The chokepoint Europe forgot it owned: reading Exosens from its filings

I. The mechanism: what is scarce is not the device, it is the tube

A night-vision binocular is an assembly. Its housing is machinable, its optics are competitively supplied, its electronics are ordinary, and its integrator can be swapped in a procurement cycle. The part that cannot be swapped is the image-intensifier tube: a vacuum device in which a photocathode converts scarce photons to electrons, a microchannel plate multiplies them, and a phosphor screen turns them back into an image the eye can use. Its performance is expressed in a single scalar, the Figure of Merit, and Exosens' own leaflet for its fourth-generation tube states an excellent FOM of 1800 minimum. Its product page for the same tube claims that the technology "has become Exosens' high runner over the last 10 years and is today the benchmark in all major European Land Forces programs."

That sentence is worth reading as an economist rather than as a reader of marketing copy. It asserts a position in the bill of materials of programmes that are procured by states, qualified over years, and re-ordered by the tens of thousands. The economics that follow are not software economics and not defence-prime economics. They are the economics of a qualified component: the vendor cannot raise price at will, because the customer is a ministry; but the customer cannot re-source at will either, because re-qualifying a photocathode inside an accepted optical chain is a programme event, not a purchasing decision. Margin, in that structure, is not a function of brand. It is a function of how few plants in the world can hold the process.

This is why the industrial geography matters more than the org chart. The tubes are made in Brive and in Roden. Everything else in the group (infrared cameras, hyperspectral imagers, radiation and ion detectors, ultraviolet detection) has been added around that core since 2022, and can be read as an attempt to convert one qualified process into a portfolio of qualified processes before the defence cycle turns.

II. The ownership ladder: Ardian, HLD, Euronext

The chain is documented and dated. In February 2021 Ardian announced that "HLD Europe est entré en négociations exclusives avec Ardian, en vue de réaliser l'acquisition de Photonis": a French sponsor taking a French photonics asset from another French sponsor, after an earlier attempted sale to a US buyer had become politically contested. In September 2023 the group renamed itself, and its own release recorded that it had been "détenu par le Groupe HLD depuis 2021" and had acquired four companies since December 2022.

In June 2024 it listed. Exosens announced the launch of the offering on 3 June and its success on 7 June, both from Mérignac. Euronext's own release put numbers on it: a private placement of €350m extendable to €402.5m, €180m raised in new capital, a market capitalisation of roughly €1.016bn, and the eighteenth listing on Euronext in 2024. Agefi, covering the launch, wrote that the operation valued the former Photonis at a billion euros in the market and that the company "restera contrôlée par le fonds HLD."

Two things follow. First, the float is a partial one: control did not change hands at the IPO, it was monetised. Second, and this is the reason we say so explicitly rather than estimating, the free-float percentage and the detailed post-IPO capital breakdown are not stated in either the Exosens or the Euronext release. Placement size and market capitalisation are not a shareholding table. We do not print one.

III. The roll-up: six businesses, six undisclosed prices

Between December 2022 and November 2024 the group bought, in its own announcements: Xenics (Leuven, Belgium, infrared, described by the acquirer as "over 65 people", 15 December 2022); Telops (Quebec, Canada, hyperspectral and infrared imaging, 11 April 2023); ProxiVision (Bensheim, Germany, ultraviolet detection, 25 April 2023); El-Mul (Rehovot, Israel, ion and electron detection, 21 July 2023); Centronic (Croydon, United Kingdom, radiation detection, closing announced 1 August 2024); and Noxant (Palaiseau, France, high-performance cooled infrared cameras, exclusive negotiations announced 18 November 2024).

Not one of those six releases states a consideration. That is a finding, not a gap in our reading: a listed issuer that discloses no price for six consecutive acquisitions is telling the market that none of them was individually material enough to require it, and is simultaneously making it impossible for an outside analyst to separate organic growth from bought growth on a per-deal basis. The company itself supplies the only bridge available: in FY2024 it reported 35.0% total growth of which 24.9% like-for-like. The residual is inorganic, and it is roughly ten points of growth whose acquisition cost is not public.

IV. The numbers we will print, and the ones we will not

The audited spine is the part of this dossier where nothing needs hedging. FY2024, in the release filed through the French official-publication channel on 3 March 2025 and published simultaneously by the company: revenue €394.1m, +35.0%; adjusted EBITDA €118.5m, +37.8%; margin 30.1%, against 29.5% the prior year. H1 2025, dated 30 July 2025: revenue €224.5m, +20.1%, with amplification, the tube business, up 17.6%; adjusted EBITDA €69.5m, +23.8%, margin 30.9%, ninety-two basis points better year on year. Q1 2026: revenue €122.6m, +19.7%. The universal registration documents for 2024 and 2025 are both filed and openable, the 2024 one under number R.25-001, deposited 29 April 2025; the 2025 one deposited 9 March 2026.

Three numbers that circulate about this company are not in that spine, and we treat them differently. A group headcount of 1,800 and a FY2025 revenue of €468.2m appear in a single regional trade publication dated 3 July 2026, in a parenthesis. We could not open that article to verify it still serves, and we found no Exosens-issued document in this pass that states either figure. So they are reported here as what they are (one publisher's parenthetical, uncorroborated by the issuer) and they are used to support no argument in this dossier. A per-site headcount split between Mérignac, Brive, Roden, Sturbridge, Palaiseau and the acquired units is published nowhere at all, so no site is described here as large or small.

V. Capacity: €20m, a first American plant, and €140m from the EIB

Growth in a qualified-process business is a capacity question before it is a demand question, and the company has answered it twice in public. On 3 March 2025, the same day as the FY2024 results, it announced that it would invest €20m over the following two years to raise production capacity in Europe and the United States, framed in the international release as its first US investment in night-vision production capacity. It already holds an American footprint to build on: Exosens Scientific USA at 660 Main Street, Sturbridge Business Park, Massachusetts.

On 25 June 2026 it announced a €140m financing from the European Investment Bank to support innovation in the European defence and security industry. Regional coverage of that financing states that the tubes concerned are produced at Brive-la-Gaillarde and Roden. We flag that the local article carrying that sentence, and a second regional piece pairing the EIB financing with a Czech army contract, could not be opened from our environment: one returned a 403, another returned no response at all. The financing itself is announced by the company on its own site; the plant attribution rests on press we could not certify as serving, and is therefore printed as attributed rather than as established.

VI. The €500m order, and the three counterparties it is given

On 10 December 2025 Exosens announced, from Mérignac, "un contrat historique de 100 000 binoculaires MIKRON, intégrant 200 000 tubes Exosens pour une valeur totale d'environ 500 millions d'euros: le plus important contrat jamais signé dans l'histoire de la vision nocturne." Two tubes per binocular; a single order roughly the size of a year and a quarter of group revenue at the FY2024 rate.

Who bought it is where the record splits, and we refuse to resolve the split by choosing the tidiest version:

• The company's own release frames the counterparty as Theon International, the Greek night-vision integrator whose MIKRON product carries the tubes.
• A French defence outlet, dated 11 December 2025, frames it as a contract won "via un consortium avec l'allemand Hensoldt" for the Bundeswehr, again at 100,000 binoculars and 200,000 tubes.
• A regional business title, headlined 12 December 2025, frames the buyer as the European intergovernmental armaments organisation, procuring for the German and Belgian armies.

These are not necessarily three different deals; they are three different answers to the question "who is the customer", and that question is precisely the one that determines whether Exosens is a supplier to an integrator, a member of a prime consortium, or a named supplier inside a multinational procurement. An analyst who silently picks one has manufactured a fact. We print three and note that only the first is issuer-sourced.

VII. What this record refuses to say

Discipline in a dossier like this is mostly a list of sentences not written. The following are absent from every source we could open, and are therefore absent here: any acquisition price for Xenics, Telops, ProxiVision, El-Mul, Centronic or Noxant; any standing tube production volume per year at Brive or Roden: the only volume in the record, 200,000 tubes, belongs to one contract and is not a capacity statement; any free-float percentage or post-IPO shareholding table; any company-sourced group or per-site headcount; any salary band, none of the job postings we opened publishing one; and any figure for European night-vision market share, which no openable source states and which we will not derive from a contract win.

VIII. What to watch, and why it is a career signal rather than a stock signal

Three observable events would change the reading. First, the FY2025 results release itself: it would either corroborate or contradict the €468.2m that currently rests on one parenthesis, and it would show whether the amplification line kept the 30%-plus adjusted-EBITDA margin reported for H1 2025 (30.9%, dated 30 July 2025) while capacity was being added. Second, the deployment of the €20m and the €140m into named lines at Brive, Roden and Sturbridge: qualified-process capacity is built in identifiable steps, and each step is a hiring event in a town of Brive's size. Third, the contractual identity of the MIKRON programme's buyer as it appears in subsequent filings, because the difference between "supplier to Theon" and "consortium member with Hensoldt" is the difference between a component vendor and a programme participant.

For anyone reading this as a place to work rather than a security to price, the structural fact is the one from section I. This is a company whose defensibility lives in a process qualified inside other people's programmes, in two towns, one of which has hosted it since 1937. That is a rare kind of employer: the scarcity is in the plant, not in the pitch. The companion dossier reads what that plant demands of the people in it: strictly from what the company publishes about what it ships and hires for, and with every salary claim refused.