3.3.3 Automotive OEMs¶
Automotive OEMs are the carmakers, and in this study they sit upstream of the moulding pain. An OEM obviously has automotive presence and obviously ships moulded parts, so the vendor-style tests do not discriminate here. The question that does is narrow: is there public, dated evidence of an OEM applying AI to moulded-parts engineering, meaning plastic-part design, mould-flow, moulding DFM, or tooling, as opposed to AI for autonomy, in-car assistants, or the factory?
Seven OEMs were assessed, three in full and four as scope notes. The answer is consistent, and it is a negative one: no example of an OEM deploying AI in moulded-parts engineering was found; the nearest case — Hyundai's academic Bayesian-optimisation work — is classical search under the honest-AI test, even though these same OEMs publish freely about AI everywhere else in the business. [Evidence of absence] That silence is the most important thing this section establishes, and it is turned into a go-to-market direction for anyone pursuing moulding AI in the Conclusion.
3.3.3.1 Selection criteria¶
For OEMs the screening question differs from the vendor tests: an OEM obviously has automotive presence and obviously ships moulded parts. The question is whether any public, dated evidence shows the OEM applying AI to moulded-parts engineering (plastic-part design, mould-flow, moulding DFM, or tooling). An OEM earns a full report when that evidence exists, or when its in-house moulding and engineering-AI depth make the negative finding itself worth a full treatment. [Synthesis]
3.3.3.2 Companies studied in detail¶
| # | Company | One-line role |
|---|---|---|
| 01 | BMW Group | In-house injection moulding at Landshut; among the most AI-forward European OEMs: a Mistral "Large Industry Model" being trained on >1 PB of crash-simulation data. |
| 02 | Volkswagen Group | 75+ years of in-house toolmaking; co-built PTC Codebeamer Copilot (requirements AI) for vehicle software. |
| 03 | Hyundai Motor Group | Deep Korean moulding base; Bayesian tailgate-rib moulding-optimisation research with KAIST and toolmaker Q-JEN. |
3.3.3.3 Considered but not studied in detail¶
Four OEMs with enormous moulded-parts consumption were researched, and all four fail the same test: no public, dated example of AI applied to moulded-parts engineering, each in an instructive way. Each has its own short scope note (steelman, verdict, what would change our mind).
| Company | Why no full report |
|---|---|
| Toyota | Real generative-AI design research (TRI drag-guided diffusion), but styling/aero from the research arm; nothing in production part or moulding engineering. |
| Mercedes-Benz | In-car GenAI (MBUX) and factory digital twins (Omniverse) sit either side of part engineering; no moulded-parts AI. |
| Tesla | Confirmed CATIA/3DEXPERIENCE shop; visible AI is FSD/Dojo/Optimus. No public evidence of AI in design engineering (a low-visibility verdict). |
| General Motors | Engineering-AI evidence is old (2018 generative design, metal) or classical (Digimat); fresh 2025 NVIDIA AI aimed at factories and autonomy. |
The section's core finding — that OEM demand for moulding-AI is indirect, and the go-to-market compass that follows (sell to the Tier-1s and toolmakers, treat the OEM logo as a lighthouse) — is drawn together with the other four groups in Insights from DFM Companies → Conclusion.
3.3.3.4 Things to watch¶
The finding above is a 2026 snapshot. Three developments would move it, and are worth keeping in view while reading the three company reports.
- Whether BMW's Large Industry Model reaches moulded parts. BMW's Mistral model is being trained on crash-simulation data today. If that same in-house capability is ever pointed at plastic-part design or mould-flow, it would be the first OEM exception to the silence. [Inference]
- Whether Hyundai's research reaches production. The Bayesian tailgate-rib moulding-optimisation work with KAIST and Q-JEN is real and dated, but sits in research. Moving it into a production part programme would change the verdict. [Inference]
- Whether factory digital twins drift upstream. OEM AI budgets today land in the factory (Mercedes/Omniverse) and in autonomy. A shift of that investment upstream into part and mould engineering would signal the pain is finally being felt where design happens. [Inference]