Executive brief
On 13 November 2025, a press release datelined Grenoble, France announced that engineering models of Teledyne's newly launched industrial CMOS image sensors — upscreened for space applications — would be available with evaluation kits and integration tools (Teledyne, 13 November 2025). Six months earlier the same product family had been introduced as three industrial sensor variants from 1.3 MP to 67 MP, taken through a delta space qualification methodology and radiation testing (Teledyne Space Imaging, 22 May 2025). Read together, those two dates describe a strategy, not a product launch: the Saint-Égrève site is monetising the same silicon twice, once on a factory floor and once in orbit. This dossier reconstructs that move from primary releases, places it against the site's own dated industrial history — a €8 million cleanroom project begun on 4 May 2021 in a plant that processes 2,500 lots a year — and states plainly the nine claim families we refused to print because no source we opened supports them.
I. What the site is, precisely
Precision about the legal entity matters here, because "Teledyne" is a portfolio and readers routinely mistake one part of it for another. The operator in Isère is Teledyne e2v Semiconductors SAS, registered at Saint-Égrève (38120) under SIREN 341 470 656 in the French national business register. The site itself long predates its current ownership: regional business coverage records a company founded in 1955, a wholly owned Teledyne subsidiary since 2017, specialising in the design and manufacture of high-performance electronic components and systems (Présences Grenoble, 31 May 2021).
The ownership date is not folklore. Teledyne Technologies and e2v technologies plc jointly announced agreed terms for a recommended cash acquisition on 12 December 2016 (Teledyne, 12 December 2016), and jointly announced completion by scheme of arrangement on 28 March 2017 (Teledyne, 28 March 2017), the same date carried by Teledyne's filing with the US Securities and Exchange Commission. A French design-and-manufacture site of the 1955 vintage therefore sits inside a US-listed group, and has done so for nine years. That single structural fact governs almost everything a career-minded reader needs to understand about the place: reporting lines, export control, and which decisions are taken in Isère versus elsewhere.
II. The two markets, and why one silicon serves both
At completion, Teledyne described the acquired business in market terms: for the machine vision market, e2v provides high-performance image sensors (Teledyne, 28 March 2017). That is the industrial half — sensors that run inspection and machine-vision lines, sold on frame rate, resolution and cost of ownership.
The scientific half has a longer paper trail. The detector design for the visible instrument of the European Space Agency's Euclid mission was specified by the Euclid VIS instrument team in close collaboration with ESA and e2v technologies, a collaboration documented in the peer-reviewed instrumentation literature on 23 July 2014 (SPIE Proceedings, 23 July 2014). When Euclid flew, the company published the consequence in one sentence: Teledyne e2v sensors on ESA's Euclid mission will explore the composition and evolution of the dark universe (Teledyne e2v, 28 July 2023), a claim Teledyne's corporate newsroom had already made on 26 June 2023. Nine years separate the design paper from the launch release. That interval is the space-detector business model, and it is the reason a scientific sensor team is a slow, cumulative asset rather than a product line that can be stood up on demand.
The 2025 announcements join the two halves. Three industrial CMOS variants spanning 1.3 MP to 67 MP were released following a delta space qualification methodology and radiation testing (Teledyne Space Imaging, 22 May 2025) — the word delta carrying the whole commercial argument, because it means the qualification starts from an existing industrial part rather than a bespoke space design. The company's product literature describes what that yields: the Emerald Gen2 12M USV as a compact 12-megapixel global-shutter sensor offered for monitoring cameras on satellites, moon rovers and landers, or space suits, and the Emerald 67M USV as an 8K global-shutter sensor for wide-swath Earth observation, combining low noise (below 3 electrons at 12 bits) with speeds up to 65 frames per second at 10 bits (Teledyne Space Imaging product flyers, undated, consulted 3 September 2026). We print those specifications as manufacturer literature with a consultation date, not as measured performance, and we date them as undated: they are the vendor's claim about its own part.
The mechanism, stated once. Space-grade volumes are tiny and space-grade qualification is expensive. A merchant sensor designed for factory vision amortises its mask set, its yield learning and its supply chain across industrial volume; a delta qualification then buys orbital credibility for a fraction of a clean-sheet space programme. The scarce input is not the die. It is the evidence: radiation test campaigns, screening, traceable lots, and engineers who can defend a part to a mission assurance authority.
III. The industrial base underneath the announcements
Announcements are cheap; cleanrooms are not. On Tuesday 4 May 2021, the first stone was laid to launch upgrade work on the assembly and test cleanroom at the factory near Grenoble, a project named GECkO (Teledyne e2v, 5 May 2021). Regional press put the figure at €8 million, self-financed (ESSOR Isère, republished without a legible date — we flag the dating rather than launder it), and the national business daily Les Echos reported on 2 June 2022 that the plant produces 2,500 lots per year, each containing 30 to 200 electronic components. Le Dauphiné Libéré covered the same project on 1 June 2021, framing it as a competitiveness move in a sector whose customers tolerate nothing but excellence.
That lot statistic is the most analytically useful number in this dossier, and it deserves arithmetic rather than admiration. Between 30 and 200 components per lot across 2,500 lots implies an annual output somewhere between roughly 75,000 and 500,000 components — a range so wide that it tells you what kind of factory this is: high-mix, low-volume, per-part traceable. A plant with that profile does not compete on unit cost against Asian consumer sensor volumes. It competes on qualification, documentation, longevity and the ability to deliver a part that must still be procurable a decade later. Assembly and test — the step GECkO upgraded — is exactly where that promise is kept or broken.
IV. Market context, disclosed at the strength we have it
The market frame for CMOS image sensors is the weakest evidence in this dossier and we say so in the body rather than in a footnote. A widely circulated analyst view puts the CMOS image sensor market on a 4.7% compound annual growth rate from 2023 to 2029, reaching US$28.6 billion — figures we could only reach through a secondary recap of an analyst webinar published 2 December 2024, because the analyst's own page was unavailable to us. We therefore treat them as directional context, in US dollars, single-sourced and not desk-certified. They are not the basis of any conclusion here.
The public-policy frame is firmer, and it is French. Under France 2030, the state is devoting €5 billion to the development and industrialisation of electronic technologies (Direction générale des Entreprises, 18 October 2024), and the programme's own public microsite states a workforce objective of 35,000 people to be trained in the electronics and robotics sectors by 2030 (France 2030, undated page, consulted 3 September 2026). Neither figure is Grenoble-specific, and we do not allocate a share of either to this site. What they establish is the direction of national capital and the fact that the constraint has been officially named as people.
V. What we refused to print
A dossier is defined as much by its refusals as its findings. No source we opened establishes: the price or valuation of the e2v acquisition; headcount at the Saint-Égrève site, in any year; site-level revenue or profitability; wafer process specifics, node or foundry arrangements beyond marketing-level literature; salaries for Grenoble roles; the value, date or scope of any ESA or CNES contract tied specifically to this plant; quality or aerospace certifications held by the site; site awards or safety records; and any dated statement quantifying a skills shortage at this employer. Where a widely repeated figure exists but we could not open a source for it — the acquisition price is the clearest case — we print the absence, not the rumour.
VI. What to watch next
Three observable events would confirm or break the reading above. First, whether the upscreened Emerald family appears in a named mission or a named integrator's design, because engineering models with evaluation kits (Teledyne, 13 November 2025) are an invitation, and an invitation is only validated by a design win. Second, whether assembly and test capacity is extended again after GECkO, since a delta-qualification strategy consumes screening and test capacity faster than it consumes wafers. Third, whether France 2030's stated 35,000-person training objective produces electronics and photonics course capacity in Isère specifically — the constraint the state has named, and the one this site shares with every employer in the sovereignty stack.
Nothing in this dossier depends on a forecast. It depends on four dated releases, one peer-reviewed instrumentation paper, one national daily, two regional titles and one national register — and on being explicit about the nine things we could not verify.