LightPath Technologies

IThe business

LightPath makes the lenses and glass inside thermal imaging systems, and since 2023 it also makes the cameras. A thermal camera sees heat instead of light. Soldiers use them to see at night, border agents to spot people crossing, firefighters to see through smoke, and operators of counter-drone systems to track small aircraft. The lens that focuses the heat has to be made of a material that infrared passes through, and the usual one is germanium, a rare metal.² The company is based in Orlando, Florida, was formed in 1992 and has 425 employees: 214 in the United States, 107 in Riga, Latvia and 104 in Zhenjiang, China.² Its shares trade on the Nasdaq Capital Market under LPTH.

Two facts in the brief I was given turn out to be out of date, and I want to correct them at the start. First, revenue is not $30M to $50M any more. It was $37.2M in fiscal 2025 and $71.7M in fiscal 2026, up 93%, because a full year of G5 Infrared, bought in February 2025, was added C (calculated).⁸² Second, China is leaving. LightPath agreed on 23 July 2026 to divest its Chinese subsidiary, which holds the factory in Zhenjiang, for $4.5M paid over five years to a company owned by the factory’s management and others; the 26 October proxy says the sale is expected to close before 30 November, and I found no filing that says it has.⁴⁵⁹ The company’s Shanghai subsidiary was merged into the Zhenjiang one in 2023, so there is only one Chinese entity.² ISP Optics, bought in December 2016 for $11.8M of net cash and some stock, is not Chinese. Its plants are in Orlando and Riga, Latvia.⁷²

How it earns

The company reports four kinds of revenue. The table below puts them side by side. Infrared and visible components are the old business: lenses sold in quantity to other makers of optical systems. Assemblies and modules is the new one, where LightPath sells a finished lens assembly or a whole camera, at a price that is tens of thousands of dollars instead of single dollars.² In fiscal 2026 assemblies and modules grew from $8.4M to $31.9M, and $23.1M of that increase came from G5 Infrared R (reported).²

The growth needs to be looked at carefully, because it was bought. Of the $34.5M rise in revenue, $24.4M came from G5 Infrared having a full year instead of under five months, and $2.2M from Amorphous Materials, bought in January 2026. The other $8.0M, about 21% on last year’s base, is growth in the rest of the business C (calculated).² That is still good, but it is not 93%. For comparison, revenue was $38.5M in fiscal 2021 and $31.7M in fiscal 2024, so for six years before G5 the company did not grow at all C (calculated).¹⁰¹¹

Who buys

One customer was 24% of fiscal 2026 revenue, up from 9% a year earlier, and the three largest were 37%.² The 10-K describes the orders behind this as coming from “a leading global technology customer” for advanced infrared camera systems, about $58.0M of the $73.6M rise in backlog, shipping in calendar 2026 and 2027.² The company has not named it, and I will not guess. The September release adds an $11M follow-on camera order from the same customer for counter-drone use, and $13M of follow-on lens-assembly orders from a counter-drone and defense systems supplier.³ Visimid has a design contract with Lockheed Martin, and management expects that engineering revenue to fall as the programme moves into production.² By ship-to address 48% of revenue went outside the United States in fiscal 2026, almost all to Europe and Asia, though one large customer assembles in Europe and ships to U.S. end users, so the true foreign share is lower.²

What matters to a minority shareholder is that this is a small, lumpy business where a single order can change a year. Backlog was $110.9M at 30 June, 1.5 times the year’s revenue, and $85.6M of it is requested for delivery within a year.² That is good visibility for about twelve months. Backlog was flat in the June quarter, after rising 130%, 14% and 13% in the three before C (calculated).²

How it got here

For ten years the company was a small molder of lenses that lost money slowly. It bought ISP Optics in December 2016 to add infrared, and revenue went from $17.3M in fiscal 2016 to $28.4M in 2017.⁷ Then it stood still. Sam Rubin became chief executive in March 2020 and set a strategy of moving from components to assemblies and cameras, which the company calls “LightPath 2.0 and 3.0”.²⁵ Visimid, an engineering firm in Plano, Texas, came in July 2023. G5 Infrared, a maker of cooled long-range cameras in New Hampshire, came in February 2025 for consideration valued at $27.1M, with earnouts of up to $23.0M more.² Amorphous Materials, a glass maker in Garland, Texas, came in January 2026 for a preliminary $9.2M.² The earnout for G5 was settled early: $9.0M, $6.3M in cash and $2.7M in stock, due by 15 January 2027.² The company had $4.9M of cash at 30 June 2025 and had used $8.3M in operations that year. It raised $120.2M of equity in fiscal 2026 and ended with $93.2M.⁸² I come back to that, because the difference between the two dates is the whole story of the shares.

IIThe moat

What keeps a rival from earning what LightPath earns? I look at the glass, the lenses and the customer, and I find something real in the first, less in the second and a good deal still to prove in the third.

The germanium argument. Germanium is the standard infrared lens material. The 10-K says that over 60% of the world’s supply comes from China, that Russia was the second source until the war in Ukraine, and that the price has gone from around $1,000 a kilogram a few years ago to over $6,000.² China began requiring export licences in July 2023. In December 2024 it announced it would generally not license germanium and gallium exports to the United States and banned dual-use exports to U.S. military users. In November 2025 it suspended the U.S.-specific restriction until 27 November 2026, but left the licence rules and the military-user ban in place, and the 10-K says China may reinstate or change the restriction.² So the supply argument is live, and it has a date: 27 November 2026, three weeks from now.

LightPath’s answer is chalcogenide glass, a family of glasses made from sulphur, selenium and other elements that transmit infrared. It melts its own, under the name BlackDiamond. BD6, the workhorse, is described by the company as a lower-cost alternative to germanium, and it can be molded to shape instead of ground or turned, which is cheaper at volume.² I want to be exact about this. The 10-K says BD6 and BD2 are generic, produced by at least two other companies.² Glass that anyone can make protects nothing. The company’s own claim is about what it calls the BDNL glasses, which it holds under an exclusive licence from the U.S. Naval Research Laboratory signed in December 2021, with the first, BDNL-4, sold from April 2024.² Those have a property germanium lacks: their focus does not drift with temperature, which lets one lens do the work of several in a camera that must run in the desert and at altitude.² I cannot test that claim. I can say that it is the sort of advantage that is hard to copy, because the licence is exclusive and the know-how is trade secret, not patent.² The 10-K says plainly that the company does not seek patent protection for all of its formulas and processes.²

Switching costs. A change of lens material means a customer must requalify its design. The company says that takes several months for commercial systems and up to two years for complex defense and airborne ones, and the CEO puts supplier qualification cycles at two to three years.²³ That cuts both ways. It protects a supplier once it is designed in, and it slows the wins. The 2030 deadline, set by legislation enacted in December 2025 that directs the Defense Department to eliminate reliance on “covered nations” for optical glass and optical systems, is according to the company the reason sourcing decisions are being made now.³ I take the company’s description of the law as given and have not read the statute. The law names covered nations. It does not say chalcogenide, and a programme could instead buy germanium from non-Chinese sources, if there are any at the quantity it needs. I cannot judge how many programmes will take which route, and I say so.

Pricing power. Gross margin was 36.0% in fiscal 2026 against 27.2% in 2025, and 39.4% in the June quarter.²³ The CEO says the rise came from mix and throughput and not from price increases, and that the yield problems of fiscal 2025 are behind it.³ That is the right kind of improvement. It is also a young one. Before that, gross margin was 52% in fiscal 2017, 35% to 40% from 2018 to 2021, 33% in 2022 and 2023, and 27% in 2024 and 2025 C (calculated).^{7121314101516118} What would show real pricing power is the margin holding or rising as cameras, which are priced against well-funded competitors, become most of the revenue.

Who else is in the room. The 10-K names Ophir (a MKS Instruments company), Controp, Teledyne FLIR and Clear Align as sellers of standard products, and Lockheed Martin, Anduril, Teledyne FLIR, RTX and Elbit as customers that also design optical assemblies in-house.² Some are both customer and competitor. Teledyne FLIR belongs to a far larger group. A company with $71.7M of revenue selling cameras into that group is not protected by size.

My view: the protection is a position, not a wall. It is strongest where the customer cannot get germanium or cannot get it from an approved country, and weakest where a rival can buy generic glass or a customer can make its own. It is a better position than the company had in 2022, and I do not think the numbers yet show how deep it is.

IIIThe competitive advantage

The advantage the company claims is a short, controlled chain from raw material to camera. It melts glass in Orlando and now Garland, molds and turns lenses in Orlando and Riga, coats them in Orlando, Riga and Hudson, and builds cameras in Plano and Hudson.² The September release says it has two domestic glass sites and roughly 20 proprietary infrared compositions, and that adding Amorphous Materials roughly doubled its melting capacity.³ That is a lot of steps for a company with $72M of revenue to run itself, and it is the pitch: a customer who wants one accountable supplier can have it. This is the third product group, assemblies and modules, and it is where gross margin went up.

The second advantage is a clean origin. The sale of the Chinese plant makes LightPath, in its words, a “fully Western-aligned” manufacturer of NDAA-compliant optics.⁴ I think that matters for defense buyers, and it costs something. The plant made about $4.5M a year of sales to outside customers, and LightPath will keep buying product from it for five years at cost plus 10%, with the markup rising 5 points for each $1.0M of the price the buyer pays, to cost plus 30% when the price is paid in full.⁴² So the China exposure does not end. It changes from owning a factory to depending on a supplier whose price rises as the supplier pays LightPath. The buyer is owned by the factory’s manager and other members of management, and the 10-K says that if the supplier is disrupted it could hurt LightPath’s supply chain.² Visible lenses and some commercial products come from there, and tariffs on Chinese goods would apply to what LightPath imports, which I come back to below.

Third, funding. LightPath has no debt of note: a loan of about $0.2M in Latvia and finance leases of about $0.7M, against $93.2M of cash at 30 June.² It has no bank revolver listed in the latest 10-K; a small one of up to $2M was in place in 2021.¹⁰² That gives it room to build glass capacity and to wait for defense qualifications. It is an advantage over the company’s own past, when it could not afford to take on a large order. It is not an advantage over a competitor like Teledyne, which has far more.

Tariffs and Chinese content

The 10-K carries a long tariff risk factor, and it says the company uses unspecified “strategies” to reduce the effect. It does not give a dollar figure for tariffs paid.² From what I can see the exposure comes in three ways. Goods made in Zhenjiang and imported by LightPath into the United States are subject to whatever tariff applies on Chinese goods, and the supply agreement pays cost plus 10% on top. Germanium bought from China comes under export licences, not tariffs, but it too has risen in price. And a fall in sales to Chinese customers or to customers selling to Chinese end users would hurt. The 10-Q for the March quarter notes the U.S. Supreme Court’s February 2026 ruling against certain emergency-powers tariffs and the government’s plan to use other laws.¹⁷ I cannot size any of this from the filings, and I treat it as a margin risk of a few points on the commercial lines, not as a risk to the defense business, where parts made in the United States, Latvia and New Hampshire dominate.

Latvia is a European Union and NATO member, and Riga has 107 of the company’s 425 employees.² I note that the 10-K says one of its customers integrates in Europe, and Russia’s war in Ukraine is listed among the risks.² I have no way to size Latvia’s share of production from the filings.

IVManagement

Sam Rubin, 52, has been president and chief executive since March 2020 and a director since then. He came from Thorlabs, a private optics company, where he ran imaging systems and founded its China operation.⁵ Albert Miranda is chief financial officer. M. Scott Faris has chaired the board since December 2022. Mark Caylor, a former president of Northrop Grumman’s mission systems sector, joined the board on 9 October 2025 as a director designated by North Run.⁵ The board has eight members, and Thomas Ellis, a co-managing member of North Run, is one of them.⁵ Mr. Rubin also runs the Chinese subsidiary and will go on doing so until the sale closes.⁵ The buyer is owned by Mr. Leo Zheng and, in the 10-K’s words, certain members of the Company’s current management team; the July release calls them the factory’s incumbent management.²⁴ I could not tell from the filings whether any of them is a LightPath executive officer, and the proxy names no executive officers other than Mr. Rubin and Mr. Miranda.⁵

How they are paid

The proxy lists Mr. Rubin’s fiscal 2026 pay at $15.8M, against $0.8M a year earlier. Salary was $0.42M, a cash bonus $0.30M, stock awards $4.0M and option awards $11.1M, the last two being accounting values of grants made in March 2026.⁵ Mr. Miranda’s came to $2.9M.⁵ In March 2026 the board approved a special multi-year package of 375,000 stock units for the CEO and 75,000 for the CFO, and performance options for up to 2,550,000 and 425,000 shares at an exercise price of $10.53, earned over five and three years only if revenue and Adjusted EBITDA reach “significantly higher” levels; the proxy does not give the thresholds, and only a small part vests at the first.⁵ The annual bonus rests on revenue and Adjusted EBITDA targets, and was paid in full for 2026. A yearly award is measured against the Russell MicroCap Index and was also earned.⁵

That design pays management to grow revenue and Adjusted EBITDA, and it pays most if the share price clears $10.53. Adjusted EBITDA leaves out stock pay, which will be larger, and the cost of earnouts; revenue counts acquisitions as growth. A manager paid this way has a reason to buy revenue, and the company has bought three companies in three years. I do not say that is why. I say it is what the pay rewards, and I read the owner cash numbers next to it. The one thing in the design that I like is that most of the new award vests only on big increases, so a result near my base would leave most of the options unearned.

What they own, and what the largest holder did

At 15 October all directors and executive officers as a group beneficially owned 8.09M shares, 10.9%, but 5.96M of those belong to North Run through Mr. Ellis, and 3.02M of those are the Series G preferred on an as-converted basis.⁵ Without North Run the directors and officers own about 2.1M shares, about 3%. Mr. Rubin’s 671,144 shares include 225,000 options and 162,931 restricted shares, about $5.1M at $7.63 C (calculated).⁵ I found one open-market purchase by an officer or director since July 2025: 180 shares by the CEO at $9.695 on 27 March 2026, about $1,700. I found no open-market sale by an officer.⁶ The record is silent in both directions, and that tells me little.

North Run tells me more. It bought the Series G preferred in the February 2025 financing at a conversion price of $2.15, with warrants at $2.58. In 2026 it converted preferred and exercised warrants and sold: 6,875,861 shares between 5 January and 3 June for about $90M, an average of $13.10, including 3,571,400 shares sold alongside the company in the June offering at $14.00.⁶¹⁸ I computed this from the Form 4s. It still holds about 2.9M shares and all 6,492 preferred shares, 8.2% on an as-converted basis, and keeps the right to name a director.⁵ A holder with a board seat that sold at $12 to $14 is not telling me anything about the next year, and I do not read it as a signal about the business. It does mean that part of the supply of shares in the first half of 2026 came from an insider who had bought at $2.15, and that the June offering included 3.57M of its shares.

What they did with capital

Over the ten fiscal years to 2026, operating cash flow was −$2.9M and capital spending $26.7M, so owner cash flow was −$29.6M and net income −$47.2M C (calculated).⁷² Revenue went from $28.4M to $71.7M, 2.5 times. Shares outstanding went from 24.2M to 70.0M, 2.9 times C (calculated).¹²² The share price at 30 June went from $2.70 in 2017 to $16.35 in 2026, 6.1 times, but it went through $0.55 in December 2019 and $1.26 in June 2024.¹ A holder from 2017 has done well only because of the last eighteen months.

The acquisitions are the capital allocation. ISP Optics cost $11.8M of net cash in December 2016, and it gave the company infrared; revenue was $28.4M the next year and $31.7M in fiscal 2024.⁷¹¹ Visimid, G5 Infrared and Amorphous Materials came in the last three years; G5 was valued at $27.1M and Amorphous at $9.2M.² G5 is the one that has shown up in the numbers: its cameras account for $23.1M of the $23.5M rise in assemblies and modules, and earnouts of $19.5M, most of the maximum, were earned.² The earnouts also cost something I want to name. Fiscal 2026 included a $15.6M non-cash charge because G5 did better than the company had forecast. That is a good problem, and it is also why the reported loss widened.²

The financing is the second half. In February 2025, with the shares near $2.15, the company raised about $25M of Series G preferred stock with warrants to pay for G5, from investors that included North Run and the Lytton family, who were already large holders.⁸² Those holders turned the preferred into 8.69M shares at $2.15 and the warrants into 3.47M shares at $2.58, and at $7.63 those 12.2M shares are worth about $93M C (calculated).² Then in December 2025 the company sold 8.9M shares at $7.75, about $69M gross, and in June 2026 3.6M shares at $14.00, about $50M.²¹⁸ Between December and June the shares went from $7.75 to $14.00. The company took money at both prices. I think the December price was fair and the June price was good, and the fact that a company with a record of losses sold shares at $14 and now trades at $7.63 is the reason I keep the funding risk in the table. There has been no buyback.

VFinancials

The history is a plateau followed by a jump. Revenue was $28.4M in fiscal 2017 and $38.5M at its peak before the jump in 2021, then $31.7M in 2024. It ended fiscal 2026 at $71.7M. The table has ten years, from the 10-Ks as filed.

Three things stand out. Gross margin fell from 52% in 2017 to 27% in 2024 and came back to 36%. Operating income was positive in three of the first four years and negative in each of the last six, reaching −$19.7M in 2026 including $15.6M of earnout charges C (calculated).² And cash never rose above $8.1M until the year the company sold stock in two offerings.

What an owner could really take out

Nothing, so far. Owner cash flow, operating cash flow less capital spending, was −$16.5M in fiscal 2026 and −$29.6M over ten years C (calculated).² Stock pay of $2.9M was a non-cash cost on top, and a further $3.5M of the G5 earnout was paid through financing, which I count as the price of the acquisition. The gap between the reported loss of $20.5M and the owner cash of −$16.5M is mostly the earnout charge, which is accounting, and the working capital, which is real. The company put $9.6M into receivables, inventory and supplier prepayments as sales grew, and that is the part of the cost of growth that a profit measure hides.²

Quality of earnings

Adjusted EBITDA of $4.2M is not owner cash. It starts from a net loss of $20.5M and adds back $5.0M of depreciation and amortisation, $2.9M of stock pay, $15.6M of earnout charges, $0.5M of debt extinguishment and a few small items.² I have no objection to adding back the earnout charge, because it measures a good result. I do object to using the result without stock pay: the new awards will be larger, and I deduct stock pay in my owner cash. The June quarter earned $2.1M of Adjusted EBITDA, a 10% margin, on $21.2M of revenue.² If that continues, the year is $8M, not $4M. The business has not yet shown a full year of it.

The receivable turns and stock turns bear watching. Days of sales outstanding were 65 in the June quarter against 48 to 58 in the three before, and days of inventory 128 against 100 to 119.² The company aims for under 60 and 110 to 120. Both slipped as the camera orders ramped and as it prepaid suppliers for long-lead materials.² I would rather see them fall than rise as revenue grows.

Taxes are not a cash cost for now. The company records a $10.5M deferred tax asset on loss carryforwards and reserves $17.8M against its total deferred assets.² I assume $40M of usable losses in my model, a rough figure, since I have not seen a Section 382 analysis and the prior ownership changes could limit them. The 10-K reports an accumulated deficit of $251.3M, which is how much the company has lost since it began.²

The balance sheet

At 30 June the company had $93.2M of cash, working capital of $103.7M and no meaningful debt.² The table sets it out. After the January earnout and the other items I count net cash of about $87M, of which a part is spoken for by the growth the backlog implies: capital spending is expected to be higher in fiscal 2027 than the $6.3M of 2026, and working capital will take more as camera shipments rise.² My base case spends about $11M of owner cash in fiscal 2027 and breaks even in 2029.

The share count is where the history of losses shows. Fully diluted shares were 74.6M on my count: 70.09M outstanding at 15 October, 3.02M from converting the preferred stock, and 1.5M of stock units and restricted shares C (calculated).⁵² On top of that are 1.5M older options at about $7.20 and the 2.98M performance options at $10.53, which I add only when my value per share is above their price. There is no way for me to say whether the company will issue more stock. It says in the 10-K that it has enough capital in the short term and may need more for acquisitions.² A company whose share count has risen in every one of the last ten years is likely to do it again, and a new offering at a price below my base value would cost holders.

The share price is the other side of the table. It was $2.70 in June 2017, $0.91 in June 2019, $1.26 in June 2024 and $3.08 in June 2025. It reached $16.35 at the end of June 2026, a high close of $18.15 on 28 May, and $7.63 on 2 November.¹ It closed at $9.68 on 10 September, the day of the results, and has since lost another 21% C (calculated).¹ I found no filing that explains the fall since July. The sale of the Chinese plant came on 23 July, when the shares were $12.37; North Run’s sales ended in June.¹⁶ Over the same stretch the number of shares traded rose from about 95,000 a day in 2024 to about 3.0 million a day since August, roughly $31M a day. That is not microcap liquidity any more. Before 2025 it was.

VIValuation

First, how this goes badly wrong

Before any number, I ask what would most likely destroy this investment. I see three ways, and the first is the one I would worry about.

The big order does not repeat. About $58.0M of the $110.9M of backlog, 52%, is camera systems for one customer shipping in calendar 2026 and 2027, and that customer was 24% of fiscal 2026 revenue.² If the programme ends with the order and nothing replaces it, revenue returns toward the roughly $67M that fiscal 2026 would have been without the Chinese plant’s sales, on a cost base that was built for more C (calculated).²⁴ Operating costs other than cost of sales were $24.7M of selling and administrative expense and $3.8M of development in fiscal 2026.² At a 36% gross margin $67M of revenue earns $24M of gross profit, about what those costs take once stock pay is set aside. The cash would last for years. The shares would drift toward the cash.

The germanium advantage fades. China’s U.S.-specific restriction is suspended only until 27 November 2026.² If the restrictions end, or new non-Chinese germanium supply opens, the price of germanium could fall from the $6,000 a kilogram the 10-K reports and the reason to redesign around glass would weaken. The defense deadline is a second reason, but it names covered nations, not a material.³ And BD6, the glass that sells today, is made by others. This risk does not need a collapse in anything. It only needs the premium for being germanium-free to shrink.

The company funds growth by selling stock again. Each time it has needed money it has sold shares, and the count has risen in every one of the last ten years.² Capacity for melting glass and for downstream assembly needs capital, and management says capital spending will rise in fiscal 2027.² A further acquisition or a slower order book could bring another offering. At $7.63 that would cost holders less than at $14.00, and more than at my base value of $2.72.

What an owner gets today

Almost nothing in cash. Fiscal 2026 operating cash flow less capital spending was −$16.5M, and stock pay of $2.9M comes on top, so what an owner could take out was about −$19.4M C (calculated).² Adjusted EBITDA was $4.2M, so the distance between the reported earnings measure and what the owner got is about $24M, most of it the $15.6M of earnout charges, which are not cash, the $9.6M of working capital, which is, and the $6.3M of capital spending C (calculated).² At today’s price the company is worth $569M fully diluted, 7.9 times revenue, and 114 times Adjusted EBITDA on enterprise value.

What I can’t judge

Four inputs I cannot know, so I show the range. How long the large technology customer will keep ordering: the company has not named it, and there is nothing in the filings about renewal. How fast defense programmes will qualify new glass: the company says two to three years, and I have no schedule. How much of the 2030 deadline will go to chalcogenide glass and how much to germanium from non-Chinese sources: the law does not say. And how much capital the capacity expansion needs: the company says more than the $6.3M it spent in fiscal 2026, and gives no figure. I have therefore made the range wide, and I would not take the base value as an estimate with a small error. It is the middle of a distribution that runs from about half a dollar to about six.

How I value it

I value what an owner can take out: Adjusted EBITDA less stock pay, capital spending, cash tax and the cash tied up in growth, discounted to today. I start from the balance sheet at 30 June, so all of fiscal 2027 counts, and I add net cash of $87M (cash of $93.2M, less debt and leases of $0.8M and the $9.0M January earnout, plus $4.0M for the Chinese receivable). I do not add cash earned since June, because I have not seen it. The discount rate is 11%: the 10-year Treasury yield of 4.91% on 2 November plus 6.1 points, for a company with a history of losses, a 24% customer and a share price that has moved six-fold in two years.¹⁹ I have not fitted a beta, because the business only took its present form in 2025. The sensitivity table shows the effect of the choice.

Bear

The large technology customer does not reorder after its $58M programme ships, defense qualifications arrive more slowly than the 2030 deadline suggests, and gross margin slips toward 30% as cameras meet price competition. Revenue is $82M in fiscal 2027 and grows 3% to 6% a year to $114M, Adjusted EBITDA margin sits at 6% to 9%, and a $10M one-off loss in 2028 stands for a failed program. Owner cash is negative in every year, about −$13M in 2027 and still −$3M in 2036. The cash pile is spent slowly. The model value is $0.50, a little under half what the company holds in net cash today.

Base

The $85.6M of backlog due in a year ships and the year lands at $95M, 32% above fiscal 2026 after losing the Chinese plant’s sales. Growth then fades from 15% to 5%, to $215M in 2036, with gross margin near 42% and Adjusted EBITDA margin rising from 10% to 19%. The 19% is my judgement of what a scaled defense optics supplier can earn; I have no peer data on this page to support it, and it is well above the 6% of fiscal 2026. Owner cash is −$10.9M in 2027, turns positive in 2029 and reaches $15.5M in 2036. The value is $2.72.

Bull

The company wins programme after programme on the back of the 2030 deadline and its camera line, and sells $363M in 2036, five times fiscal 2026. Adjusted EBITDA margin reaches 25% and gross margin 46%. The performance options for the CEO and CFO vest in full and I count them. Owner cash is $41.4M in 2036. The value is $5.60.

What the price says

The market is paying $482M for the operating business, which is 6.7 times fiscal 2026 revenue and 114 times Adjusted EBITDA. My base value for it is $116M. The table asks what has to be true to close the $366M gap. Three routes, and each needs a lot: revenue growing 21% a year for nine years at bull margins, to $600M in 2036; or 32% a year at my base margins; or margins above 40%. In plain words: the price needs a company about eight times the size it is, earning the best margin I can defend, within ten years.

A discounted-cash-flow answer is harsh on a young business, because most of the value sits in the last years and is cut by 11% a year for ten. So I checked it another way. A buyer paying 14 times Adjusted EBITDA at the end of fiscal 2031 would, on my base path, make the shares worth $4.38 and on my bull path $7.94. Even 18 times gives $5.29 and $9.79.

The cross-check says the bull case is worth about the price only if a buyer pays 14 to 18 times the profit of a business that has just done the whole bull path. That is a possible outcome. It needs two things at once.

Where the base value comes from

Of my base operating value of $116M, only $33M comes from cash in the ten years to 2036, because the early years are negative. $83M comes from the value after 2036 that I discount back. Add net cash of $87M for the total of $203M and divide by 74.6M shares to get $2.72. So roughly 72% of the value of the operating business is in years beyond 2036, which I cannot see, and net cash is 43% of the total. That is the reason I keep the margin of safety wide.

How much the answers move

The margin of safety

My margin-of-safety price is $1.77, 35% under my base. I use a wider cushion than for a large company because my range is wide, the history is poor and the largest customer can go. At that price an owner would be buying the net cash of about $1.17 a share and getting the operating business for less than $1. The price is 4.3 times that.

The price against the range

The price, $7.63, is 181% above my base value and 36% above my bull. The probability-weighted value is $2.52, using 35% bear, 45% base and 20% bull. I put more weight on the bear than on the bull because of the record: losses in seven of nine years and a share count that has risen every year.

Why might the market be right? Because it is not pricing a ten-year cash flow. It is pricing a company that doubled its revenue in a year, has $111M of backlog and a defense tailwind, and could be bought or could win a large programme. A sale is possible. A buyer who paid 18 times in 2031 on my bull path would value the shares at $9.79, above today’s price. Analysts, per Nasdaq, show a one-year target of $15.50, which is above the bull case here; I could not open the reports behind it.¹ I take that as a record of what some people think, not as evidence. My numbers are for a business that earns it.

Against the alternative

The next best use of the same dollar is not an optics company. A 10-year Treasury pays 4.91% and a 30-year 5.28%.¹⁹ At my base value the shares return -3.9% a year to June 2036 from today’s price, and at my bull 5.1% E (estimate). On this site the company I find closest to the price being below value is Booking Holdings, at 34% below my base value in its page, a business with large free cash flow, which LightPath does not have. That is a different kind of company, and I use it only to say that a dollar can find a better-supported price elsewhere.

Quality and price, separately

These are two questions. The quality of this business is improving and still unproven: a gross margin back up to 36% from 27%, a record backlog, a clean balance sheet and a plausible reason for customers to need it. In the ten years I looked at, its best operating profit was $4.1M. The price is a separate question, and at $7.63 it asks for a result beyond my best case. A business of improving quality can be a poor purchase at a price that already assumes the improvement and more. That looks like the situation here, in my numbers.

How the odds multiply

The price needs several things to go right at the same time, so I multiply them. These are my odds, not data.

Where this could be in 5 and 10 years

The table gives the value of a share at June 2031 and June 2036 on each path, with owner cash added to the net cash as it comes in and nothing paid out. The return is from $7.63 to the 2036 value.

VIIThe case for the price

What most people miss

Three things cut in LightPath’s favour and are not obvious. First, the balance sheet is no longer the story it was. In June 2025 the company had $4.9M of cash and had used $8.3M in operations that year; at 30 June 2026 it had $93.2M, no meaningful debt, and four straight quarters of positive Adjusted EBITDA R (reported).⁸²³ Second, the glass is a licence, not just a product. The best of it, BDNL, is under an exclusive licence from the U.S. Naval Research Laboratory signed in December 2021, and the company says the first of the family went on sale in April 2024.² Third, much of the visible-lens revenue comes from a factory that is being sold, so fiscal 2027 will compare with a base that includes about $4.5M of sales that will not be there. The reported growth will look lower than the business’s growth if that comes through.⁴ One thing runs the other way and is also missed: the China restriction on germanium exports to the United States is suspended, not in force, until 27 November 2026.²

The bull case

In plain words: thermal imaging is moving from a niche into border security, counter-drone defense, public safety and consumer devices; the standard lens material is controlled by China and has become scarce and six times as expensive; Congress has told the Defense Department to stop depending on covered nations for optical glass and systems by 1 January 2030; and LightPath melts its own alternative, owns exclusive rights to the best of the family, and has bought the camera makers to go further up the value chain.²³ In the September release Mr. Rubin said the fourth quarter set the high-water mark on every measure that matters to the company, that the sourcing decisions for those programmes are being made now, and that fiscal 2027 is about “capacity and conversion”.³ The proof offered is a backlog of $110.9M, up 197% in a year, follow-on orders of $11M and $13M, and a gross margin that reached 39.4% in the June quarter.³ The company has not given revenue or earnings guidance that I can find.³ Nasdaq shows a one-year target of $15.50 and I could not open the analyst reports behind it.¹

On the holder side, only dated facts. At 15 October BlackRock owned 4.29M shares, 6.1%, North Run 5.96M on an as-converted basis, 8.2%, and Hood River Capital Management 3.15M, 4.5%, according to the proxy statement.⁵ The proxy does not say what any of them plans to do. I found no named investor’s published model or letter on LightPath that I could open and cite.

Against reality

Today LightPath earns about $4M of Adjusted EBITDA on $72M of revenue, owns net cash of about $87M, and consumes cash after stock pay C (calculated).² If it ships the $85.6M due in the year, holds a gross margin of 40% or more and adds defense programmes at the rate the 2030 deadline implies, then my base is possible: $2.72 a share, with revenue near $215M and Adjusted EBITDA near $41M by fiscal 2036. The price needs more. It needs about $600M of revenue at a 25% margin, or growth of 21% a year for nine years C (calculated). That is the distance between today’s numbers and the case. The company has shown one year of doubled revenue, of which a large share was bought, and four quarters of small profit.

What a 10x would take. Ten times $7.63 is $76.30, about $5.7B of share value. At 18 times Adjusted EBITDA, after net cash, that needs about $311M a year, and at a 25% margin about $1,246M of revenue, 17 times fiscal 2026 and 3.4 times my bull case for 2036 C (calculated). If that took ten years it would mean revenue growing about 33% a year, every year, on top of a record of losses. The shares themselves did something like it before: from $1.26 on 28 June 2024 to a close of $18.15 on 28 May 2026, 14 times in 23 months.¹ That move began at a market value of about $50M and a company running out of cash, and it is not the start I have now. From a market value of $0.6B I would say a tenfold gain is close to impossible on any numbers I can defend, and I would not plan around it.

Where I come down

Three things moved me: the balance sheet is repaired, the gross margin has come up from 27% to 36%, and the backlog is real. They did not move my numbers to the price, because the price needs a company about eight times larger with the best margin I can defend, and because the largest order, which is more than half the backlog, has not shown that it repeats. A named second customer of similar size, a gross margin held above 40%, Adjusted EBITDA above $10M a year and shares outstanding that stop rising would raise my values most. The first test is the quarter report due in November, and the second is the end of the China suspension on 27 November.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• A history of losses. The company lost money in seven of the last nine fiscal years, has an accumulated deficit of $251.3M, and used $10.2M of cash in operations in fiscal 2026 even as revenue doubled.² Its cash a year ago was $4.9M. A company like this can fail if the money stops, and it was close to that point in 2025.
• Dilution. Shares outstanding rose from 24.2M to 70.0M in ten years, and 4.5M new options and 1.5M stock units are outstanding.² A further offering would be priced against a record of selling at the high and now a price 58% below it.
• Customer concentration. One customer was 24% of revenue and about 52% of backlog sits with one technology buyer.²
• Germanium and the China rules. The restriction suspension ends 27 November 2026.² A permanent easing weakens the case for redesign; a reinstated ban strengthens it, and also raises the cost of germanium products the company still sells.
• Tariffs and the Chinese supplier. After the sale, LightPath depends on a supplier owned by the factory’s management at cost plus 10% rising to 30%, and it imports from China into a tariff regime that has changed several times.⁴²
• Competition from large groups. Teledyne FLIR, Ophir and others sell standard products, and Lockheed Martin, RTX, Anduril, Elbit and Teledyne FLIR design assemblies themselves.²
• Long qualification. Supplier qualification cycles run two to three years, which also delays revenue.³
• Acquisition and earnout accounting. Earnouts cost $15.6M of charges in fiscal 2026 and $19.5M of payments in total; integration of three companies in three years is not shown yet.²
• Size and insurance. A company of 425 people that has doubled revenue and filled executive roles is stretched. Management judged its controls effective at 30 June 2026, with no auditor attestation, and the 10-K says the company does not carry product liability insurance.²
• Microcap liquidity. It was a microcap until 2025, with about 95,000 shares trading a day in 2024; it now trades about 3.0M shares ($31M) a day, but the share price fell 53% from the June close, which shows how far a thinly held stock can move.¹ Holders with board seats or large positions can change supply quickly.
• Related parties. North Run is the largest holder, has a board seat, holds all of the preferred stock and sold $90M of stock in 2026; the buyer of the Chinese plant is owned by the factory’s management and others.⁵⁶²
• Latvia and Europe. 107 of the 425 employees and the company’s only European plant are in Riga, in a region affected by the war in Ukraine.²