Skip to content

Volkswagen AG (Volkswagen Group)

Company Overview

Business Overview

Volkswagen AG, headquartered in Wolfsburg, Germany, is Europe's largest automaker and one of the largest in the world by revenue, operating roughly ten vehicle brands plus financial-services and software subsidiaries. It is listed in Frankfurt (DAX). Oliver Blume is Group CEO (also CEO of Porsche AG for most of the period covered). Exact current revenue and headcount figures were not primary-verified for this report and are deliberately omitted (see Appendix C). [Synthesis]

Automotive Business

The Group's brands span volume (Volkswagen, ŠKODA, SEAT/CUPRA), premium (Audi), sport/luxury (Porsche, Bentley, Lamborghini) and commercial vehicles. Unusually among OEMs, VW retains deep vertical integration: Volkswagen Group Components, an independent business unit within the Group's Technology division, runs dozens of plants making axles, steering systems, battery systems, seats, plastic parts — and the tools and moulds to produce them. [Evidence] 34

The Braunschweig (Brunswick) plant — the oldest VW plant, part of Group Components since 2019 — is the site that matters most for this study:

  • ~6,000 employees on 682,000 m²; components for 5.6 million cars in 2025; supplies every Group brand including Porsche, Bentley and Lamborghini. [Evidence] 3
  • It manufactures "axles, steering mechanisms, plastic parts, battery systems, vehicle components as well as the required tools and machines." [Evidence] 4
  • Its component tool engineering unit (~700 employees, ~100 machine tools, 75+ years of toolmaking history) builds injection moulds — e.g. for bumpers and instrument panels — plus die-cast moulds and sheet-metal compression tools. [Evidence] 4

Engineering Business

VW's engineering is done by the brands' Technical Development organisations (Wolfsburg for VW, Ingolstadt/Neckarsulm for Audi, Weissach for Porsche), by Cariad (Group software), by Porsche Engineering (a Porsche subsidiary selling engineering services, and a visible publisher of engineering-AI work), and by Group Components for parts and tooling. Since 2024 an AI Lab GmbH acts as the Group's AI incubator. [Evidence] 212

Role in Automotive Moulded Parts Workflow

As an OEM with in-house tooling, VW performs (rather than sells tools for) essentially every stage. The table shows where AI demonstrably touches VW's own execution of each stage as of August 2026. [Synthesis]

Workflow Stage VW does it in-house? AI evidence at VW
Requirements Yes (Codebeamer ALM) Yes — Codebeamer Copilot (co-built; vehicle software)6
Industrial Design Yes (brand studios) Yes — FelGAN at Audi (wheel rims, a GAN)9
Concept Design Yes Dassault "AI-powered engineering environment" (in rollout)1
CAD Modelling Yes No public AI evidence
Engineering Review Yes No public AI evidence
Simulation Yes Yes — research-grade: Porsche Engineering RL crash optimization (−80% FEM loops)12
DFM Yes (internal) No public AI evidence
Tool Design Yes — own mould shop No AI — knowledge-based CAM automation, statistical (2010s)4
Mould Flow Yes (internal simulation assumed) No public AI evidence
Prototype Yes (incl. 3D printing centre, 2018)14 Additive manufacturing, not AI
Validation Yes No public AI evidence
Manufacturing Engineering Yes Binder-jetting industrialisation with HP + Siemens (2021) — additive, not AI15
Production Release / Quality Yes Yes — strong: spot-weld AI, weld-splatter detection, paint-shop AI1011

(FelGAN designs wheel rims — cast metal, not moulded; included as a design-AI signal only.)

Reading of the table: VW's AI clusters at the two ends of the workflow — requirements/software upstream, production quality downstream — plus design ideation and simulation research. The moulded-parts core (DFM, tool design, mould flow) shows no public AI at all, despite VW owning the entire chain. [Synthesis]

AI Strategy

Public AI Vision

VW's stated ambition (September 9, 2025): AI as "the key to greater speed, quality, and competitiveness," applied across vehicle development, production, cybersecurity and IT. Headline commitments: [Evidence] 1

  • Up to €1 billion invested in AI by 2030.
  • Expected efficiency gains and cost avoidance of up to €4 billion by 2035.
  • 1,200+ AI applications already active across the Group.
  • Product development cycle target of 36 months or less (~25% faster), explicitly linked to an AI-powered engineering environment built with Dassault Systèmes for virtual testing.
  • A Digital Production Platform (DPP) connecting 40+ manufacturing sites.
  • A "WE & AI" employee programme reaching 130,000+ employees since spring 2024.

Board slogan, from IT board member Hauke Stars: "No process without AI." [Evidence + Marketing] 1

Investor / Executive Statements

  • Hauke Stars (IT board member): "AI is our key to greater speed, quality, and competitiveness. Our ambition: No process without AI." [Marketing] 1
  • Oliver Blume (Group CEO, on founding AI Lab): "We want to offer our customers genuine added value with artificial intelligence." [Marketing] 2
  • Gerd Walker (Audi production board member): "AI is a quantum leap for efficiency in our production... transforming our plants into smart factories." [Marketing] 11

Engineering AI Strategy — the buy / build / co-develop pattern

This is the report's spine. VW's observable behaviour sorts into three modes. [Synthesis]

  1. Buy the platform. Codebeamer ALM from PTC (strategic supplier, October 2023); the 3DEXPERIENCE platform from Dassault Systèmes (February 2025) as "a main engineering and manufacturing platform" for the Volkswagen, Audi and Porsche brands; Microsoft Azure as the AI cloud. VW does not build CAD, PLM, ALM or foundation models itself. [Evidence] 586
  2. Build the application where the data is uniquely VW's. Audi built FelGAN (wheel-design GAN) entirely in-house; Audi/VW built the spot-weld quality AI with research institutes; Porsche Engineering builds reinforcement-learning crash-optimization methods; AI Lab prototypes digital products. [Evidence] 910122
  3. Co-develop where VW wants to shape a vendor's product: Codebeamer Copilot with PTC and Microsoft (December 2024, Azure AI, beta early 2025) — VW is named as a development partner, not a launch customer; and the binder-jetting production process with HP and Siemens (2021). This co-development posture is what distinguishes VW from every other OEM studied. [Evidence] 615

Timeline of AI Evolution

Date Event
~2016 Group Machine Learning Research Lab active in Munich (van der Smagt) — fundamental ML research1617
Dec 11, 2018 Toolmaking division opens 3D-printing centre in Wolfsburg (additive, not AI)14
Jun 2021 Binder jetting industrialisation with HP + Siemens; target 100,000 printed parts/year by 202515
Dec 12, 2022 Audi unveils FelGAN — in-house GAN for wheel-rim design ideation9
Jun 30, 2023 Audi begins roll-out of spot-weld quality AI (Neckarsulm → Brussels, Emden, Ingolstadt)10
Oct 31, 2023 Codebeamer ALM adopted group-wide; PTC named strategic supplier5
Jan 31, 2024 AI Lab GmbH founded — global AI competence centre and incubator2
Dec 3, 2024 Codebeamer Copilot co-development with PTC + Microsoft announced (Azure AI; beta early 2025)6
Feb 4, 2025 Dassault Systèmes partnership — 3DEXPERIENCE as a main engineering & manufacturing platform (VW, Audi, Porsche)8
Feb 19, 2025 Porsche Engineering publishes RL crash-structure optimization (−80% FEM loops)12
Sep 9, 2025 €1bn AI investment announced; "No process without AI"; 36-month dev-cycle target1
Jan 27, 2026 Audi scales production AI: weld-splatter detection to series, paint-shop AI pilots11

Products Relevant to Engineering (demand-side: organisations & tools)

Volkswagen Group Components / Braunschweig component tool engineering — the in-house mould and tool shop (~700 people). Builds injection moulds (bumpers, instrument panels), die-cast moulds and press tools. AI: none found. Its modernisation story is hyperMILL CAM automation (OPEN MIND)19 plus a home-grown statistical knowledge-management method (from a 2011 doctoral thesis) that mines CAM data for tool-selection rules and standardisation. Quantified wins: standardisation cut primary processing time ~35% in its area; automated NC programming cut combustion-chamber-core programming from ~2 hours to minutes. All of this is rules/statistics on top of deterministic CAM — VW itself calls it knowledge management, not AI. The honest-AI test (learn from data vs. compute from equations/rules) places it firmly outside AI, and — credit where due — VW never mislabels it.18 [Evidence — dated (2016 case study)]4

Codebeamer (PTC) + Codebeamer Copilot — bought ALM, co-built AI layer. Copilot drafts requirements and test cases for vehicle-software engineering on Microsoft Azure AI; VW is the named co-development partner (Dec 2024, beta early 2025). Microsoft's customer story describes VW engineers using it in Codebeamer. [Evidence] 67

3DEXPERIENCE (Dassault Systèmes) — announced February 4, 2025 as "a main engineering and manufacturing platform" for the Volkswagen, Audi and Porsche brands: cloud-based, virtual-twin-centred, with "AI-powered solutions" on consistent data streams. VW's September 2025 AI announcement ties this environment to the 36-month development-cycle target. The release names four industry solution experiences but no rollout timeline and no legacy systems being replaced. [Evidence] 81

AI Lab GmbH — a lean incubator company (founded January 31, 2024), a "globally networked competence centre" producing proofs-of-concept with partners in Europe, China and North America. Scope is digital products (speech, infotainment, charging optimization, predictive maintenance) — not mechanical engineering or manufacturing. [Evidence] 2

Audi production AI portfolio — self-built with partners: spot-weld quality AI (roll-out from June 2023), Weld Splatter Detection heading to series at six Ingolstadt plants, ProcessGuardAIn paint-shop anomaly detection (series planned Q2 2026), AI-supported dryer control, and Edge Cloud 4 Production replacing 1,000+ industrial PCs. [Evidence] 1011

Porsche Engineering — the Group's most visible publisher of engineering-AI methods: reinforcement learning + neural surrogates for crash-structure and restraint-system design; LLM use across the development workflow. [Evidence] 1213

AI Capabilities

1. Codebeamer Copilot (co-developed with PTC + Microsoft). Generative AI that drafts requirements and test cases inside Codebeamer ALM. Stage: requirements/validation for vehicle software. Stack: Microsoft Azure AI. Status: announced Dec 3, 2024; beta early 2025. Limitation: nothing to do with mechanical parts or moulding — but it is the only OEM co-built engineering-AI product found in this study. [Evidence] 67

2. FelGAN (Audi, in-house). A generative adversarial network — generator vs. discriminator trained until rim designs are photorealistic — used as an ideation tool by wheel designers; designs carry a reproducible "DNA" code; cloud-hosted. Stage: industrial design. Limitation: inspiration only — a human designs the final wheel; no quantified outcome published; and note a wheel is cast metal, not a moulded plastic part. Announced December 12, 2022; built by Audi IT + Design jointly. [Evidence] 9

3. Spot-weld quality AI (Audi → Group roll-out). Analyses ~1.5 million resistance spot welds per shift across 300 vehicles at Neckarsulm, replacing random ultrasound sampling (~5,000 welds/vehicle); developed with DGQ, Fraunhofer IAO and IPA20, including an audit-proof AI process guide. Rolled out toward Audi Brussels, VW Emden, Audi Ingolstadt from 2023. Stage: production quality. Limitation: body-shop welding — not moulding; no defect-rate figures published. [Evidence] 10

4. Audi production-AI scale-up (Jan 2026). Weld Splatter Detection into series production (six Ingolstadt plants planned); ProcessGuardAIn real-time paint-process anomaly detection (series Q2 2026); AI dryer-control energy pilots; 1,000+ industrial PCs eliminated via Edge Cloud 4 Production. Stage: manufacturing/quality. Limitation: pilots and roll-outs, quantified mostly in infrastructure terms. [Evidence] 11

5. Reinforcement-learning crash optimization (Porsche Engineering). An RL agent plus neural surrogate explores crash-structure variants (side sills, restraint systems); reported result: 80% fewer finite-element calculation loops in a real development task; surrogate answers in seconds vs. up to 72 h per full crash simulation. Stage: simulation. Limitation: published as methodology/magazine material, not a shipping product. [Evidence] 12

6. LLMs in the development workflow (Porsche Engineering). Described uses of large language models across vehicle-development tasks (2025). Limitation: descriptive article; no named product or metrics. [Evidence — thin] 13

7. Group AI platform & DPP. 1,200+ AI applications; Digital Production Platform across 40+ sites; "AI-powered engineering environment" with Dassault. Limitation: aggregate marketing numbers — individual applications mostly unnamed. [Evidence + Marketing] 1

8. Tool-engineering knowledge system (Braunschweig) — explicitly NOT AI. Statistical mining of CAM data for tool-selection rules, part-family standardisation (rib milling, electrode machining, mould finishing) and push-button NC-program generation via the hyperMILL API. Honest label, honest mechanism: rules and statistics, deterministic, learns nothing autonomously. Included here because it is exactly the layer AI would have to replace or augment.18 [Evidence — dated] 4

What was searched for and not found: any VW/Audi/Porsche AI applied to injection-moulding simulation, mould design, DFM checking, or plastics engineering; any AI initiative named inside Volkswagen Group Components. Searches combining "Volkswagen Group Components" with AI terms return nothing. [Evidence of absence]

Engineering Workflow Contribution

VW executes the full moulded-part chain internally for major interior/exterior parts: styling in brand studios → part engineering in Technical Development → DFM and tool design in Group Components' tool engineering (Braunschweig) → mould build in the same unit → moulding at Braunschweig and other component plants → assembly at vehicle plants. [Synthesis] 34

The AI overlay, as shipped or announced, touches: requirements (Codebeamer Copilot), design ideation (FelGAN), crash simulation (Porsche RL, research), and production quality (weld/paint AI). The tooling-and-moulding middle is untouched by AI and runs on CAM automation + knowledge rules. [Synthesis]

Public Customer Evidence (VW as the customer: AI inside VW engineering)

Case Studies

  • Codebeamer at Group scale — VW adopted Codebeamer ALM for next-generation vehicle-software development (requirements, test, release); PTC named a "strategic supplier for our Group-wide product development strategy" (October 31, 2023). [Evidence] 5
  • Codebeamer Copilot co-development — PTC + Microsoft + VW Group (December 3, 2024; Azure AI; beta early 2025); Microsoft's customer story shows VW engineers using Copilot in Codebeamer. Strong, primary-sourced — but vehicle software, not moulded parts. [Evidence] 67
  • Dassault 3DEXPERIENCE adoption — announced February 4, 2025 for VW, Audi and Porsche brands; "virtual twin" and AI-on-consistent-data framing; no quantified outcome yet. [Evidence + Marketing] 8
  • Audi spot-weld AI — the Group's most concrete production-AI number: 1.5 million welds analysed per shift vs. sampled ultrasound checks of ~5,000 welds per vehicle. [Evidence] 10
  • Braunschweig tool engineering (pre-AI baseline) — standardisation −35% primary processing time; NC programming 2 h → minutes. Vendor-published (OPEN MIND, 2016) — a CAM-automation win, not AI. [Evidence — vendor-published] 4

Conference Demonstrations

CES 2024 (AI Lab context, in-car ChatGPT era — digital products); the co-development announcements above were press events rather than demos. [Synthesis]

White Papers / Testimonials

The gap: no public VW story combines AI + moulded plastic parts or mould engineering + a quantified outcome. The nearest items are the Braunschweig CAM story (moulds, quantified, but not AI) and the spot-weld AI (AI, quantified, but welding). [Evidence of absence]

Technical Architecture (Inferred)

What the sources establish

  • Microsoft Azure AI powers Codebeamer Copilot. [Evidence] 6
  • Dassault 3DEXPERIENCE on the cloud is to become a main engineering and manufacturing platform for three brands. [Evidence] 8
  • A Digital Production Platform connects 40+ plants; Audi's Edge Cloud 4 Production virtualises shop-floor control (1,000+ PCs removed). [Evidence] 111
  • Audi's FelGAN runs cloud-based with a Streamlit front-end — pragmatic internal tooling, not productised software. [Evidence] 9

Putting it together

VW's AI architecture is federated, partner-anchored: hyperscaler AI (Microsoft) + engineering platforms (PTC, Dassault) + self-built point applications per brand (Audi production AI, Porsche Engineering methods), with AI Lab and the ML Research Lab as exploratory outposts. There is no sign of a single unified "VW AI stack" spanning brands. [Synthesis]

Reading between the lines

Like every platform vendor in this study, VW rents its foundation models (Azure AI for Copilot). Its proprietary asset is data and process access — weld signals, paint parameters, decades of tooling know-how — and its co-development seats. The Braunschweig knowledge system shows VW understood the value of mining its own engineering data before modern ML existed; that pipeline has simply never (publicly) been upgraded from statistics to learning. [Inference]

AI Technologies

In one list: generative AI / LLMs (Codebeamer Copilot on Azure — co-built, model rented); GANs (FelGAN, in-house); computer-vision / signal ML (spot-weld AI, Weld Splatter Detection); anomaly detection (ProcessGuardAIn); reinforcement learning + neural surrogates (Porsche Engineering crash work); virtual twins (Dassault platform); edge/cloud production infrastructure (DPP, EC4P); fundamental ML research (Munich lab: probabilistic deep learning, robotics, quantum ML). [Evidence + Synthesis]

Explicitly not AI but adjacent and load-bearing: knowledge-based CAM automation and statistical process standardisation in tool engineering; additive manufacturing (3D-printing centre 2018; binder jetting with HP + Siemens 2021).41415

Research Publications

Papers / published methods

  • Porsche Engineering publishes readable method articles: RL for crash structures and restraint systems (Feb 2025, −80% FEM loops; explainable-AI post-processing), and LLMs in vehicle development (2025). [Evidence] 1213
  • The Volkswagen Group Machine Learning Research Lab (Munich, led by Prof. Patrick van der Smagt since ~2016) publishes fundamental ML research — probabilistic deep learning, time-series models, robot learning, world models, quantum ML — with publications through 2025. None of it touches moulding or manufacturing of plastic parts. [Evidence] 1617
  • Schneider, T., "Automated acquisition of experience-based manufacturing expertise in tool and mould engineering," doctoral thesis, 2011 — the intellectual foundation of Braunschweig's tool-engineering knowledge system. A pre-deep-learning, statistics-on-CAM-data approach — evidence that VW's mould organisation has a data-mining culture awaiting modern methods. [Evidence] 4

Patents / Standards

Not surveyed in depth for this report (OEM patent portfolios are vast and mostly vehicle-technology); no moulding-AI patent was surfaced by the searches performed. [Evidence of absence — shallow search]

Open Source

Thin, and nothing moulding-related. The Munich ML Research Lab is described in conference-circuit profiles as publishing openly, and its leader co-founded the etami ethics-in-AI initiative; its publication list is public. No VW open-source release relevant to moulded-parts engineering was found. [Evidence — thin]1617

Engineering Service / Platform Mapping

  • ALM: Codebeamer (PTC) — group-wide, strategic. [Evidence] 5
  • Engineering platform (announced): Dassault 3DEXPERIENCE for VW, Audi, Porsche brands — cloud, virtual twins. [Evidence] 8
  • CAD estate: the Dassault release does not name the legacy CAD/PLM being used or replaced; VW's historical CATIA-centred, partly home-grown CAD/PLM landscape is widely assumed but was not primary-verified here — see Appendix C. [Evidence of absence]
  • CAM (tooling): OPEN MIND hyperMILL in Braunschweig component tool engineering (13+ years as of the 2016 case study). [Evidence] 4
  • AI cloud: Microsoft Azure (Codebeamer Copilot). [Evidence] 6
  • Additive: HP (binder jetting) + Siemens (AM software). [Evidence] 15

Engineering Intelligence Stack Mapping

  1. Intent — strong and AI-assisted for software (Codebeamer + Copilot drafting requirements/tests); conventional for mechanical parts. [Evidence] 6
  2. Knowledge — the standout irony: Braunschweig built a systematic engineering-knowledge pipeline (CAM data-mining, part families, rules) in the 2010s — exactly the Knowledge layer this study cares about — but it is statistical, static, and was never (publicly) connected to modern AI. [Evidence + Synthesis] 4
  3. Reasoning — emerging in research (Porsche RL surrogates for crash; Dassault virtual-twin environment announced); absent for moulding physics. [Evidence] 128
  4. Execution — CAM automation generates NC programs push-button in the mould shop (deterministic); production execution increasingly AI-monitored (welds, paint). [Evidence] 411
  5. Feedback — the strongest genuine learning loop found at any company in the automotive OEMs group so far: production sensor data (weld signals) feeds trained models that adjudicate quality in-line. But the loop lives in the body shop and paint shop — no equivalent exists (publicly) for moulding defects, warpage or scrap. [Evidence + Evidence of absence] 1011

Strengths

  • The only OEM found co-building an engineering-AI product (Codebeamer Copilot with PTC + Microsoft) — a seat at the vendor's design table, not just a licence. [Evidence] 6
  • Owns the full moulded-parts chain in-house — part design, mould design, mould build, moulding — a data position most software vendors can only dream of. [Evidence] 34
  • Real, shipped production AI with real numbers (1.5 M welds/shift) and an audit-proof AI process developed with Fraunhofer — unusually mature AI governance for manufacturing. [Evidence] 10
  • Funded, board-level AI strategy (€1bn / €4bn / "No process without AI"). [Evidence] 1
  • A genuine research bench (Porsche Engineering methods; Munich ML lab). [Evidence] 1216

Weaknesses

  • AI has not reached the moulded-parts core — no AI in mould design, DFM, mould flow or plastics engineering anywhere in the public record. [Evidence of absence] 4
  • Fragmented by brand and unit — Audi, Porsche, Cariad, AI Lab, Components each run their own initiatives; the €1bn programme is an umbrella over 1,200 mostly unnamed applications. [Synthesis] 1
  • Foundation models and platforms are rented (Azure, PTC, Dassault) — VW's co-development influence is real but the IP sits with vendors. [Evidence + Inference] 68
  • The tooling organisation's knowledge system is a decade old in its public form (2011 thesis, 2016 case study) — no public sign of refresh. [Evidence — dated] 4

Current Gaps (largely manual today)

Within VW's own walls, on public evidence: mould design and DFM judgement in Braunschweig remain expert work supported by deterministic CAM rules; moulding process setup and defect diagnosis have no public AI equivalent to the weld AI; no geometry-retrieval or design-reuse system over VW's decades of mould designs is publicly known; and moulding outcome data (scrap, warpage, rework) has no published learning loop. [Evidence of absence + Inference]

Future Direction

  • Execute the €1bn programme: DPP-connected plants, the Dassault AI-powered engineering environment, and the 36-month development cycle. [Evidence] 1
  • Scale the Audi production-AI portfolio to series across sites (WSD to six Ingolstadt plants; ProcessGuardAIn Q2 2026). [Evidence] 11
  • Continue the co-development posture (Copilot beta → production; deeper Dassault integration). [Inference]
  • My read: the weld-AI playbook — in-line sensors + trained models + audit-proof process, built with institutes — is a template VW could point at its injection-moulding machines and mould shop next. Nothing public says it is doing so. [Inference]

Relevance to Automotive Moulded Parts

Capability VW position AI?
Plastic Part Design In-house (brands + Components) No public AI
Surface Design (Class A) In-house studios No public AI (FelGAN is ideation for rims, not Class-A surfacing)
CAD Automation Platform being renewed (3DEXPERIENCE announced) Announced, not evidenced8
DFM In-house, expert + rules No AI
Tool Design In-house mould shop, Braunschweig No AI — knowledge-based CAM4
Mould Flow Internal (tools unspecified publicly) No public AI
Manufacturing Engineering Strong (DPP, EC4P, additive) Infrastructure-AI, not moulding1115
Quality Strong in body/paint shops Yes — but welding/painting, not moulding10
Engineering Knowledge Reuse Pioneering but frozen (CAM data-mining, 2010s) Statistical, pre-AI4

Critical finding. Volkswagen is simultaneously (a) the most AI-committed OEM found in this study, (b) an owner-operator of one of the industry's great in-house mould shops, and (c) a company in whose entire public AI record — €1 billion programme, 1,200 applications, co-developed products — the words "injection moulding," "mould design" and "plastics" never appear. The moulded-parts middle of its workflow is served by deterministic CAM automation and a 2011-vintage statistical knowledge method that VW, to its credit, has never rebranded as AI. [Evidence of absence + Synthesis] 41

Net. For this study, VW is the demand-side mirror of the vendor reports: the vendors don't sell moulding AI, and the biggest, most willing buyer isn't building it either — while possessing exactly the labelled outcome data (its own scrap, rework and mould-revision history) that the field lacks. [Synthesis]

Key Takeaways

  1. VW is the only OEM found co-building an engineering-AI product: Codebeamer Copilot with PTC + Microsoft (Dec 3, 2024; Azure AI; beta early 2025). [Evidence] 6
  2. Its buy/build/co-develop pattern is consistent: buy platforms (PTC, Dassault, Azure), build data-side applications (FelGAN, weld AI), co-develop where it wants influence (Copilot, binder jetting). [Synthesis]
  3. The AI strategy is board-level and funded: €1bn by 2030, €4bn targeted by 2035, 1,200+ applications, "No process without AI." [Evidence] 1
  4. Its most convincing shipped engineering AI is production quality (1.5 M spot welds/shift), built with research institutes and now scaling. [Evidence] 1011
  5. No AI reaches VW's own mould, DFM or plastics engineering — the study-defining negative. [Evidence of absence] 4
  6. VW's mould shop instead runs on honestly-labelled knowledge-based CAM automation (−35% processing time; NC programs in minutes) from the 2010s. [Evidence] 4
  7. VW owns the moulded-parts data chain end-to-end (design → mould → moulding → outcomes) — the exact labelled-data position the whole AI-DFM field lacks — and publicly does nothing with it. [Synthesis]
  8. Vehicle-development AI ambitions run through Dassault's platform (Feb 4, 2025) — VW rents the engineering-AI layer just as vendors rent LLMs. [Evidence] 8
  9. Porsche Engineering's RL crash work (−80% FEM loops) shows the Group can do surrogate-model engineering AI when it chooses — making the moulding silence a choice, not an inability. [Evidence] 12

References

Primary Sources — Volkswagen Group / brands

  • VW Group invests in AI (€1bn; "No process without AI"), Sep 9 2025 — https://www.volkswagen-group.com/en/press-releases/boosting-innovation-reshaping-mobility-volkswagen-group-invests-in-ai-19852
  • VW Group establishes AI company ("AI Lab"), Jan 31 2024 — https://www.volkswagen-group.com/en/press-releases/volkswagen-group-establishes-artificial-intelligence-company-18105
  • VW Group Components Plant Brunswick (plant profile) — https://www.volkswagen-newsroom.com/en/volkswagen-group-components-plant-brunswick-6741
  • Audi: FelGAN inspires new rim designs, Dec 12 2022 — https://www.audi.com/en/press-releases/felgan-inspires-new-rim-designs-15097
  • Audi: AI roll-out for spot-weld quality control, Jun 30 2023 — https://www.audi.com/en/press-releases/audi-begins-roll-out-of-artificial-intelligence-for-quality-control-of-spot-welds-15443
  • Audi: scales up AI in production, Jan 27 2026 — https://www.audi.com/en/press-releases/audi-scales-up-deployment-of-artificial-intelligence-in-production-17002
  • Porsche Engineering: "AI agent on a crash course," Feb 19 2025 — https://newsroom.porsche.com/en/2025/innovation/porsche-engineering-ai-agent-on-a-crash-course-38631.html
  • Porsche Engineering: LLMs in vehicle development, 2025 — https://newsroom.porsche.com/en/2025/innovation/porsche-engineering-large-language-models-38277.html
  • VW toolmaking opens 3D-printing centre (Group release via Porsche Holding newsroom), Dec 11 2018 — https://www.porscheholding-newsroom.at/pressemeldungen/detail/2018-12-11-produzieren-fuer-die-zukunft-volkswagen-werkzeugbau-eroeffnet-hochmodernes-3d-druck-zentrum

Partner / vendor primary sources

  • PTC strategic supplier to VW Group (Codebeamer), Oct 31 2023 — https://www.prnewswire.com/news-releases/ptc-becomes-strategic-supplier-to-volkswagen-group-with-codebeamer-alm-solution-301972685.html
  • Codebeamer Copilot with Microsoft + VW Group, Dec 3 2024 — https://www.prnewswire.com/news-releases/ptc-partners-with-microsoft-and-volkswagen-group-to-develop-codebeamer-generative-ai-copilot-302320888.html
  • Microsoft customer story: VW engineers + Copilot in Codebeamer — https://www.microsoft.com/en/customers/story/24120-volkswagen-microsoft-copilot
  • Dassault Systèmes × Volkswagen Group (3DEXPERIENCE), Feb 4 2025 — https://www.3ds.com/newsroom/press-releases/dassault-systemes-and-volkswagen-group-implement-3dexperience-platform-optimize-vehicle-development
  • OPEN MIND success story: "Knowledge with a system" — VW Braunschweig component tool engineering (PDF, April 2016) — https://www.hummingbird-systems.com/fileadmin/user_upload/pdf/references/automotive/cus-automotive-vw-volkswagen-knowledge-open-mind-en.pdf

Secondary Sources

  • 3D Printing Industry: VW binder jetting with HP + Siemens, Jun 19 2021 — https://3dprintingindustry.com/news/volkswagen-to-adopt-binder-jet-3d-printing-for-car-part-production-with-hp-and-siemens-191760/
  • Machine Learning Research Lab publications (van der Smagt) — https://argmax.org/
  • AI for Good speaker profile, Patrick van der Smagt (VW Group AI research) — https://aiforgood.itu.int/speaker/patrick-van-der-smagt/

Research Papers

  • Schneider, T.: Automated acquisition of experience-based manufacturing expertise in tool and mould engineering, doctoral thesis, 2011, Logos Verlag (cited in the OPEN MIND case study above).

Appendix A — Timeline

ML Research Lab Munich (~2016) → Wolfsburg 3D-printing centre (Dec 2018) → binder jetting with HP + Siemens (Jun 2021) → FelGAN (Dec 2022) → spot-weld AI roll-out (Jun 2023) → Codebeamer group adoption (Oct 2023) → AI Lab founded (Jan 2024) → Codebeamer Copilot co-development (Dec 2024) → Dassault 3DEXPERIENCE partnership (Feb 2025) → Porsche RL crash publication (Feb 2025) → €1bn AI programme (Sep 2025) → Audi production-AI scale-up (Jan 2026).

Appendix B — Glossary

  • Volkswagen Group Components — VW's in-house components business unit (Technology division): dozens of plants making axles, batteries, seats, plastic parts and the tools/moulds for them.
  • Component tool engineering — the Braunschweig mould-and-tool shop (~700 people) building injection moulds, die-cast moulds and press tools.
  • ALM — application lifecycle management: requirements, tests and releases for software (Codebeamer). Distinct from PLM (mechanical product data).
  • Codebeamer Copilot — generative-AI assistant for requirements/test-case drafting in Codebeamer; co-developed by PTC, Microsoft and VW Group.
  • GAN — generative adversarial network: two neural networks (generator vs. discriminator) trained against each other until outputs look real (FelGAN).
  • Reinforcement learning (RL) — ML where an agent learns by reward from trial actions (Porsche crash-structure work).
  • Surrogate model — a neural network trained on simulation runs that then predicts new results in seconds.
  • Binder jetting — an additive-manufacturing process printing metal parts layer-by-layer with a binder; industrialised by VW with HP and Siemens.
  • CAM — computer-aided manufacturing: generating machine toolpaths (NC programs) from CAD geometry; hyperMILL is the CAM system in VW's tool shop.
  • DPP — VW's Digital Production Platform connecting 40+ plants.
  • DFM — design for manufacturability: checking a design against the rules that make it mouldable (draft angles, wall thickness, undercuts…).

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

  1. "VW.AI" (from the research seed) could not be verified as a distinct initiative; the verified entity is AI Lab GmbH (Jan 31, 2024). A Munich "Data:Lab" is often mentioned in secondary accounts as a 2013-era predecessor but was not primary-verified here.
  2. VW's legacy CAD/PLM estate (historically CATIA-centred with home-grown PLM systems) is widely assumed in industry commentary but was not primary-verified; the Feb 2025 Dassault release names no legacy systems.
  3. The Braunschweig tool-engineering details (700 employees, injection moulds for bumpers/instrument panels, −35%/2h→minutes figures) come from a vendor case study dated April 2016 — strong on specifics, but a decade old; current headcount and toolchain may differ.
  4. A Springer/Kunststoffe article ("The New Dimension in Volkswagen Plastics Processing," ~110 injection-moulding machines, 32,000 kN Battenfeld press) surfaced in search but its date could not be established — excluded from the body.
  5. The Machine Learning Research Lab's current organisational status inside VW (and whether it survived 2024–26 restructuring) rests on conference-profile and publication-page evidence only.
  6. Group revenue/headcount figures were not primary-verified and are omitted from the body.
  7. Absence of AI in Group Components is an absence-of-evidence finding from English/German web searching; internal pilots would not be visible.

Executive Summary

  • Who they are. Volkswagen Group is one of the world's largest automakers — a multi-brand group (Volkswagen, Audi, Porsche, ŠKODA, SEAT/CUPRA, Bentley, Lamborghini, commercial vehicles) with its own software arm (Cariad) and, crucially for this study, a huge in-house components and toolmaking organisation (Volkswagen Group Components) that designs and builds its own injection moulds and moulds its own plastic parts. [Synthesis] 34
  • Why they matter here. VW is the clearest example found of an OEM co-building an engineering-AI product rather than just buying one: it co-develops Codebeamer Copilot — a generative-AI assistant for requirements and test cases — together with PTC and Microsoft (announced December 3, 2024). No other OEM in this study has a comparable co-development on the record. [Evidence] 6
  • The strategy is loud and funded. In September 2025 VW announced up to €1 billion of AI investment by 2030, targeting up to €4 billion in efficiency gains by 2035, with 1,200+ AI applications already running and the slogan "No process without AI." AI is explicitly aimed at cutting the vehicle development cycle to 36 months (~25% faster). [Evidence] 1
  • The central honest finding. After deliberate searching, no public evidence was found of AI reaching VW's own mould, tooling or plastics engineering — the Braunschweig component tool engineering unit (~700 people building injection moulds for bumpers and instrument panels) runs on knowledge-based CAM automation and statistical data-mining dating from the 2010s, which VW itself honestly calls "knowledge management," not AI. The world's most AI-ambitious car company has not pointed its AI at its own mould shop — at least not publicly. [Evidence of absence] 4

  1. "Boosting innovation, reshaping mobility: Volkswagen Group invests in AI," Sep 9 2025 — https://www.volkswagen-group.com/en/press-releases/boosting-innovation-reshaping-mobility-volkswagen-group-invests-in-ai-19852 ↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  2. "Volkswagen Group establishes artificial intelligence company" (AI Lab), Jan 31 2024 — https://www.volkswagen-group.com/en/press-releases/volkswagen-group-establishes-artificial-intelligence-company-18105 ↩↩↩↩↩

  3. Volkswagen Group Components — Plant Brunswick (profile; ~6,000 employees, chassis/steering/battery systems) — https://www.volkswagen-newsroom.com/en/volkswagen-group-components-plant-brunswick-6741 ↩↩↩↩↩

  4. OPEN MIND Technologies, "Knowledge with a system" — success story on Volkswagen Braunschweig component tool engineering (~700 employees; injection moulds for bumpers/instrument panels; statistical CAM-data knowledge management per Schneider 2011 thesis; −35% primary processing time via standardisation; NC programming 2 h → minutes), PDF dated April 2016 — https://www.hummingbird-systems.com/fileadmin/user_upload/pdf/references/automotive/cus-automotive-vw-volkswagen-knowledge-open-mind-en.pdf ↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  5. PTC becomes strategic supplier to Volkswagen Group with Codebeamer, Oct 31 2023 — https://www.prnewswire.com/news-releases/ptc-becomes-strategic-supplier-to-volkswagen-group-with-codebeamer-alm-solution-301972685.html ↩↩↩↩

  6. Codebeamer Copilot co-development with Microsoft + Volkswagen Group (Azure AI; beta early 2025), Dec 3 2024 — https://www.prnewswire.com/news-releases/ptc-partners-with-microsoft-and-volkswagen-group-to-develop-codebeamer-generative-ai-copilot-302320888.html ↩↩↩↩↩↩↩↩↩↩↩↩↩↩

  7. Microsoft customer story: Volkswagen engineering with Copilot and PTC Codebeamer — https://www.microsoft.com/en/customers/story/24120-volkswagen-microsoft-copilot ↩↩↩

  8. Dassault Systèmes and Volkswagen Group implement the 3DEXPERIENCE platform (VW, Audi, Porsche brands; "a main engineering and manufacturing platform"), Feb 4 2025 — https://www.3ds.com/newsroom/press-releases/dassault-systemes-and-volkswagen-group-implement-3dexperience-platform-optimize-vehicle-development ↩↩↩↩↩↩↩↩↩↩

  9. Audi: "FelGAN inspires new rim designs" (in-house GAN by Audi IT + Design), Dec 12 2022 — https://www.audi.com/en/press-releases/felgan-inspires-new-rim-designs-15097 ↩↩↩↩↩

  10. Audi begins roll-out of AI for quality control of spot welds (1.5 M welds/shift at Neckarsulm; with DGQ, Fraunhofer IAO/IPA; roll-out to Brussels, Emden, Ingolstadt), Jun 30 2023 — https://www.audi.com/en/press-releases/audi-begins-roll-out-of-artificial-intelligence-for-quality-control-of-spot-welds-15443 ↩↩↩↩↩↩↩↩↩↩

  11. Audi scales up deployment of AI in production (WSD to series; ProcessGuardAIn Q2 2026; AI dryer control; EC4P removes 1,000+ industrial PCs), Jan 27 2026 — https://www.audi.com/en/press-releases/audi-scales-up-deployment-of-artificial-intelligence-in-production-17002 ↩↩↩↩↩↩↩↩↩↩↩

  12. Porsche Engineering, "AI agent on a crash course" (reinforcement learning + surrogates; −80% FEM calculation loops; crash sims up to 72 h), Feb 19 2025 — https://newsroom.porsche.com/en/2025/innovation/porsche-engineering-ai-agent-on-a-crash-course-38631.html ↩↩↩↩↩↩↩↩↩↩

  13. Porsche Engineering on large language models in vehicle development, 2025 — https://newsroom.porsche.com/en/2025/innovation/porsche-engineering-large-language-models-38277.html ↩↩↩

  14. "Produzieren für die Zukunft: Volkswagen Werkzeugbau eröffnet hochmodernes 3D-Druck-Zentrum," Dec 11 2018 (Group release via Porsche Holding newsroom) — https://www.porscheholding-newsroom.at/pressemeldungen/detail/2018-12-11-produzieren-fuer-die-zukunft-volkswagen-werkzeugbau-eroeffnet-hochmodernes-3d-druck-zentrum ↩↩↩

  15. 3D Printing Industry: "Volkswagen to adopt binder jet 3D printing for car part production with HP and Siemens" (Wolfsburg; 100,000 parts/year target by 2025), Jun 19 2021 — https://3dprintingindustry.com/news/volkswagen-to-adopt-binder-jet-3d-printing-for-car-part-production-with-hp-and-siemens-191760/ ↩↩↩↩↩↩

  16. Machine Learning Research Lab publications page (Patrick van der Smagt; publications 1990–2025 incl. world models, robot learning) — https://argmax.org/ ↩↩↩↩

  17. AI for Good speaker profile: Patrick van der Smagt, Director of AI research, Volkswagen Group (Machine Learning Research Lab, Munich) — https://aiforgood.itu.int/speaker/patrick-van-der-smagt/ ↩↩↩

  18. On the honest-AI test (learn from data vs. compute from equations/rules) and why deterministic optimization/rule systems are not AI — see the concept note What is Optimization technology. VW's Braunschweig knowledge system is the rare case of a company applying the honest label itself: it calls its statistical CAM-data mining "knowledge management," not AI. ↩↩

  19. OPEN MIND Technologies — Wessling, Germany — https://www.openmind-tech.com ↩

  20. Fraunhofer (incl. ITWM, IAO, IPA) — Munich, Germany (society HQ) — https://www.fraunhofer.de ↩