, network-equipment makers, cloud/AI infrastructure customers, and some end-service providers. The company discloses clearly: customer concentration is high. In FY2025, one customer accounted for 16.0% of revenue and another for 15.4%, with two more at 12.1% and 10.5%; in the first nine months of FY2026, two customers accounted for 24% and 16% of total revenue. This shows it is not a "tens of thousands of scattered customers" business but one driven by a few large customers.
The way it charges is not complex: the company recognizes revenue as it delivers products such as devices, modules, and systems — in essence still selling hardware and solutions, not subscription software. Inference. This means the "repeat-purchase" nature of revenue comes from customers' continued capacity expansion, product upgrades, and ongoing design-ins, rather than contracts that renew by nature. So its revenue is not entirely one-off, but it is far from highly predictable. Over the past few years revenue went from $1.74bn in 2021, $1.71bn in 2022, and $1.77bn in 2023, to a drop to $1.36bn in 2024, then a recovery to $1.65bn in 2025, and reached $2.01bn in the first nine months of FY2026; that curve alone is enough to show its volatility.
On cost structure, this is a manufacturing/engineering business with substantial fixed costs and technology investment. The company must simultaneously bear the costs of wafer fabs, packaging and testing, equipment depreciation, third-party foundry, R&D, qualification, and quality systems. FY2025 R&D expense was $303.9m, or 18.5% of revenue; SG&A was $348.2m, or 21.2% of revenue; capex that same year reached as high as $231.0m. In the first nine months of FY2026, capex rose further to $284.5m. This is not an "asset-light software" model but a classic "technology + manufacturing" hybrid business.
This business also depends on a few key resources: first, large customers; second, a limited number of suppliers and foundries; third, customer qualification of production lines. In the 10-K the company states clearly that some raw materials and components come from single or limited-source suppliers, and many customers require qualification of a production line before formal ramp; switching plants, moving lines, or bringing in a new outsourced party could affect shipments and profits. The company likewise acknowledges that large customers typically have stronger bargaining power.
Opinion. So, answering "would you be willing to hold if the stock market closed for 5 years" from the perspective of a long-term business owner: at the enterprise level it can be considered for holding; at the stock level it depends on price. This is not an incomprehensible business, but neither is it the kind you can "explain in five minutes and that stays highly stable ten years later." It is more like an "important but not calm" infrastructure-components company. Business comprehensibility score: 3/5.
Industry and competitive landscape
Fact. In its 10-K the company on one hand stresses that its products sit at a critical position in cloud, AI/ML, communications, and industrial manufacturing; on the other hand it also writes clearly that spending on optical-communications and laser products is cyclical and unevenly volatile, that customers go through inventory buildup, destocking, and demand recovery, and that competitive intensity in the industry is high and rising. By the FY2026 10-Q, the company further notes that in recent years customers had pre-stocked because of tight supply chains and then destocked once supply eased, so revenue and profit were noticeably disturbed.
For this reason, this industry is neither a simple "growth industry" nor a "mature industry," but a hybrid of "long-term demand trending up + highly cyclical over the short to medium term." Long-term demand does have support: AI training and inference drive higher bandwidth, lower power, and higher-grade optical interconnect; the industrial side has continued demand for precision lasers from semiconductor manufacturing, microelectronics, EV/battery, and welding/cutting. In March 2026 NVIDIA and Lumentum signed a multi-year, non-exclusive strategic agreement, accompanied by a multi-billion-dollar purchase commitment and future capacity rights — itself a strong signal of the sector's strength.
But this industry is also very easy to disrupt through changes in technology path. In its press release the company describes the partnership as "developing state-of-the-art optical technology, next-generation AI architectures, and system designs"; that in fact conversely shows the products are neither commodities nor a monolith — once the technology path changes, the winners may rotate. The company's risk disclosures repeatedly mention patent challenges, supply chain, export controls, tariffs, and rare earths.
On main competitors, the company does not name each one in the 10-K, but states clearly that its current competitors include both public and private companies, many with stronger financial, technical, marketing, and distribution resources than its own. Among publicly comparable names, Coherent is one of the strongest integrated comparables; in addition, Applied Optoelectronics, nLIGHT, Fabrinet, and Ciena offer reference points across different sub-segments, though comparability is imperfect. On current market data, Coherent has a market cap of about $69.2bn and a trailing PE of about 167.6x; Fabrinet a market cap of about $24.6bn and a PE of about 58.3x; AAOI and nLIGHT are currently loss-making, so their PEs carry no reference value. Lumentum's current price and valuation show no obvious cheapness relative to peers.
Opinion. This is not "an excellent company in a bad industry," because the underlying demand matters a great deal; nor is it "an easy winner in a good industry," because competition, technology, and capex are all heavy. It is more like: a hard business inside good demand. Industry attractiveness score: 3/5.
Moat analysis
Going item by item through a Buffett-style framework:
Brand advantage: weak to moderate. This is a brand known to engineers, procurement staff, and system architects, not a consumer brand. It has a reputation within the industry, but the brand alone is not enough to generate excess profit on its own.
Cost advantage: moderate. The company emphasizes technology leadership, cost leadership, functionality, and vertical integration, and owns its wafer fabs, packaging and testing, and a global manufacturing footprint; but it also admits that customers and some competitors have strong resources and that its supply chain is constrained, so the cost advantage is not the kind of insurmountable "king of low cost."
Scale advantage: moderate. In photonic devices and certain sub-segment products, scale and manufacturing experience translate into yield, qualification speed, and delivery capability, but the industry still has several strong rivals.
Network effects: essentially none. This is not a platform model.
Switching costs: moderate to high. Customers must qualify production lines, and many products go through design-in, validation, and production qualification, meaning that once a supplier is in a large customer's supply chain, replacement is not as simple as swapping an ordinary component. The company discloses clearly that if a new manufacturing line fails a customer's qualification, the business would be harmed; that in itself shows the existing qualified lines carry some stickiness.
Channel advantage: weak. It relies mainly on direct customer relationships and design-ins; there is no high barrier akin to a consumer-goods distribution network.
Patents, licenses, and technical barriers: moderate. As of June 28, 2025, the company held roughly 1,020 U.S. patents and 1,100 foreign patents, plus about 780 patent applications pending worldwide. This shows the technical accumulation is real; but the company also cautions that patents may not be enough to stop competitors, and that intellectual-property disputes are frequent in the industry.
Data advantage: weak. This is not a data-driven monopoly.
Corporate culture or operating capability: moderate. Being able to run through complex manufacturing, overseas footprint, M&A integration, customer qualification, and the recent AI-demand ramp shows the operating capability is not poor; but the big swings in profit over the past few years also show this capability is not yet enough to fully smooth the cycle.
Capital-allocation capability: moderate to weak. Management has taken some not-foolish actions, such as early buybacks at low prices, debt refinancing, and bringing in NVIDIA's industrial capital and purchase commitment; but the large acquisitions (NeoPhotonics, Cloud Light) have not so far clearly proven to have converted into stable long-term high-return capital.
Inference. This moat is more a combined "technology + process + qualification" barrier than a "Coca-Cola-style brand" or a "Microsoft-style ecosystem lock-in." Amid the AI optical-interconnect wave it may widen for a phase, but it is hard to say it has already formed a durable advantage that "needs no further proof ten years from now." Replicating it by competitors cannot be done in a quarter or two — it often takes years of R&D, customer qualification, and capital investment; but it is not un-replicable either.
On pricing, inflation, and recession: in the first nine months of FY2026 the company did disclose a slight ASP increase in laser-chip products within the components business, mainly tied to the 200G-lane speed upgrade; but the systems products also saw higher shipment volumes accompanied by lower ASP. Over 2023–2025 the company faced pressured operating profit for several consecutive years, even losses, which proves it cannot steadily maintain high profitability through downturns. So within its past high margins there is both structural technology value and a clear mix of cyclicality, product mix, and up-cycle dividend. Moat strength score: 3/5.
Management and Capital Allocation
Is management trustworthy
Fact. The company completed a CEO transition in 2025: in the proxy the board explicitly thanks Alan Lowe for nearly a decade of leadership, and Michael Hurlston becomes the new CEO. On governance structure, the company has stock-ownership requirements for directors and executives: the CEO must hold stock worth 5 times base salary, other executives 2 times, and directors 5 times the cash director fee. The direction of these rules is right.
But looking at actual ownership, management's direct holdings are not high. The 2025 proxy shows that as of August 29, 2025, Michael Hurlston's direct beneficial ownership was 0, and all directors and executives combined held only 307,933 shares, less than 1% of the company's outstanding common stock. This means the alignment of interest between management and shareholders relies more on incentive plans than on founder-style concentrated holdings. For a value investor this can usually only be counted as "acceptable," hardly "exceptional."
Fact and opinion combined. My judgment on management's integrity is: there is currently no hard evidence of accounting fraud, major related-party transactions, or clearly misleading disclosure; in 2025 Deloitte gave unqualified opinions on both the financial statements and internal controls. At the same time, the company also disclosed in the proxy a small number of Section 16 filings that were late due to administrative errors, but the nature is minor and does not constitute a core governance red flag. Overall, integrity can be rated neutral-to-positive for now; shareholder alignment is only average.
Is capital allocation rational
The main uses of the company's cash over the past few years fall roughly into five categories: M&A, capex, buybacks, debt-structure management, and equity incentives. On M&A, in 2022 it acquired NeoPhotonics for $934.4m in cash, and in 2023 completed the Cloud Light acquisition for total consideration of $728.5m, of which about $705.0m was cash; Cloud Light also brought $365.8m of goodwill. The M&A logic is clear: to strengthen high-speed optical devices, optical modules, and data-center capability. The problem is not the logic but the return: as of FY2025 the company had only just returned to a thin profit, and ROIC is far from "excellent." So the ultimate value creation from M&A can today still only be marked "incomplete."
On buybacks, historically the company has not been entirely price-insensitive. In fiscal Q4 2021 it used $241m to buy back 3.1 million shares; in FY2022 it bought back a combined 6 million shares for about $548.9m; the 2023 cash-flow statement also shows $175.6m of buybacks. At least by results, these buybacks occurred before this round of valuation surge — far more rational than "buying back at the top" today. In 2024–2025 it did not continue buybacks, instead directing funds more toward M&A, capacity expansion, and debt management.
The most important capital-allocation move of 2026 is the industrial-partnership financing with NVIDIA. In March 2026 the company privately placed about 2,876,400 shares of Series A convertible preferred stock to NVIDIA, raising $2.0bn at an issue price of $695.31/share; at the same time the two signed a multi-year, non-exclusive strategic agreement, including a multi-billion purchase commitment and rights to future advanced-laser-component capacity. At the operating level this is a major positive: it raises the funding safety cushion and materially reduces demand uncertainty for the new-fab expansion. At the existing-shareholder level it is not costless — it introduces potential dilution, so whether it "truly creates per-share value" depends on whether future returns can exceed the cost of the new capital. My assessment: this step is industrially smart and financially neutral-to-positive, but by no means an unconditional positive for per-share value.
On compensation, management is a mixed bag. On the good side, in 2025 the compensation committee exercised a negative discretionary adjustment on the AIP outcome, cutting the originally calculated payout ratio of 181.5% down to 132.2%, on the grounds of better reflecting operating performance; this practice is usually more shareholder-minded than mechanically paying bonuses. On the other side, the new CEO's package is sizable, including high-value RSUs/PSUs; although the company explains that a substantial portion is used to compensate for equity forfeited at his prior employer, and that the long-term PSUs use a relative-TSR constraint, it is still a non-negligible dilution cost for long-term shareholders. Management and capital allocation score: 3/5.
Financial Quality and Owner Earnings
Key financial metrics
The table below tries to lay out, side by side, the key data needed to "look at the business" rather than "look at the mood." The 2021–2023 data are drawn mainly from the 2023 annual report, 2024–2025 from the FY2025 10-K, and the first nine months of FY2026 plus the latest balance sheet from the March 2026 quarterly report.
| Metric | FY2021 | FY2022 | FY2023 | FY2024 | FY2025 | 9M FY2026 | TTM through Q3 2026 |
|---|---|---|---|---|---|---|---|
| Revenue ($m) | 1,742.8 | 1,712.6 | 1,767.0 | 1,359.2 | 1,645.0 | 2,007.7 | 2,488.4 |
| Gross margin | 44.9% | 46.0% | 32.2% | 18.5% | 28.0% | 38.8% | ~37.7% |
| Operating margin | 30.2% | 17.7% | -6.5% | -31.9% | -10.9% | 12.2% | ~9.5% |
| Net income ($m) | 397.3 | 198.9 | -131.6 | -546.5 | 25.9 | 226.6 | ~439.9 |
| Operating cash flow (CFO) ($m) | 738.7 | 459.3 | 179.8 | 24.7 | 126.3 | 388.4 | ~452.4 |
| Capex ($m) | 84.8 | 91.2 | 128.5 | 133.0 | 231.0 | 284.5 | ~338.4 |
| Free cash flow (FCF) ($m) | 653.9 | 368.1 | 51.3 | -108.3 | -104.7 | 103.9 | ~114.0 |
Note. FY2021's CFO/FCF are elevated, affected by the merger-termination fee Coherent paid the company; in the 2023 annual report management explicitly notes that part of the 2021 operating cash flow came from the net amount of that fee. That is, 2021 cannot be treated as a "normalized cash-flow baseline."
The most important message in this table is not "growth is fast" but: this is a company with historically unstable cash flow, extreme margin elasticity, and only a recent recovery driven by AI demand. For three consecutive years in 2023–2025 operating profit was very weak, and 2024 even saw a net loss of $546.5m; only in FY2026 did it clearly return to a high-growth, high-margin zone. For a value investor, this means you can hardly capitalize a single year's profit directly.
On the authenticity of cash profits, I do not think the company currently shows clear evidence of "accounting fraud": the audit opinion is unqualified, and in 2025 Deloitte listed inventory valuation as a critical audit matter, rather than revenue recognition or cash anomalies; this looks more like a valuation difficulty common in cyclical manufacturing than a direct fraud signal. At the same time, the company's book profit is mixed with a large amount of acquisition-related intangible amortization, gains on facility disposals, tax swings, and restructuring charges, so GAAP net income cannot simply be equated with true distributable cash.
Balance sheet, returns, and cash-flow quality
As of March 28, 2026, the company held $2,617.8m in cash and $554.5m in short-term investments, totaling $3,172.3m; book debt was $3,281.8m, for net debt of about $109.5m. This means that although total debt on the balance sheet is not low, after the NVIDIA capital injection the company's net leverage is actually very light. The company also has a $400m revolving credit facility, undrawn as of that date. The current balance sheet is not fragile.
But you have to read the other side too: the sharp increase in "current liabilities" in the quarterly report does not fully represent an operating crisis, but mainly reflects that the 2026, 2028, 2029, and 2032 convertible notes, under the current share-price conditions, became convertible or were reclassified to current liabilities. The company repeatedly explains in the 10-Q that these notes remain convertible because the share price exceeded a certain threshold above the respective conversion prices, and so are listed as current liabilities. For that reason, looking at the "current ratio" alone would misjudge the risk.
On returns, the company's "present" is far better than "the average of its earlier years," but it is still far from proven to be excellent over the long run. Roughly estimated on a TTM basis through Q3 2026, TTM EBIT is about $237m, while invested capital including net debt is around $3.1bn, corresponding to an ROIC of only mid-to-high single digits, roughly in the 6%–8% range; in FY2025 there was hardly any "excellent return" to speak of. In other words, the company is getting better, but it is not yet good enough to deserve its current market cap.
On working capital, growth is eating cash. In the first nine months of FY2026, accounts receivable rose $191.5m and inventory rose $165.2m; although accounts payable rose $109.4m, providing some hedge, overall it still ties up cash significantly. This matches the AI-demand ramp and capacity expansion, but it also shows: the company is not the kind of model where "the more it grows, the easier it gets," but rather "the more it grows, the more it first has to sink money into inventory, equipment, and customer receivables."
Changes in share count matter equally. At the August 2025 record date there were about 70.8m common shares; by April 30, 2026, the company's common shares had increased to 77.8m, while the diluted weighted-average share count for fiscal Q3 2026 reached 96.2m. This includes potential conversion of the NVIDIA preferred stock, as well as share-exchange transactions the company conducted with some convertible-note holders in April 2026. For a value investor, this means any future "per-share value" judgment must use a more prudent diluted basis.
Owner Earnings analysis
Here I deliberately adopt a conservative basis, avoiding adding back all non-cash charges wholesale, which would make it "look very cheap."
Fact. On a TTM basis through Q3 2026, roughly:
TTM net income about $440m;
TTM operating cash flow about $452m;
TTM capex about $338m;
TTM free cash flow only about $114m;
TTM stock-based compensation, as a cash-flow adjustment item, about $169m;
the company is currently in a new-fab/new-capacity investment phase, and the joint statement by management and NVIDIA explicitly mentions using the funds to support R&D, future capacity, and U.S. manufacturing capability build-out.
Assumption. The company does not separately disclose maintenance capex, so it can only be estimated. My conservative assumption is: in the current heavy-expansion phase, a substantial part of total capex is growth investment, so maintenance capex can be estimated at $140m–$180m/year, roughly slightly above depreciation but clearly below current total capex. This range is not aggressive, because equipment renewal, yield maintenance, and line-qualification upkeep in photonic-device manufacturing are inherently not low-maintenance.
Inference. If, from the standpoint of "cash truly distributable to shareholders" and treating stock-based compensation as a real economic cost, a relatively conservative Owner Earnings calculation can be written as: Owner Earnings ≈ CFO – maintenance capex – the economic cost of stock-based compensation. On this basis, TTM Owner Earnings is only roughly $120m–$170m. If you are more lenient on stock-based compensation — deducting only maintenance capex and not treating SBC in full as a cash-equivalent cost — then Owner Earnings can rise to $250m–$320m. In the later valuation I use the more lenient assumption near the midpoint, which is already granting the company a "growth premium."
Opinion. This step almost already delivers the conclusion: whether you take the conservative or the more lenient basis, the Owner Earnings multiple implied by the current share price is extremely high. Dividing the roughly $85.6bn diluted market cap given by the market-data tool by $120m–$320m of Owner Earnings implies the market is pricing this company at roughly 270x–710x Owner Earnings. For any investor seeking a margin of safety, this is not an acceptable buy range.
Valuation, Margin of Safety, and Opportunity Cost
Current valuation and intrinsic value under three methods
Fact. The current price is about $890.09. If we use the market-data tool's diluted market cap of about $85.6bn, plus net debt of about $0.1bn through Q3 2026, enterprise value (EV) is also roughly $85.7bn. Compared with TTM financials:
EV / Sales ≈ 34x–35x;
P / TTM GAAP net income ≈ 190x–195x;
EV / TTM EBITDA (including acquisition amortization) ≈ 170x+;
P / TTM FCF ≈ 750x;
P / Book ≈ 23x–29x, depending on whether you use basic shares or the more conservative diluted basis. Unless you assume Lumentum will become an AI core platform with an ultra-high moat, ultra-high cash returns, and almost no competition, this set of numbers is very hard to justify.
Method one: Owner Earnings discounting. I use not "management's optimistic non-GAAP profit," but the more conservative Owner Earnings starting point above. The following are the three scenarios I consider barely self-consistent, all of which already leave the company considerable room for growth:
| Scenario | Starting Owner Earnings | Next-10-year growth | Discount rate | Terminal growth | Implied intrinsic value per share |
|---|---|---|---|---|---|
| Conservative | $300m | 8% | 11% | 3.0% | ~$60–90 |
| Neutral | $550m | 10% | 10% | 3.5% | ~$130–220 |
| Optimistic | $900m | 12% | 9.5% | 4.0% | ~$250–400 |
Note. These ranges already imply: the company continues to capture the AI optical-interconnect dividend in the coming years, margins keep recovering, maintenance capex can be partly diluted by expansion capex, and there is ultimately no drastic competitive price war. Even so, the current $890 is still clearly above the upper edge of my optimistic range. Conservative intrinsic-value range: $60–90. Fair intrinsic-value range: $130–220. Optimistic intrinsic-value range: $250–400.
Method two: relative valuation. Compared with comparable companies, Lumentum is not cheap either. Coherent's current trailing PE is about 167.6x, with a market cap of about $69.2bn; Fabrinet's about 58.3x PE and $24.6bn market cap; AAOI and nLIGHT are currently loss-making, so their PEs are distorted. Even within this peer group already pushed up as a whole by the AI optical-interconnect theme, Lumentum's valuation is not "clearly low." Add in Lumentum's unstable historical FCF, current large capacity expansion, and dilution factors, and there is no logic of "peers are all expensive, but it is relatively cheap."
Method three: asset/liquidation value. As of Q3 2026, the company's total shareholders' equity was $2.97bn, of which goodwill was $1.07bn and other intangibles $0.36bn. In other words, about $1.43bn of book equity is goodwill and intangibles; on a more conservative tangible-net-worth basis, only about $1.5bn remains. Even without applying any liquidation discount, the "asset floor" this forms for common stock is only within tens of dollars per share, not hundreds of dollars. For someone who invests using a "margin of safety," this kind of asset approach cannot support the current price.
Margin of safety and opportunity cost
Margin-of-safety conclusion: insufficient, essentially absent. The most fragile assumption in the valuation is how much of the next decade's cash flow the market has pre-paid. To earn a 10% annualized return at today's $890 price, the share price ten years out would roughly have to exceed $2,300; to earn 12%–15% annualized, an even higher terminal value would be needed. That would mean the company must convert cash flow that is still unstable today into very large and highly sustainable Owner Earnings per share. This is not impossible, but it far exceeds the definition of a "conservative buy."
If growth falls short of expectations, or margins pull back, the investment logic wobbles almost immediately. History has already provided the sample: in 2023–2025 the company experienced pressured revenue and a marked decline in gross and operating margins; and the current growth itself is also consuming liquidity and capex. That is, this company is not one where "it's fine even if things are a bit worse" — rather, it is one where "only continued high prosperity, high utilization, and high execution can barely justify the current valuation."
Opinion. This is a very typical "good company, possibly a bad price." If you are a value investor, I think the most reasonable action is not to "chase for fear of missing out," but to wait patiently for a better price, or wait for a longer stretch of verifiable high ROIC / high FCF to materialize. At the current valuation, what the market offers you is not a margin of safety but the pressure of "having to keep being perfect."
From an opportunity-cost standpoint, I would not put Lumentum ahead of an S&P 500 ETF, still less ahead of high-grade bonds/risk-free yields. SPY's current price is about $733.73 — that is not valuation itself, of course; but as an investment choice, a broadly diversified index does not require you to make such strong optimistic assumptions about a single company, a single technology path, a single customer relationship, and a single expansion cycle. If I could hold only 5 assets, LITE at its current price does not qualify for the portfolio.
Ideal buy-price range: $60–120. Acceptable holding-price range: $120–220. Clearly overvalued price range: above $350–400. This is based on conservative-to-optimistic Owner Earnings scenarios, not on short-term momentum.
Risks, Checklist, and Final Conclusion
The most important risks and the counterview
The most important risk is not short-term volatility, but permanent loss of capital. For Lumentum, the most core risks include: competitive risk, technology-substitution risk, customer-concentration risk, supply-chain and export-control risk, capex returns falling short of expectations, overvaluation risk, dilution risk, and declines in capacity utilization and gross margin from a downturn in industry conditions. In its 10-K/10-Q the company gives clear disclosure of supplier concentration, customer concentration, line qualification, trade restrictions, and the impact of rare earths and tariffs.
The strongest counterview is actually straightforward: Lumentum now looks like an "AI optical-interconnect must-own," but at its core it is still a hard-tech manufacturer, not a software platform with a monopolistic ecosystem. It has to keep investing in equipment, inventory, and R&D, its customers are highly concentrated, and product cycles and technology paths iterate quickly. NVIDIA's investment and purchase commitment are certainly important, but the agreement is non-exclusive, which shows the supply chain is not closed to leave only one winner. If competition intensifies in the future, customers diversify sourcing, or AI-cluster capex slows in cadence, Lumentum's margins and valuation could both pull back at the same time.
What facts, if they appeared, would make me admit today's "avoid" call is wrong? If over the next two to three years the company can — without significantly further diluting shareholders — raise annualized Owner Earnings to above $1bn, and steadily lift ROIC to around 15% or higher, while proving that the new fab and capacity expansion are not a one-off up-cycle stocking but a sustainable long-term high-return capability, then today's "too expensive" judgment would need to be overturned. In other words, I am not denying it could become a better company; what I deny is that the current price has already bought away too much of the future in one shot.
The biggest permanent-loss-of-capital scenario is: AI optical-interconnect demand slows in 1–3 years, customers begin to digest inventory, utilization of new capacity falls short of expectations, ASPs decline, share dilution keeps rising, and the market simultaneously re-prices it from a "scarce AI optical asset" back to a "high-volatility hardware supplier." For a stock at a current EV/Sales of about 34x or more, valuation contraction alone could cause a huge drawdown, even if the business itself does not collapse.
Investment checklist
| Checklist item | Conclusion |
|---|---|
| Can I understand this business? | Pass |
| Does it have stable long-term demand? | Pass |
| Does it have a durable moat? | Uncertain |
| Does it have pricing power? | Fail / Uncertain |
| Can it generate stable free cash flow? | Fail |
| Is its return on capital excellent? | Fail |
| Is management trustworthy? | Uncertain |
| Is capital allocation rational? | Uncertain |
| Is the balance sheet sound? | Pass |
| Is valuation below intrinsic value? | Fail |
| Is the margin of safety sufficient? | Fail |
| Does holding long-term let me rest easy? | Fail |
| Which key facts would make me sell? | Worsening customer concentration, gross margin turning down, capex returns missing targets, continued heavy share dilution, weakening of the NVIDIA partnership |
| Am I wanting to buy just because the price rose or on market sentiment? | Must be highly vigilant |
The above judgments are based on the preceding business, financial, valuation, and governance analysis, especially customer/supply-chain concentration, cash-flow volatility, recent dilution, and the current extremely high valuation.
Final investment conclusion
【Final Rating】 Avoid
【One-Sentence Investment Thesis】 Lumentum may become an important beneficiary of AI optical interconnect, but its current share price has run far ahead of its "verifiable, distributable, sustainable" cash-flow capability.
【Core Bull Case】
It sits on a critical link in the AI data-center optical-interconnect upgrade, with both components and systems revenue posting strong growth in the first nine months of FY2026.
NVIDIA not only invested $2bn but also provides a multi-year purchase commitment and future capacity rights, strengthening demand visibility.
The company has some technology and process barriers, a broad patent portfolio, and customer-qualification thresholds.
The balance sheet strengthened materially in 2026, net leverage is light, and financing risk has fallen.
2026 margins and cash flow have clearly recovered relative to 2024–2025.
【Core Bear Case】
Historical free cash flow is very unstable, negative in both 2024 and 2025, and the 2026 improvement has not yet been tested across a full cycle.
Customer concentration is high, supply-chain dependence is strong, and customers have strong bargaining power; it has no absolute pricing power.
Current valuation is extremely high, virtually requiring continued flawless execution for years.
Dilution risk is real: preferred stock, convertible notes, and share-exchange transactions are all expanding the potential share base.
Capital allocation has logic, but the M&A and expansion have not yet proven they can sustainably create high per-share returns.
【Key Assumptions】 For this investment to hold, at minimum the following must be satisfied: AI optical-interconnect demand stays highly prosperous for at least several years; Lumentum's new fab and added capacity generate high returns; the company converts growth into free cash flow per share, not merely revenue expansion; and the NVIDIA partnership delivers long-term platform standing, not just a phase of stocking.
【Fair Buy Price】 I believe a more reasonable buy range is $60–120; if one must give a range that is "barely acceptable to hold," it is about $120–220. The basis is a conservative-to-optimistic Owner Earnings discounting, not the current theme's heat.
【Target Holding Period】 If valuation returns to a reasonable range in the future, and the company proves that this new round of growth carries sustainable high ROIC, I would treat it as a candidate to track or hold for at least 5–10 years; but at the current price, I do not think "long-term" can automatically make up for "buying it too expensive."
【Expected Annualized Return】 Buying at the current price:
Conservative scenario: -16% to -10%/year
Neutral scenario: -8% to -2%/year
Optimistic scenario: 0% to +6%/year These are not short-term price forecasts but ranges back-solved from the "intrinsic value per share that could be realized over the next decade." To achieve double-digit annualized returns at today's price, the company would need to deliver Owner Earnings far above my neutral scenario.
【Maximum Loss Risk】 If the pace of AI build-out slows, expansion returns fall short of expectations, customers press on price, and valuation reverts to a more normal range, a long-term drawdown of 60%–80% or more is not unimaginable. The reason is not only declining earnings but, even more, that today's valuation starting point is too high.
【Tracking Metrics】 Going forward I will focus on:
whether components and systems revenue growth stays consistently above the industry;
whether gross margin can stabilize around 40% or higher;
whether operating margin can stably move into the mid-to-high double digits;
whether operating and free cash flow keep turning positive and expanding;
whether maintenance and growth capex begin to diverge, and whether total capex intensity comes down;
whether changes in receivables, inventory, and payables show improving growth quality;
whether the diluted share base keeps expanding significantly;
whether customer concentration rises further;
whether the NVIDIA partnership brings higher visibility rather than just a short-term volume surge;
whether ROIC truly rises into the range of excellent companies.
【Signals to Trigger Reassessment】 Any of the following should prompt a fresh review of the investment logic:
revenue still growing strongly for several consecutive quarters, but FCF turning significantly negative again;
gross margin and operating margin turning down;
the new fab and expansion causing inventory and depreciation to surge while returns stay insignificant;
the diluted share base continuing to expand rapidly;
the multi-year purchase commitment weakening, or customer destocking recurring;
export controls, tariffs, rare earths, and key-component supply again severely disrupting costs and delivery.
【Final Recommendation】 To put it plainly: Lumentum as a company is worth tracking; Lumentum as a stock at this price is not worth buying. If you are a long-term value investor, what you really want to own is not "the AI narrative" but "verifiable cash flow per share + a cheap enough buy price." Lumentum currently only meets part of the former's expectations and is far from meeting the latter. The best action is not to chase, but to put it on a watchlist and wait for a cheaper price, or for more cash flow and returns to materialize before judging.
Open questions and limitations. To be candid: the company does not publicly disclose the names of its major customers; maintenance capex is also not separately disclosed, so Owner Earnings can only be a conservative estimate; in addition, because peers' business structures differ substantially and some peers are currently loss-making, I have not laid out all comparables' PB, EV/EBITDA, P/FCF, and ROIC in full, but focused on the most representative comparables and Lumentum's own diluted-basis valuation. These limitations affect the "precision" of the valuation, but they do not change my overall judgment that the current price lacks a margin of safety.
This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.
Full report
Sign in to read the full report
Sign up free to unlock the full text, the Baillie growth scorecard, and full-text search.
Log in / Sign up free