On 17 September 2025, STMicroelectronics said it had already allocated over $60 million to developing a panel-level packaging pilot line at Tours. It expected the line to be operational in the third quarter of 2026. That quarter has passed. The documents reviewed for this article, as of 8 November 2026, contain no confirmation of commissioning. They describe an investment allocation and an expected start, not completed spending or a demonstrated production outcome.
For a candidate considering ST Tours or Rousset, the useful distinction is between the site's announced mission, the outcome that a document actually establishes, and the work that a particular employer posting describes.
The two places belong to the same group, but the April 2025 industrial plan gave them different directions. Rousset was to stay focused on 200 mm manufacturing, receiving additional volumes from other sites. Tours was to retain a silicon line for selected technologies and its gallium nitride competence, mainly epitaxy, while transferring legacy activities and developing packaging. These are ST's declarations. Their specificity makes them useful; their future tense also sets a limit on what we can say.
At Tours, allocation preceded the expected start
ST's September announcement provides the strongest dated account of the proposed Tours packaging investment. The body of the release says:
The development of the new PLP pilot line in Tours is supported by a capital investment of over $60 million, already allocated as part of the company-wide program to reshape the Company’s manufacturing footprint.
The sentence names Tours, panel-level packaging, a capital allocation and the wider footprint programme. It does not print a completed-spending figure. Nor does it attribute that money to the older GaN work. Reading the amount without those nouns would turn a specific programme into a misleading general investment total.
The operational quarter is a separate statement. ST expected its pilot to operate in Q3 2026. An expected date is useful evidence of what the company intended when it spoke. It is not a completion certificate, even once the date has elapsed. Equally, our failure to find confirmation is not evidence that the pilot failed to start. We leave that outcome unconfirmed in the reviewed record, rather than choosing the more dramatic interpretation.
A later employer document provides a different kind of evidence. A Tours CIFRE thesis description concerns the interfaces within PLP stacks. That is a named research task, not an operating-status announcement. A candidate can learn something about the proposed research without treating the presence of a posting as proof of commissioning. An industrial announcement and a thesis description answer different questions.
At Rousset, incoming volumes were a means, not a measured result
The 10 April 2025 announcement says Rousset would remain focused on 200 mm manufacturing, with additional volumes reallocated from other sites to enable full saturation of existing manufacturing capacity. The wording connects a proposed action with its intended benefit. It does not report the amount transferred, the capacity reached, or a date when saturation was achieved. This article therefore describes planned concentration and incoming volumes, not a factory already proven to be full.
There is a much older capacity figure in the record. EE Times reported the formal dedication of Rousset8 on 15 May 2000. Its account said the fab would eventually have capacity for 7,000 eight-inch wafers per week; a fully equipped facility carried a total planned investment of $1.4 billion, according to company officials. These are historical, attributed plans. They are not current capacity or completed expenditure.
The historical account helps explain why a precise-looking number can still answer the wrong question. A candidate reading about Rousset today needs the scope and date of the job, not an unqualified weekly capacity extracted from a quarter-century-old headline. We retain the report because it documents an earlier announced industrial stage. We do not carry its projected output forward to 2026.
Dated announcements preserve their targets, not their outcomes
| Document | What it states | What remains unproved here |
|---|---|---|
| EE Times, 15 May 2000 | Rousset8 formally dedicated; eventual 7,000 wafers/week; planned $1.4 billion when fully equipped | Achieved output, realised spending, current capacity |
| CEA, 24 September 2018 | GaN technology transfer planned from CEA R&D to an ST 200 mm pilot, targeting operation by 2020 | Completed transfer or operation; no PLP claim |
| ST, 10 April 2025 | Different retained and transferred missions for Tours and Rousset | Completed transfers or achieved Rousset saturation |
| ST, 17 September 2025 | Tours PLP: over $60 million allocated; operation expected Q3 2026 | Commissioning, realised spending, jobs created |
Sources 1 STMicroelectronics · 2 STMicroelectronics · 4 EE Times · 5 Commissariat à l'énergie atomique et aux énergies alternatives
The GaN record is older, and it is not the PLP record
CEA's French announcement of 24 September 2018 describes cooperation with ST on GaN-on-silicon for power conversion. It says the technology developed within IRT Nanoelec would transfer from CEA's 200 mm research line to an ST-operated 200 mm manufacturing pilot intended to be operational by 2020. The source is a research partner's dated declaration, not a retrospective report of the target being met.
That timeline matters because the September 2025 packaging announcement can otherwise be mistaken for the beginning of all advanced work associated with Tours. The April 2025 plan separately retained Tours as a GaN competence centre, mainly on epitaxy. Keeping those statements separate avoids making the older research programme and the newer PLP investment look like one project with one budget and one commissioning date.
A separate Les Echos investment report remains outside this article's admitted citations because we did not obtain its full original. We do not reproduce its unverified amount or date. The CEA announcement does not supply the missing investment figure. Naming a relevant partner is not a licence to attach a sum that its document does not print.
For someone comparing technical opportunities, that distinction changes the questions to ask. Is the work about growing material, characterising an interface, or supporting a manufacturing process? The April plan names epitaxy at Tours; the thesis names interfaces; the Rousset description names plasma-etch production support. Those are the terms the documents actually use. A broad label such as semiconductor engineering would hide the differences rather than help the reader evaluate them.
ST's site missions diverge within the same industrial plan
| Site | Retained in the declaration | Transferred or developed in the plan | Basis |
|---|---|---|---|
| Tours | 200 mm silicon for selected technologies; GaN competence, mainly epitaxy | Legacy 150 mm activities to other ST sites; PLP pilot development | April plan; September PLP release |
| Rousset | Focus on 200 mm manufacturing | Additional volumes from other sites, intended to saturate existing capacity | April plan |
Sources 1 STMicroelectronics · 2 STMicroelectronics
Two roles make the work more legible
The Tours thesis description names experimental and numerical work on PLP interfaces. In its original French, the employer says:
Cette thèse vise à développer une approche expérimentale et numérique permettant de caractériser les interfaces des empilements PLP, de comprendre leur comportement mécanique et d'identifier les paramètres gouvernant leur tenue et leur endommagement.
In our translation, the thesis aims to characterise interfaces in PLP stacks, understand their mechanical behaviour and identify the parameters governing their integrity and damage. The wording names a research question. It does not promise a completed result. The intended methods include X-ray tomography, scanning electron microscopy and, where relevant, atomic force microscopy. That qualification matters: AFM is conditional in this description, not an unconditional requirement we add ourselves.
The Rousset process-engineer description answers another question. It asks for daily production support on the processes and equipment under the person's responsibility, rapid identification of malfunction causes, and adapted technical solutions. It also says regular presence close to production, including cleanroom presence, is indispensable to the duties. The employer is describing responsibility in a specific plasma-etch engineering role. It is not declaring a schedule for every employee at Rousset.
These documents have dates, but their date basis needs to travel with them. The Tours original declares 26 October 2026 in employer JobPosting metadata; the Rousset original declares 29 September 2026. Neither timestamp includes a timezone. These are the employer's declared dates, not independently confirmed first-publication days. Both originals were read on 8 November 2026. Application buttons and future validity fields do not establish that the vacancies remain available; no application was submitted.
The employer describes research interfaces and production support separately
| Document and place | Employer-described work | Boundary |
|---|---|---|
| Tours CIFRE thesis | Experimental/numerical interface characterisation; X-ray tomography and SEM; AFM where relevant | Proposed research methods, not completed work or installed pilot equipment |
| Rousset plasma-etch engineer | Daily process/equipment production support; diagnose malfunctions; regular cleanroom presence | One role's duties, not achieved saturation or universal working conditions |
| ST April industrial plan | Stated shift toward process control, automation and design | Company direction, not measured market demand or a hiring total |
Sources 2 STMicroelectronics · 3 STMicroelectronics · 6 STMicroelectronics
Employment numbers belong to their own perimeter
ST's April announcement projected up to 2,800 voluntary departures globally, on top of normal attrition. EE Times then reported, on 30 April 2025, an ST estimate that up to 1,000 employees could choose to leave in France by the end of 2027. The first is a group projection. The second is trade reporting of a France-wide estimate; we have not opened the underlying management statement. Neither supplies a Tours or Rousset allocation, nor a count of departures that actually occurred.
The legal register provides another set of numbers with another scope. The Annuaire des Entreprises pages, using INSEE Sirene data, print 2024 staff bands of 1,000–1,999 for STMICROELECTRONICS (TOURS) SAS and 2,000–4,999 for STMICROELECTRONICS ROUSSET SAS. The figures' sections were updated on 7 November 2026 and read on 8 November. The update date does not turn the underlying statistical reference year into 2026.
Those names resemble sites, but the rows concern legal units. The Tours entity also has a listed active establishment in Rennes; the Rousset entity has one in Biot. The register describes its staff band as a statistical variable. It is neither an exact headcount nor a site-only number. We do not add the ranges, pick their midpoints, or subtract planned departures from them. Different scopes and years cannot yield a reliable local staffing story merely because all the numbers concern ST.
The missing denominator matters as much as the visible numerator. A France-wide departure estimate does not become a local risk rate when paired with an entity's staff band. We would need matching geography, a matching population and an observed outcome, none of which that pairing supplies. The honest question for a prospective applicant is therefore not which site lost the larger percentage. It is which dated document explains the particular team's work and which questions still need an answer from that team.
This also prevents an apparent contradiction from becoming a false story. A company can announce voluntary departures and print a specific role description in the same period. Those documents do not establish the balance between departures and hires, whether the role replaces someone, or whether it adds a position. We cannot infer expanding local employment from the posting, and we cannot infer that every technical opportunity has disappeared from the restructuring plan.
What this means for your career
A useful preparation exercise starts with the duties, not the capital figure. For the Tours research description, select an example from your own work in which you compared observations with a numerical account of material behaviour. Explain what was measured, what the model assumed, and what evidence would make you revise the explanation. This is our suggested preparation, not an extra employer requirement or an assurance of selection.
For the Rousset description, choose a real process or equipment malfunction you helped investigate. Separate the symptom, possible causes, the check that distinguished them, and the technical response. If you have no cleanroom experience, say so rather than disguising a laboratory exercise as production work. The posting's requirement is concrete; your evidence should be equally concrete.
When discussing either role, distinguish what you have done from what you would like to learn. A proposed method in the Tours thesis is not permission to claim practical experience with it. If your background is numerical rather than experimental, explain that boundary and the observations your model would need. For process support, a response that names a diagnosis without explaining the discriminating check leaves the central responsibility unclear. These are preparation questions suggested by the printed tasks, not a scoring rubric supplied by ST.
You can also ask a prospective team which stage its work is in. For Tours, that means asking about the task's relationship to pilot development rather than asserting the pilot is operational. For Rousset, ask which processes and equipment the role would support rather than assuming the April volume transfers were completed. A dated announcement gives you a reason to ask a focused question. It does not predetermine the answer, and an interview answer would require its own verification before becoming a published industrial fact.
Before acting, reopen the exact employer document and check the role is still offered. Keep the place, title and duties together. A matching platform's suggested skills are not the employer's requirements. Neither of these two descriptions earns a market-wide salary claim, a training entitlement, or a prediction of hiring growth. If a role disappears, its dated description can still explain a task, but it no longer establishes an application opportunity.
The industrial record and the job record can help each other without being merged. The first explains the mission ST announced for each site; the second names work an employer described. Keep the unanswered operating and staffing questions alongside them. That gives a candidate a better-grounded conversation than either a restructuring headline or an investment amount alone.