Skip to content

Forvia SE (Faurecia + HELLA)

Company Overview

Business Overview

Forvia SE was created from Faurecia's takeover of HELLA (majority stake closed early 2022; the combined group adopted the FORVIA name that year). Faurecia itself dates from 1997 (Bertrand Faure + ECIA); HELLA was founded in 1899 in Lippstadt, Germany. The group is headquartered in Nanterre, France, and listed on Euronext Paris. Scale (as of end-2022, the last fully consolidated public snapshot in the engineering deck cited): ~157,000 employees, 291 plants, 76 R&D centres, 15,000 R&D engineers, 14,000+ patents. [Evidence] 118

Leadership: Patrick Koller led Faurecia/Forvia through the HELLA acquisition and the 2023 AI announcements. On December 2, 2024 the board announced Martin Fischer (ex-ZF board member) as his successor; Fischer became CEO on March 1, 2025. [Evidence] 4

Key financials — and the honest picture is strained:

  • FY2025: sales €26.2 billion (≈€27bn at constant currency, flat); operating margin 5.6% (up 40bps); but a net loss of €2.1 billion, driven by ~€1.85bn of non-cash write-downs — a €920m impairment in Lighting/Electronics, a €578m capital loss on the Interiors divestiture, and a €209m write-down of the Symbio hydrogen stake. Net debt €6.0bn (1.7× EBITDA, from 2.0×); net cash flow €962m (+47%); dividend suspended. Results published February 24, 2026, alongside a Capital Markets Day launching the IGNITE strategy. [Evidence] 56
  • Debt is the story behind every strategic move since the HELLA deal was financed: deleveraging targets (1.5× by end-2026), the EU-FORWARD European restructuring (€165m saved in 2025; 6,400 headcount reductions announced), and ultimately the Interiors sale. [Evidence] 567

Automotive Business

Forvia is automotive, wholly. Business groups: Seating, Interiors, Clean Mobility (exhaust/hydrogen), Electronics, Lighting (HELLA), Lifecycle Solutions. Customers span essentially all major OEMs — Volkswagen Group, Stellantis, Ford, GM, Toyota, Tesla, BYD and others. [Evidence] 11

April 27, 2026: Interiors sold. Apollo-managed funds agreed to acquire the entire Interiors business group at a €1.82bn enterprise value; closing targeted by end-2026; Interiors will become a standalone company. This removes from Forvia the business that makes instrument panels, door panels and centre consoles — the core large-moulded-parts operation. Seating (foam, plastic trims, structures) and HELLA Lighting (injection-moulded optics) keep the group in moulded plastics, but the flagship interiors plastics organisation exits. [Evidence] 7

Engineering Business

Forvia is one of the largest CATIA customers in the automotive supply base: the group ran ~2,500 CATIA V5 licences (HELLA alone ~1,030 licences, 3,500+ users). HELLA's lighting engineering is a fully-fledged optics-and-moulding engineering organisation: its Mohelnice (Czech Republic) site alone has 3,325 employees doing duroplast (a thermoset plastic) and thermoplastic injection moulding, thick-wall optics, light guides, and a fully automated single-shot/two-shot (1K/2K) lens process. [Evidence] 8

Role in Automotive Moulded Parts Workflow

Customer-side reading: for each stage, what does Forvia itself do, and is any AI in evidence?

Workflow Stage Forvia's position AI in evidence?
Requirements OEM-driven; internal programme management No
Industrial Design Cockpit/lighting styling in-house No
Concept Design In-house (CATIA V5) No
CAD Modelling Core activity — CATIA V5 primary CAD; 3DEXPERIENCE co-existence No (xGenerative design productive — optimization, not AI12)
Engineering Review MBSE push at HELLA (CATIA Behavior/Dymola) No
Simulation (structural/crash) Core activity Yes — ODYSSEE ML crash surrogates (2021)2
DFM Manual/rules; nothing public No — nothing found
Tool Design In-house and supplier moulds; HELLA has deep in-house moulding No
Mould Flow Uses commercial solvers (Moldex3D verified, 2015) No — classical simulation1
Prototype In-house No
Validation Global Validation Services (the ODYSSEE user) Partially (same surrogates)
Manufacturing Engineering Digital factory programme (3DEXPERIENCE/DELMIA AGV logistics, 2022); "dark plants" ambition Simulation/automation, not learning AI93
Production Release Standard Tier-1 PPAP (Production Part Approval Process) processes No

The shape of the table is the finding: the two "yes/partially" rows sit at simulation/validation (crash CAE) — and the moulding-specific rows (DFM, tool design, mould flow) are all "no AI." [Synthesis]

AI Strategy

Public AI Vision

Forvia's public AI story has two distinct eras. [Synthesis]

Era 1 — the Koller mandate (2023–2024). On December 18, 2023, CEO Patrick Koller told Reuters that Forvia aimed to cut R&D costs in half by 2028, mainly thanks to generative AI, across its 15,000+ engineers and 76 R&D centres. His own words on the 50% figure: "it is scary for two reasons. First it is huge and will require a lot of restructuring, but above all, if you don't do it, within three to five years, you are out of the game." He also announced plans for five fully automated "dark plants" within two years (one existing, in Spain). [Evidence] 3

Era 2 — the Fischer redirection (2025–2026). Koller left March 1, 2025. Under Martin Fischer, the February 2026 results and IGNITE Capital Markets Day frame AI much more narrowly: the SIMPLIFY programme targets €110m of administrative cost reduction by 2028, "partly through the use of GenAI." Meanwhile the R&D cost lever visible in public is not AI but outsourcing and restructuring: a ~€45m five-year engineering-services deal moving Clean Mobility engineering (≈300 engineers' worth of work) to L&T Technology Services (May 2024), and the EU-FORWARD restructuring. No public reaffirmation of the "halve R&D by 2028 via generative AI" target under Fischer was found. [Evidence] 5104 [Evidence of absence]

Investor Statements

The December 2023 Reuters interview is the high-water mark of AI ambition. The February 2026 communications are dominated by deleveraging, the Interiors sale and margin recovery — AI appears as a cost-programme ingredient, not a strategy pillar. [Evidence] 356

Engineering AI Strategy

What Forvia's engineers demonstrably use, versus what was announced:

  • Demonstrated (real ML): ODYSSEE CAE reduced-order models for seat crash simulation at the Brières Tech Center (2021) — machine learning trained on batches of finite-element runs. [Evidence] 2
  • Demonstrated (not AI): Moldex3D mould-flow simulation (Pune, 2015); 3DEXPERIENCE AGV logistics simulation (2022); CATIA Behavior/Dymola system-simulation MBSE at HELLA (2023); xGenerative design "productive" at HELLA — parametric/algorithmic design, i.e. optimization technology, not learning AI12. [Evidence] 198
  • Announced, unverified in engineering: generative-AI-driven R&D cost halving (2023); GenAI inside SIMPLIFY is aimed at administration, not engineering. [Evidence] 35

Timeline of AI Evolution

Date Event
Jul 21, 2015 Faurecia Pune Moldex3D case published — tool tuning −68% (classical simulation)1
2021 Hexagon publishes Faurecia Brières ODYSSEE ML crash-surrogate case study2
Jan 2022 Faurecia takes control of HELLA; FORVIA name adopted11
Apr 27, 2022 Faurecia digital-factory AGV logistics on 3DEXPERIENCE (simulation, not AI)9
Oct 19, 2023 HELLA presents CATIA V5/3DEXPERIENCE MBSE headlamp engineering at 3DS conference8
Dec 18, 2023 Koller to Reuters: halve R&D costs by 2028 "mainly thanks to generative AI"; 5 dark plants3
May 2024 LTTS strategic partnership (~€45m) — Clean Mobility engineering outsourcing10
Dec 2, 2024 Martin Fischer announced as next CEO (effective Mar 1, 2025)4
Feb 24, 2026 FY2025 results: €2.1bn net loss; SIMPLIFY (€110m, "partly GenAI"); IGNITE CMD56
Apr 27, 2026 Interiors business group sale to Apollo funds announced (€1.82bn EV)7

Products Relevant to Engineering

Forvia is a parts maker, not a software vendor, so this section maps its engineering toolchain rather than products it sells. [Synthesis]

Where the moulded parts sit — Forvia's injection-moulded output spans Interiors (instrument panels, door panels, consoles — large visible mouldings; the Interiors division is being sold to Apollo, 2026), Seating (structures, foams, plastic trims and mechanism covers), and HELLA Lighting (lenses, light guides, reflectors, housings — precision optics moulding; impaired in 2025). Clean Mobility is mostly metallic, with engineering partially outsourced to LTTS (2024). [Evidence] 7810

CATIA V5 (Dassault Systèmes) — the primary CAD across the group: ~2,500 licences group-wide (Dec 2022); HELLA Mohelnice tracked releases R19→R32. HELLA runs CATIA V5 in co-existence with the 3DEXPERIENCE platform (11 roles / 106 applications, on-premise), transferring V5 headlamp geometry into 3DEXPERIENCE for system-level behaviour modelling. [Evidence] 8

3DEXPERIENCE apps at HELLA — CATIA Behavior Modeling ("productive"), Dymola-based system simulation of headlamp levelling/actuation, xGenerative Design ("productive") and native MCAD ("pilot"). Note the honest-AI reading: xGenerative Design is algorithmic/parametric pattern generation — optimization technology wearing a generative label, not a learning system12. [Evidence] 8

Moldex3D (CoreTech) — verified in use at Faurecia Interior Systems India (Pune) for filling/packing/cooling/warpage of interior mouldings (2015 case). Classical physics simulation. [Evidence] 1

ODYSSEE CAE (then Hexagon, now Cadence) — machine-learning reduced-order modelling used by Faurecia's seating validation organisation for crash prediction. The one genuine AI tool with verified Forvia usage. [Evidence] 2

DELMIA / 3DEXPERIENCE manufacturing — AGV inbound-logistics simulation in the digital-factory programme (2022). [Evidence] 9

AI Capabilities

Capabilities are read customer-side: what AI does Forvia demonstrably use or build?

1. ODYSSEE ML crash surrogates (Brières Tech Center, 2021) — the report's central exhibit. Hexagon's published case study documents Faurecia's "Cockpit of the Future" seating validation: four severe load cases (front crash, rear crash, 18 kg luggage impact, seatbelt anchorage). Reduced-order models (ROMs) were trained on 60 finite-element simulations each; the seatbelt-anchorage ROM was even trained from the front-crash ROM rather than fresh FEA. Results: front crash predicted in 1 minute vs 11 hours of FEA (on an 8-core machine vs a 28-core cluster), seatbelt anchorage 27 min vs 10 h, rear crash 1 h vs 14 h — >97% time reduction — with front-crash displacement within 1 mm of FEA (below the ~5 mm optical-tracking uncertainty of physical tests). Named engineer: Steeve Lainé, Calculation/Simulations Engineer, FAS/R&D/Global Validation Services: "Hard test results can be used to predict a different test configuration thanks to R.O.M." This is real machine learning — but taught by simulations, not by production or test-field outcomes. [Evidence] 2

2. Moldex3D mould-flow simulation (Pune, 2015) — quantified, but not AI. Faurecia Interior Systems India (Pune) used Moldex3D eDesign on a centre-console fascia: filling/packing/cooling/warpage analysis led to a second hot nozzle, an added cold runner with submarine gates, and local thickness changes. Reported outcomes: tool-tuning costs down 68%, weld lines relocated from visible surfaces to hidden corners, Y/Z warpage brought within gap-and-flush (the visible gaps and surface alignment between adjacent panels) tolerance, scrap "negligible." This is the solver earning its keep — deterministic physics, no learning. It matters here as the quantified upstream bracket of the workflow, and because it resolves an open question in this study (see §7). [Evidence + Marketing — vendor-published] 1

3. GenAI in SIMPLIFY (2026). Administrative cost reduction "partly through the use of GenAI" — €110m target by 2028. No engineering content disclosed. [Evidence — thin] 56

4. Dark plants (announced 2023). Five fully automated plants within two years, from one (Spain). Automation, not necessarily learning AI; no 2025–26 status update found. [Evidence — announced, stale] 3

5. Product-side AI (out of scope but noted). Forvia's in-vehicle software arm (Appning app ecosystem, with Microsoft; HELLA ADAS electronics) uses AI in the product, not in moulded-part engineering. Undated/thin — see Appendix C. [Evidence — thin]

What was searched for and not found: any Forvia/Faurecia/HELLA use of AI or ML for DFM checks, mould design, warpage prediction, moulding-process optimization, or geometry retrieval over its part libraries. Nothing public exists as of August 2026. [Evidence of absence]

Engineering Workflow Contribution

A Forvia moulded part (say, a console fascia or a headlamp lens) flows:

OEM requirement → styling/concept → CATIA V5 detailed design → structural/ system simulation (Abaqus-class FEA; Dymola system models at HELLA) → mould flow in a commercial solver (Moldex3D verified) → tool design (in-house mould shops, esp. HELLA optics) → moulding trials → validation (Global Validation Services — where the ODYSSEE ML surrogates live) → manufacturing engineering (3DEXPERIENCE digital factory) → production.

The AI content today is a single island: ML surrogates in crash validation. Everything moulding-specific between CAD and trials runs on classical physics and human judgement. [Synthesis] 812

Public Customer Evidence

Here Forvia is the customer — these are the vendor-published cases with Forvia as subject.

Case Studies

  • Faurecia Interior Systems India, Pune — Moldex3D (published Jul 21, 2015). Centre-console fascia; details and −68% tool-tuning figure in §5.2. Study correction: the Moldex3D report in this study had listed this case as "world's leading automotive part supplier — Faurecia identification unverified." Direct fetch of the case page confirms it names Faurecia Interior Systems India Pvt. Ltd., Pune, explicitly — the identification is now verified (see the Moldex3D report). [Evidence] 1
  • Faurecia Brières Tech Center — Hexagon ODYSSEE (2021, Hexagon-published PDF). Full numbers in §5.1. Study correction: the Hexagon report carried the Faurecia >97% figure as "found only on a partner/distributor page." A Hexagon-branded primary case-study PDF (© 2021 Hexagon) exists with the full data and a named Faurecia engineer — the evidence grade upgrades from partner-reported to vendor-primary. [Evidence] 2
  • Faurecia digital factory — Dassault Systèmes (Apr 27, 2022). AGV inbound-logistics simulation with "On Target Vehicle Launch" on 3DEXPERIENCE; claimed bottleneck discovery before launch, floor-space and cycle-time gains (unquantified); tied to a 15% factory-energy-reduction goal. Simulation, not AI. [Evidence + Marketing] 9
  • HELLA headlamp MBSE — Dassault conference deck (Oct 19, 2023). Primary engineering-practice evidence: CATIA V5 primary CAD, 3DEXPERIENCE behaviour modelling, Dymola system simulation, licence counts, and the Mohelnice moulding technology list (thick-wall optics, 1K/2K lens automation). No AI anywhere in the deck. [Evidence] 8

Conference Demonstrations

The HELLA deck above (3DS Conference, Darmstadt). No Forvia demonstration of AI in moulded-part engineering was found at any conference. [Evidence of absence]

White Papers / Testimonials

The two vendor case studies above are the only quantified ones found. Note both pre-date the generative-AI era (2015, 2021) — Forvia's quantified digital evidence is solver-and-surrogate vintage, not GenAI vintage. [Synthesis]

Technical Architecture (Inferred)

Evidence

  • CATIA V5 primary CAD; 3DEXPERIENCE co-existence (11 roles/106 apps, on-premise at HELLA); V5→3DX data transfer for system simulation.8
  • Moldex3D in at least the Indian interiors engineering centre (2015).1
  • ODYSSEE CAE in seating validation (Brières, 2021).2
  • DELMIA-side 3DEXPERIENCE for factory logistics (2022).9

Synthesis

A classical, Dassault-centred Tier-1 stack: CATIA V5 authoring, PLM plus KBE (knowledge-based engineering) apps around it, commercial solvers per domain (optics, structures, moulding), with ML appearing only where a vendor product (ODYSSEE) brought it in. There is no sign of a Forvia-built AI platform for engineering. [Synthesis]

Inference

Given ~2,500 CATIA seats and decades of programmes, Forvia sits on one of the industry's largest historical moulded-part CAD libraries — exactly the raw material for design-reuse retrieval — and no public evidence suggests it is mined by any learning system. The 2023 "halve R&D with GenAI" ambition would have required precisely such infrastructure; its absence from all subsequent communications suggests it was never built. [Inference]

AI Technologies

In one list, with the honest test applied: reduced-order-model ML (ODYSSEE — real, simulation-taught, vendor-supplied); generative AI (announced for R&D 2023, redirected to administration 2026 — no engineering deployment evidenced); algorithmic generative design (xGenerative Design at HELLA — optimization technology, not learning12); simulation everywhere else (Moldex3D, FEA, Dymola, DELMIA — physics, not AI); product AI (in-vehicle software, ADAS — out of scope). [Synthesis]

Research Publications

Papers

No Forvia/Faurecia/HELLA-authored publication applying AI/ML to injection moulding, DFM or mould design was found. HELLA's public engineering material (the 2023 deck) is MBSE, not ML. [Evidence of absence] 8

Patents

The group holds 14,000+ patents (2022), overwhelmingly product patents (seating mechanisms, lighting, electronics). No AI-for-moulding patent family was identified in this research pass (light search — see Appendix C). [Evidence — count; absence lightly searched] 8

Standards

Nothing relevant found.

Open Source

No significant Forvia open-source AI footprint was found in this research pass. This was a light search (no exhaustive GitHub/Hugging Face audit) — treat as provisional, see Appendix C. [Evidence of absence — light]

Engineering Service / Platform Mapping

  • CATIA V5 — primary CAD, group-wide (~2,500 licences).8
  • 3DEXPERIENCE — on-premise at HELLA (behaviour modelling, xGenerative design); DELMIA-side factory simulation at Faurecia.89
  • Moldex3D — mould-flow (verified at Pune interiors engineering).1
  • ODYSSEE CAE (Hexagon → Cadence, Feb 2026) — ML crash surrogates in seating validation.2
  • L&T Technology Services — outsourced Clean Mobility engineering (May 2024, ~€45m/5yr) — the ESP tier entering where AI was promised.10

Engineering Intelligence Stack Mapping

  1. Intent — OEM requirements cascade; no AI capture of design intent. [Evidence of absence]
  2. Knowledge — KBE apps and start templates around CATIA at HELLA; decades of programme data in PLM; no evidence any of it feeds a learning system. [Evidence] 8 [Inference]
  3. Reasoning — classical solvers (mould flow, FEA, optics) plus one real ML reasoning layer: ODYSSEE crash surrogates — simulation-taught. [Evidence] 2
  4. Execution — CATIA authoring, in-house mould shops, automated moulding lines (Mohelnice 1K/2K lens automation); automation, not AI. [Evidence] 8
  5. Feedback — the industry-wide gap, here too: no evidence that measured production outcomes (scrap, warpage, tool-tuning history) feed back into any learning system — even though Faurecia's own 2015 Pune case shows exactly the kind of outcome data (tool-tuning cost, defect relocation) such a loop would need. [Evidence of absence] [Synthesis]

Strengths

  • The only Tier-1 studied with quantified digital outcomes at both ends of the moulded-parts workflow — mould-flow (2015) and ML crash surrogates (2021). [Evidence] 12
  • Massive moulded-parts engineering scale: 15,000 R&D engineers, ~2,500 CATIA seats, in-house precision-optics moulding at HELLA. [Evidence] 8
  • Demonstrated willingness to adopt vendor ML where ROI is proven (ODYSSEE). [Evidence] 2
  • CEO-level cost pressure creates a genuine pull for engineering AI — the motive exists even where the capability doesn't. [Inference]

Weaknesses

  • Financial distress shapes everything: €2.1bn net loss (2025), €6bn net debt, dividend suspended, businesses being sold. AI investment competes with survival spending. [Evidence] 56
  • The AI mandate outlived its CEO: the 2023 "halve R&D via GenAI" target has no visible successor programme in engineering; the R&D cost lever became outsourcing. [Evidence] 10 [Evidence of absence]
  • Selling the Interiors crown jewel removes the largest moulded-parts business — and its engineering data — from the group. [Evidence] 7
  • All demonstrated AI is vendor-supplied and simulation-taught; no in-house engineering-AI capability is evidenced. [Synthesis]

Current Gaps (largely manual today)

DFM checking of mouldings, mould/tool design, mould-flow interpretation, moulding-trial tuning (the exact activity the Pune case spent −68% less on — still human decisions guided by simulation), and reuse of historical part/tool knowledge across programmes. No public AI touches any of these at Forvia. [Evidence of absence] [Synthesis]

Future Direction

  • IGNITE strategy (Feb 2026): focus on "high value-added, technology-driven activities," deleverage (1.5× by end-2026), capex capped below 7% of sales; Interiors sale proceeds to cut debt by €1bn+. [Evidence] 576
  • SIMPLIFY carries the GenAI flag — into administration, not engineering. [Evidence] 5
  • My read: the 2023 vision (AI transforms R&D) has been quietly replaced by a 2026 reality (sell, outsource, restructure). If engineering AI returns to Forvia, it will most likely arrive the way ODYSSEE did — bought from a vendor with a proven number — not built. That makes Forvia a buyer profile for exactly the opportunity categories this study tracks. [Inference]

Relevance to Automotive Moulded Parts

Capability Forvia position AI?
Plastic Part Design Vast (interiors — exiting; seating trims; lighting housings) No
Surface Design (Class A — styled, customer-visible show surfaces) In-house (cockpit surfaces) No
CAD Automation KBE apps, templates, xGenerative design Optimization, not AI12
DFM Internal know-how; nothing public No — white space
Tool Design In-house mould shops (esp. HELLA optics) No
Mould Flow Commercial solvers (Moldex3D verified) No — classical physics1
Crash/Structural CAE Global Validation Services Yes — ODYSSEE ML surrogates2
Manufacturing Engineering Digital factory, AGV simulation, dark-plant ambition Simulation/automation9
Quality Standard Tier-1; no AI evidence No
Knowledge Reuse PLM archives; no retrieval/learning system evidenced No — white space

Critical finding. Forvia answers this report's research question — which component-engineering workflows benefit most from AI? — with unusual clarity, because it has quantified numbers on both sides of the line. Where AI works (and is bought): expensive, repetitive, simulation-bound validation — crash CAE, where a surrogate trained on 60 FEA runs pays back instantly (>97% time saved, 1 mm accuracy). Where AI is absent: the moulded-part engineering core — DFM, mould design, warpage — where the quantified value (−68% tool tuning) still comes from classical physics simulation plus expert judgement. The learning loop that would connect them (production outcomes feeding design rules) does not exist here, exactly as it does not exist at any vendor in this study. [Synthesis] 21

For design-reuse retrieval: Forvia is the strongest Tier-1 validation found. It owns a huge historical CATIA part library, demonstrably pays for simulation-taught ML when a vendor proves the number, is under existential cost pressure with an unmet AI mandate — and has no retrieval or rules-automation layer over its own data. [Synthesis]

Key Takeaways

  1. Forvia is the only Tier-1 with verified quantified outcomes at both ends of the moulded-parts workflow: Moldex3D mould-flow (Pune 2015, tool tuning −68%) and ODYSSEE ML crash surrogates (Brières 2021, 1 min vs 11 h, <1 mm). [Evidence] 12
  2. The mould-flow win is classical simulation, not AI; the crash win is real ML, but simulation-taught — consistent with the study's cross-vendor spine (nobody learns from measured production outcomes). [Synthesis]
  3. No AI touches Forvia's moulded-part engineering — DFM, mould design and warpage are AI-free in all public evidence. [Evidence of absence]
  4. The 2023 CEO mandate to halve R&D costs by 2028 "mainly thanks to generative AI" (Koller, Reuters) has no visible engineering successor; under CEO Fischer (from Mar 2025) the GenAI flag flies over administrative costs (SIMPLIFY, €110m), and R&D costs are cut via outsourcing (LTTS) and restructuring (EU-FORWARD). [Evidence] 3510
  5. Forvia is financially strained — €2.1bn FY2025 net loss, €6bn net debt, dividend suspended — and sold its Interiors (largest moulded-parts) business to Apollo funds (Apr 2026, €1.82bn EV). [Evidence] 57
  6. The engineering stack is Dassault-centred and classical: CATIA V5 (~2,500 seats), 3DEXPERIENCE co-existence, MBSE at HELLA — no in-house engineering-AI platform. [Evidence] 8
  7. Two findings confirmed here: the Moldex3D "world's leading automotive part supplier" case is confirmed as Faurecia (Pune) — see the Moldex3D report; and the Faurecia ODYSSEE >97% figure exists in a Hexagon-primary 2021 PDF, strengthening the sourcing in the Hexagon report. [Evidence] 12
  8. Forvia's buying behaviour (adopt vendor ML when a hard number is proven) defines the sales motion for moulding-AI vendors targeting Tier-1s. [Inference]

References

Primary Sources — Forvia / Faurecia / HELLA

  • Martin Fischer to succeed Patrick Koller as CEO (effective Mar 1, 2025), Dec 2, 2024 — https://www.forvia.com/en/press/martin-fischer-succeed-patrick-koller-ceo-forvia-effective-march-1-2025
  • FORVIA 2025 annual results, Feb 24, 2026 — https://www.forvia.com/en/press/2025-annual-results
  • FORVIA to sell its Interiors business group to Apollo funds, Apr 27, 2026 — https://www.forvia.com/en/press/forvia-sell-its-interiors-business-group-apollo-funds
  • HELLA engineering deck: "Engineering and simulation of Vehicle Headlamp systems at HELLA" (Pinďák & Cieslar, 3DS conference Darmstadt, Oct 19, 2023, PDF) — https://events.3ds.com/sites/default/files/engineering-and-simulation-of-vehicle-headlamp-systems-tribe.pdf
  • FORVIA–LTTS strategic partnership PDF, May 2024 — https://www.forvia.com/sites/default/files/2024-05/FORVIA-LTTS%20STRATEGIC%20PARTNERSHIP.pdf

Vendor-Published Customer Evidence (Forvia as customer)

  • Moldex3D customer success: "World's Leading Automotive Part Supplier…" (identifies Faurecia Interior Systems India Pvt. Ltd., Pune), Jul 21, 2015 — https://www.moldex3d.com/blog/customer_success/worlds-leading-automotive-part-supplier-utilizes-moldex3d-analyses-to-make-confident-decisions-on-product-design-optimization/
  • Hexagon case study PDF: "Reduced order modeling of advanced versatile seats — Faurecia Brières Tech Center" (© 2021 Hexagon; mirror at FeaturedCustomers) — https://cdn.featuredcustomers.com/CustomerCaseStudy.document/hexagon_mi_case_study_faurecia_a4_2021_web_1.pdf
  • Dassault Systèmes press release: Faurecia AGV logistics on 3DEXPERIENCE, Apr 27, 2022 — https://www.3ds.com/newsroom/press-releases/advancing-its-digital-factory-program-faurecia-optimizes-automated-guided-vehicle-logistics-dassault-systemes-3dexperience-platform

Secondary Sources

  • Reuters: "Forvia aims to halve its R&D costs by 2028 thanks to AI — CEO," Dec 18, 2023 — https://www.reuters.com/business/autos-transportation/forvia-aims-halve-its-rd-costs-by-2028-thanks-ai-ceo-2023-12-18/ (mirror consulted: https://finance.yahoo.com/news/forvia-aims-halve-r-d-165457099.html)
  • Reuters: Forvia to sell Interiors to Apollo, Apr 27, 2026 — https://www.reuters.com/legal/transactional/forvia-sell-its-interiors-business-apollo-funds-21-billion-2026-04-27/
  • briefglance.com FY2025 results analysis, Feb 24, 2026 — https://briefglance.com/articles/forvias-billion-euro-bet-why-a-massive-loss-signals-a-stronger-future
  • LTTS press release: FORVIA–LTTS partnership — https://www.ltts.com/press-release/FORVIA-LTTS-sign-strategic-partnership-ultralow-emissions-engineering
  • Wikipedia: Forvia (corporate facts) — https://en.wikipedia.org/wiki/Forvia

Appendix A — Timeline

See §3. Key dates: Moldex3D Pune case (Jul 2015) → ODYSSEE Brières case (2021) → HELLA takeover / FORVIA (2022) → AGV digital factory (Apr 2022) → HELLA MBSE deck (Oct 2023) → Koller GenAI/R&D mandate (Dec 2023) → LTTS outsourcing (May 2024) → Fischer named CEO (Dec 2024, effective Mar 2025) → FY2025 loss + SIMPLIFY + IGNITE (Feb 2026) → Interiors sale to Apollo (Apr 2026).

Appendix B — Glossary

  • Tier-1 supplier — a company selling systems/components directly to car makers (OEMs); Forvia is one of the largest.
  • Reduced-order model (ROM) — a small, fast mathematical model (here: ML-trained) that reproduces the input→output behaviour of a big simulation.
  • FEA — finite-element analysis; the standard physics simulation for structures and crash.
  • ODYSSEE CAE — the ML/ROM surrogate tool Faurecia used (CADLM → Hexagon 2021 → Cadence Feb 2026).
  • Moldex3D eDesign — CoreTech's injection-moulding simulation package (filling/packing/cooling/warpage).
  • 1K / 2K moulding — single-material vs two-material (two-shot) injection moulding; HELLA runs automated 2K lens lines.
  • Thick-wall optics — very thick transparent mouldings (lens bodies) — among the hardest parts to mould without defects.
  • MBSE — model-based systems engineering: system behaviour modelled and simulated (e.g. Dymola) alongside 3D CAD.
  • KBE — knowledge-based engineering: rule/template automation inside CAD.
  • EU-FORWARD / SIMPLIFY / IGNITE — Forvia's European restructuring programme, administrative cost programme ("partly GenAI"), and Feb-2026 group strategy, respectively.
  • Dark plant — a factory automated enough to run with lights off.
  • DFM — design for manufacturability: checking a design against moulding rules (draft, wall thickness, undercuts…).

Appendix C — Notes (thinnest-evidence areas to revisit)

  1. Status of the "halve R&D by 2028 via GenAI" target — no reaffirmation or formal retraction found under Fischer; classified here as quietly superseded. Revisit after the next investor communication.
  2. Dark-plants progress — the five-plants-in-two-years plan (Dec 2023) has no found 2025–26 status update.
  3. Appning × Microsoft AI collaboration — forvia.com press URL located but page unfetchable (403) and date unverified; product-side (in-vehicle apps), not engineering. Kept out of the body.
  4. FY2025 exact figures — forvia.com results page and PDF blocked fetching (403); figures cross-checked via the press-page search snippet and one secondary analysis (briefglance). Confirm decimals against the primary PDF before final publication.
  5. Patents and open source — only lightly searched; "no AI-for-moulding patents / no OSS footprint" is provisional, not established.
  6. HELLA optics-moulding AI — academic literature exists on ML for moulded PC lenses generally, but none found with HELLA authorship; the negative is as-searched, not exhaustive.
  7. Seating-division figures (42,515 employees, 77 locations, 13 research centres) come from the 2021 Hexagon case study and are dated.
  8. Interiors sale closing — announced Apr 27, 2026, targeted to close end-2026, pending approvals; not yet closed as of this report.

Executive Summary

  • Who they are. Forvia is the world's seventh-largest automotive supplier, formed when France's Faurecia took control of Germany's HELLA in 2022. It employs ~157,000 people, runs 76 R&D centres with 15,000+ R&D engineers, and supplies seats, interiors (dashboards, door panels, consoles), exhaust systems, lighting and electronics — "one in two vehicles worldwide" carries a Forvia part. [Evidence + Marketing] 811
  • Why they matter for this study. Forvia is the only Tier-1 studied with verified, quantified outcomes at both ends of the moulded-parts workflow: a mould-flow simulation case with hard numbers (Faurecia Pune, Moldex3D, 2015 — tool-tuning cost −68%) and a machine-learning crash-surrogate case with hard numbers (Faurecia Brières, Hexagon ODYSSEE, 2021 — front-crash prediction in 1 minute vs 11 hours of finite-element analysis, within 1 mm). [Evidence] 12
  • A CEO-level AI mandate — that moved. In December 2023 then-CEO Patrick Koller told Reuters Forvia would halve its R&D costs by 2028, "mainly thanks to generative AI." Koller stepped down in March 2025; under new CEO Martin Fischer the public AI language survives mainly in an administrative cost programme (SIMPLIFY, €110m by 2028, "partly through GenAI"), while R&D costs are being cut the classical way — restructuring (EU-FORWARD) and outsourcing engineering to L&T Technology Services. No 2025–26 reaffirmation of the "halve R&D via AI" target was found. [Evidence] 34510 [Evidence of absence]
  • The central honest finding. Across everything public, no AI touches Forvia's moulded-part engineering itself — no AI for DFM, mould design, or warpage. The quantified wins bracket the moulding stage: classical physics simulation upstream (Moldex3D — not AI) and genuine ML surrogates downstream (crash CAE — real ML, but trained on simulations, not on moulding or production outcomes). The moulded part in the middle is still engineered by humans with CATIA V5 and classical solvers. [Synthesis]
  • A structural caveat. In April 2026 Forvia agreed to sell its Interiors business — its largest visible-plastics operation — to Apollo funds (€1.82bn enterprise value), after a €2.1bn net loss for 2025. The group that generated the Pune mould-flow evidence is leaving the group. [Evidence] 75

  1. Moldex3D customer-success blog, "World's Leading Automotive Part Supplier Utilizes Moldex3D Analyses to Make Confident Decisions on Product Design Optimization," Jul 21, 2015 — names Faurecia Interior Systems India Pvt. Ltd. (Pune); centre-console fascia; tool-tuning cost −68%, weld lines relocated, Y/Z warpage within tolerance — https://www.moldex3d.com/blog/customer_success/worlds-leading-automotive-part-supplier-utilizes-moldex3d-analyses-to-make-confident-decisions-on-product-design-optimization/ ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  2. Hexagon case study (PDF, © 2021 Hexagon), "Reduced order modeling of advanced versatile seats — Faurecia Brières Tech Center": ROMs trained on 60 FEA runs; front crash 1 min vs 11 h FEA (<1 mm deviation), seatbelt anchorage 27 min vs 10 h, rear crash 1 h vs 14 h; >97% time reduction; quote by Steeve Lainé (Calculation/Simulations Engineer, FAS/R&D/Global Validation Services) — https://cdn.featuredcustomers.com/CustomerCaseStudy.document/hexagon_mi_case_study_faurecia_a4_2021_web_1.pdf ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  3. Reuters, "Forvia aims to halve its R&D costs by 2028 thanks to AI — CEO," Dec 18, 2023 (Koller quotes; 15,000+ engineers, 76 R&D centres; five "dark plants") — https://www.reuters.com/business/autos-transportation/forvia-aims-halve-its-rd-costs-by-2028-thanks-ai-ceo-2023-12-18/ (mirror: https://finance.yahoo.com/news/forvia-aims-halve-r-d-165457099.html) ↩↩↩↩↩↩↩↩

  4. FORVIA press release, "Martin Fischer to succeed Patrick Koller as CEO of FORVIA, effective March 1, 2025," Dec 2, 2024 — https://www.forvia.com/en/press/martin-fischer-succeed-patrick-koller-ceo-forvia-effective-march-1-2025 ↩↩↩↩

  5. FORVIA 2025 annual results, Feb 24, 2026 (sales €26.2bn; operating margin 5.6%; net cash flow €962m; net debt 1.7× EBITDA; EU-FORWARD €165m savings; SIMPLIFY €110m "partly through the use of GenAI") — https://www.forvia.com/en/press/2025-annual-results ↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  6. briefglance.com, "Forvia's Billion-Euro Bet," Feb 24, 2026 (net loss €2.1bn; write-downs €1.85bn incl. €920m Lighting/Electronics impairment, €578m Interiors capital loss, €209m Symbio; dividend suspension) — secondary analysis of the FY2025 results — https://briefglance.com/articles/forvias-billion-euro-bet-why-a-massive-loss-signals-a-stronger-future ↩↩↩↩↩↩↩

  7. FORVIA press release, "FORVIA to sell its Interiors business group to Apollo funds," Apr 27, 2026 (€1.82bn enterprise value; closing targeted end-2026) — https://www.forvia.com/en/press/forvia-sell-its-interiors-business-group-apollo-funds; Reuters coverage — https://www.reuters.com/legal/transactional/forvia-sell-its-interiors-business-apollo-funds-21-billion-2026-04-27/ (Apollo Global Management — New York, USA; apollo.com) ↩↩↩↩↩↩↩↩

  8. Pinďák & Cieslar (FORVIA HELLA), "Engineering and simulation of Vehicle Headlamp systems at HELLA," 3DS conference, Darmstadt, Oct 19, 2023 (PDF): CATIA V5 primary CAD; ~1,030 V5 licences/3,500+ users at HELLA and ~2,500 group-wide (Dec 2022); Mohelnice moulding technologies; 3DEXPERIENCE 11 roles/106 apps; Dymola MBSE; xGenerative Design "productive" — https://events.3ds.com/sites/default/files/engineering-and-simulation-of-vehicle-headlamp-systems-tribe.pdf ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  9. Dassault Systèmes press release, "Advancing its Digital Factory Program, Faurecia Optimizes Automated Guided Vehicle Logistics with … 3DEXPERIENCE," Apr 27, 2022 — https://www.3ds.com/newsroom/press-releases/advancing-its-digital-factory-program-faurecia-optimizes-automated-guided-vehicle-logistics-dassault-systemes-3dexperience-platform ↩↩↩↩↩↩↩↩

  10. FORVIA–L&T Technology Services strategic partnership, May 2024 (~€45m, 5 years, Clean Mobility engineering, ~300 engineers, Germany + India) — https://www.ltts.com/press-release/FORVIA-LTTS-sign-strategic-partnership-ultralow-emissions-engineering; FORVIA PDF — https://www.forvia.com/sites/default/files/2024-05/FORVIA-LTTS%20STRATEGIC%20PARTNERSHIP.pdf ↩↩↩↩↩↩↩

  11. Wikipedia, "Forvia" (formation, HQ Nanterre, business lines, customers, 7th-largest supplier) — https://en.wikipedia.org/wiki/Forvia ↩↩↩↩

  12. On why algorithmic "generative design" is optimization technology, not learning AI, see the concept note What is Optimization technology. ↩↩↩↩↩